Methodology document for the WHO e-atlas of disaster risk. Volume 1. Exposure to natural hazards Version 2.0
Heat wave hazard modelling
Dr Zine El Abidine El Morjani Dr Soufiane Idbraim
Taroudant poly-disciplinary faculty of the Ibn Zohr University of Agadir, Morocco
Last Update: January 2011
1
Cataloguing-in-Publication Data Methodology document for the WHO e-atlas of disaster risk. Volume 1. Exposure to natural hazards Version 2.0: Heat wave hazard modelling 1. Disasters – Heat Wave. 2. Geographic Information Systems. 3. Risk Management. ISBN: 978-9954-0-5396-6
© Ibn Zohr University, 2011 All rights reserved. Publications of the Ibn Zohr University can be obtained from Ibn Zohr University, BP 32/S Agadir, Morocco (tel.: +212 528 22 71 25; fax: +212 528 22 72 60; e-mail:
[email protected]). Requests for permission to reproduce or translate Ibn Zohr publications – whether for sale or for noncommercial distribution – should be addressed to Ibn Zohr University, at the above contact details. The designations employed and the presentation of the material in this publication do not imply the expression of any opinion whatsoever on the part of the Ibn Zohr University concerning the legal status of any country, territory, city of area or of its authorities, or concerning the delimitation of its frontiers or boundaries. Dotted lines on maps represent approximate border lines for which there may not yet be full agreement. All reasonable precautions have been taken by the Ibn Zohr University to verify the information contained in this publication. However, the published material is being distributed without warranty of any kind, either expressed or implied. The responsibility for the interpretation and use of the material lies with the reader. In no event shall the Ibn Zohr University be liable for damages arising from its use.
2
Acknowledgements The development of the methodology for the preparation of the heat wave hazard distribution map is the product of contributions by several institutions and individuals. The funding to conduct the research leading to the development of the protocol has been provided by the World Health organization (WHO). The elaboration of the models was carried out by Zine El Abidine El Morjani, Taroudant poly-disciplinary faculty of the Ibn Zohr University of Agadir, Morocco and Steeve Ebener, WHO Mediterranean Centre for Health Risk Reduction, Tunisia. The data collection and analysis, implementation of the models, and documentation of the methodology were carried out by Zine El Abidine El Morjani.
3
Contents Acknowledgements .................................................................................................................... 3 Preface........................................................................................................................................ 6 1.
Introduction ....................................................................................................................... 8
2.
Methodology ..................................................................................................................... 8 2.1 Choice of the method to be used to spatially distribute heat wave hazard................... 10 2.1.1
Selected simplified biometeorological indices................................................. 11
2.1.1.1 Heat index ........................................................................................................ 11 2.1.1.2
Humidex ....................................................................................................... 13
2.1.1.3
Net effective temperature (NET).................................................................. 13
2.1.1.4
Apparent temperature................................................................................... 14
2.2 Available daily surface meteorological elements at the weather station level............. 14 2.3 Selected method ........................................................................................................... 15 2.3.1
Extraction, preparation and pre-processing of the meteorological data........... 18
2.3.2
Calculation of the daily heat index using the NCDC formula (SHI) ............... 18
2.3.3
Calculation of the heat wave index for each period of 3 consecutive days...... 18
2.3.4
Calculation of the annual maximum heat wave index for a two, five, and ten years return period............................................................................................ 20
2.3.5
Identification of the independent variables and selection of the regression model................................................................................................................ 24
2.3.6
Spatialization of the annual maximum heat wave index over the region covered in this version of the e-atlas.............................................................................. 27
2.3.7
Classification of the annual maximum heat wave index distribution maps into the final heat wave hazard maps ...................................................................... 27
3. Implementation..................................................................................................................... 28 3.1 Required software and hardware.................................................................................. 28 3.2 Extraction, preparation and pre-processing of the meteorological elements ............... 30 3.2.1
Extraction and preparation of the meteorological data .................................... 30
3.2.2
Pre-processing of the meteorological data ....................................................... 31
3.3 Calculation of the daily heat index using the NCDC formula (SHI) ........................... 35 3.4 Calculation of the heat wave index and annual maximum heat wave index for each weather station and year of observation....................................................................... 37 3.5 Calculation of the annual maximum heat wave index for a two, five, height and ten years return period........................................................................................................ 39 4
3.5.1
Creation of weather station specific files ......................................................... 39
3.5.2
Application of the Gumbel frequency analysis ................................................ 41
3.5.2.1
Application of the Gumbel frequency analysis on all the weather stations . 42
3.5.2.2
Correction and/or adjustment of the original dataset for unusual observations .................................................................................................................... 45
3.6 Stepwise regression analysis ........................................................................................ 48 3.6.1
Preparation of the GIS layers containing the spatial distribution of the causal factors and dependant variable......................................................................... 49
3.6.1.1
Preparation of the distance to the nearest coastline layers ........................... 49
3.6.1.2
Preparation of the distance from the relative latitude/longitude layer ......... 50
3.6.1.3
Preparation of the distance to urban areas layer........................................... 51
3.6.1.4
Preparation of the mean elevation distribution layer ................................... 53
3.6.1.5
Preparation of the aspect layers.................................................................... 53
3.6.2
Integration of the annual maximum heat wave index figures into the weather stations location GIS layer ............................................................................... 54
3.6.3
Preparation of the stepwise regression analysis ............................................... 55
3.6.4
Application of the stepwise regression analysis............................................... 56
3.7 Spatialization of the annual maximum heat wave index for each return period .......... 57 3.8 Creation of the heat wave hazard distribution maps .................................................... 59 References and further reading ................................................................................................ 61 Annex 1. Steadman’s table for heat index °F (°C), Relative humidity (%) ............................. 68 Annex 2. Description of the NCDC daily meteorological elements dataset ............................ 69 Annex 3. Projection of a GIS layers into the metric projection system ................................... 72 Annex 4. Creation of a 300 km buffer around each climatic zone and clipping of the different layers for the regression analysis .................................................................................... 74 Annex 5. Final regression by climatic zone and return period................................................. 76 Annex 6. Metadata for the annual maximum heat wave index distribution layers (two, five and ten year return periods)............................................................................................. 81 Annex 7. Metadata for the heat wave hazard distribution layers (two, five and ten year return periods)............................................................................................................................ 84
5
Preface Being able to conduct geographically based risk assessment at the sub national level requires being in a position to spatially distribute all the elements reported in following conceptual formula1:
This process being very much driven by the type of hazard faced by the population and/or the key infrastructures in a given country the World Health Organization has been working, since 2006 on the development and improvement of an electronic atlas which could stimulate ministries of health and other health stakeholders to improve their disaster management capacity as well as serve as the entry point for conducting sub national geographically based risk assessments. The WHO e-atlas of disaster risk models the distribution of natural hazards and population’s exposure and provides baseline data and maps needed to advocate for resources to improve disaster preparedness; aid emergency response measures; and assist in identifying, planning and prioritizing areas for mitigation activities. The first version of the e-atlas published in 2008 covered the WHO Eastern Mediterranean Region (22 countries) and five natural hazards (flood, seismic [earthquake], landslide, heat and wind speed) and was distributed to more than 500 users. Encouraged by this success, working in close collaboration with the WHO Regions and taking advantage of the establishment of the Vulnerability and Risk Analysis and Mapping programme (VRAM), it was decided to publish a second version of the e-atlas that would, this time, also the 46 countries forming the WHO African Region as well as 32 countries of the WHO European Region (due to limited resources, this version of the e-atlas focuses on Central Europe only). Building on the successful collaboration established between the Taroudant polydisciplinary faculty of Ibn Zohr University, Agadir, Morocco and the VRAM, most of the models used in the first version of the e-atlas have been improved and heat replaced by heat wave, a current preoccupation of many ministries of health. In order to allow for any other region or country to also apply the models on their own it has been decided to document not only the research behind the models but also provide users with a protocol that would allow them to generate the final hazard distributions maps. The present series of methodology document is the result of this documentation.
1
Modified from: Office of the United Nations Disaster Relief Co-ordinator (UNDRO). Mitigating natural disasters. phenomena,
effects and options. A manual for policy makers and planners. New York, United Nations, 1991.
6
It is important to underline that the hazard distribution maps resulting from the application of these models are nevertheless only the first step of a process allowing countries to assess their risk at the sub national level. Analysing vulnerability and capacity require a process which is difficult to be applied at the level of a region for the following reasons: - availability of desegregated data - incompatibility of indicators from one country to an other - important differences in terms of health context between one country and another. WHO has therefore been looking at having the vulnerability, capacity and therefore indirectly risk analysis, conducted on a country by country basis. In this context, the VRAM is supporting Member States and partners to strengthen their capacity in order to conduct such analysis and have it presented in a manner such as the figure below.
The result of such analysis is then to be integrated in the country Disaster Risk Reduction (DRR) and Health Emergency Preparedness and Response Programmes (HEPRP) and serve, among other things, to build safer hospital, improve mass casualties’ management and help specialized units within health Organizations (including MoH) for public health planning. At the same time, the baseline data, information and maps collected or produced through the process can be used by health authorities and partners to take informed decisions in times of crises.
7
1. Introduction This document describes the methodology and protocol developed by the Taroudant polydisciplinary faculty of Ibn Zohr University in close collaboration with the VRAM and then used to generate and document maps presenting the spatial distribution of heat wave hazard for the WHO e-atlas of disaster risk, volume 1: exposure to natural hazards, Version 2.0. The methodology used to spatially distribute heat wave hazard is based on the following steps: 1. Calculating the daily heat index using Equation 1 on the basis of the mean daily temperature and dew point, 2. Calculating the average daily heat index for each period of 3 consecutive days and year of observation (1997-2008) using the EatlasClimMod 1.0 application. This is what we are referring to as the heat wave index, 3. Applying the Gumbel frequency analysis on the measures from point 2 to obtain the annual maximum heat wave index for different return period (two, five and then years) 4. Identification of the relevant parameters for each return period, and selection of the regression model to spatialize the annual maximum heat wave index using a stepwise regression analysis. 5. Spatial interpolation of the annual maximum heat wave index for each return period using the selected regression models. The methods and process presented in this document could be applied to other geographic areas provided that the analyses use geospatial data of similar or better quality and resolution.
2. Methodology Everyone agrees on the gravity of the effects of a heat wave on human health and mortality and also on the damage to agriculture, forests and water resources. It could be considered as the major cause of weather-related fatalities (Robinson, 2001). The most notable example in recent decades is the heat wave that occurred during summer 2003 in Europe (the most extreme in 500 years). According to the European 2003 Heat Wave Project, supported by the European Union, more than 70 000 additional deaths were recorded in 2003 in twelve countries in Europe. In France and Italy, mortality reached respectively 19 490 and 20 089 deaths (Robine et al., 2003, 2008). Other countries, including Germany, Portugal, Spain and the United Kingdom, were also affected. The European economic loss is estimated more than $13 billion. These reports may make the summer 2003 European heat wave the worst natural disaster of that year (Baldi et al., 2005). Current global climate changes are considered to be anthropogenic; particularly because of increasing greenhouse gas concentrations in the atmosphere, heat waves will increase in frequency, intensity and duration (WHO, 2009; Baccini et al., 2008; IPCC 2007; Meehl et al., 2007; Robinson, 2001). This will have profound effects on human health, comfort and
8
economic activities (Baccini et al., 2008; WHO, 2003b; Delworth et al, 1999). The impact of future climate change on heat waves is considered by Gosling et al. (2008). Another expression of anthropogenic climate but at local level is the urban heat island (UHI). Studies show that urban centres have temperature warming trends above those of more rural surroundings (Tamrazian et al., 2008; WHO, 2008; Patzert et al., 2007; Koppe et al., 2004) and urban heat island temperatures in general increase as population increases (Oke, 1987). Facing such a hazard, several projects have been conducted to propose operational forecast models (André et al., 2004), to study the impact on mortality (Hutter et al., 2007; Tan et al., 2007) and on the ecosystem (Garcia-Plazaola, 2008) and also to make comparisons and analyses of heat waves (Rebetez et al., 2008). Even if there is no universal definition of a heat wave (Souch and Grimmond, 2006; Meehl and Tebaldi, 2004; Robinson, 2001; WHO, 2009), a heat wave can be seen as an extended period of unusually high atmosphere-related heat stress, which causes temporary modifications in lifestyle (Robinson, 2001) and may have notable impact on human mortality, regional economies and ecosystems (Reid at al., 2009; Gosling et al., 2008; 2007; Hajat et al., 2006; Meehl and Tebaldi, 2004; WHO, 2003; Easterling et al, 2000). Operational definitions aiming at quantifying the duration and/or intensity of these extreme events can also be found in the literature. These are greatly depending on the countries and regions where they have been applied (Baccini et al., 2008; Gosling et al., 2008; Ebi and Meehl, 2007; Schär et al, 2004; Palecki et al., 2001; Hurth et al., 2000). Some of them are presented here: 1. EuroHEAT, a project co-funded by the European Commission Directorate-General for Health and Consumers with the aim to improve public health responses to weather extremes and to heat-waves in particular (WHO, 2009) came up with the following operational definition of the term “heat-wave” within nine European cities: “a period when maximum heat index and minimum temperature are over the 90th percentile of the monthly distribution for at least two days”. 2. The World Meteorological Organization defines a heat wave as when the daily maximum temperature of more than five consecutive days exceeds the average maximum temperature by 5°C (41°F), the normal period being 1961–1990, according to Frich et al. (2002). 3. In the United States, the National Weather Service (NWS) defines a heat wave as a period of at least two consecutive days during which the daytime high and night time low heat index values (apparent temperature) are exceeding the heat stress thresholds of 41°C (105°F) and 27°C (80°F), respectively (Robinson, 2001). 4. In the Netherlands, a heat wave is defined as period of at least five consecutive days in which the maximum temperature exceeds 25°C (77°F), including at least three days with a maximum temperature exceeding 30°C (86°F) (Huynen et al., 2001).. 5. In China, a hot day is defined when the daily maximum temperature exceeds 35°C (95°F) (Tan et al., 2007).
9
6. In Canada, a heat wave is defined by Environment Canada (2004) as three or more consecutive days in which the maximum temperature is greater than or equal to 32°C 7. Hajat et al. (2002) defines heat waves as periods of five consecutive days or longer when a smoothed 3-day moving average of temperature exceeded the 97th percentile of average temperature for the entire period 8. Beniston (2004) defines a heat wave as three successive days when the temperature exceeds the 90th percentile of summer maximum temperatures. 9. Gosling et al. (2007) defined heat waves as periods lasting three or more consecutive days when the daily maximum temperature is equal to or greater than the 95th percentile of summer maximum temperature over the whole period of record. The same literature review also highlighted the following important elements when it comes to spatially distribute heat wave: -
The impact of heat waves characterized by longer duration (more than four days) is 1.5–5 times higher than for short heat-waves (WHO, 2009). Temperatures that people from a hotter climate consider normal can be termed a heat wave in a cooler area if they are outside the normal climate pattern for that area (Robinson, 2001). people living in cities are likely to be at higher risk than rural dwellers because of the urban heat island (UHI) effect (WHO, 2009)
This literature review also confirmed that air temperature alone is not a good indicator of heat, reason why the first version of the WHO e-atlas of disaster risk (El Morjani et al, 2007) was already looking at spatially distributing the annual daily maximum heat index. The method used in this previous version of the e-atlas needed nevertheless to be re-evaluated to see if it would still be relevant when looking at spatially distributing heat wave hazard instead of daily heat and this over a much larger area and taking all the above into account.
2.1
Choice of the method to be used to spatially distribute heat wave hazard
If different approaches are being used to measure the heat stress placed on the human body (Biometeorological indices, heat budget models) the present work required to identify a method that could be applied over the 3 continents considered in this version of the WHO eatlas of disaster risk and this only on the basis of surface meteorological elements collected by weather stations. It has therefore been decided to focus on simplified biometeorological indices as these are requiring less data than other approaches and are those being the most largely used in the world by weather services or scientists. The following sections are therefore presenting the simplified biometeorological indices which have been considered as well as the availability of surface meteorological elements over the studied area. 10
2.1.1 Selected simplified biometeorological indices Please refer to Driscoll (1992) and Parsons (2003) for a summary description of these simplified indices as the next section is only presenting some of them. 2.1.1.1 Heat index The heat index, which is the most commonly used method for measuring heat, is an apparent temperature, or a measure of how hot it feels when observed ambient temperature and humidity are combined. This index, also referred to as the Stedman Heat Index (SHI), has been developed by Robert Steadman (1979, 1984) on the basis of physiological studies on evaporative skin cooling. The result of his studies is a table (Annex 1) which allows quantifying this index on the basis observed ambient temperature and humidity (the version of the heat index presented here neglects the effects of wind and radiation changes). It is important to note that the Stedman table is effective when the temperature is greater than 80ºF (26.7ºC) and relative humidity is at least 40%. Below these thresholds, the general practice is to (Browning, 2009): - Consider the observed ambient temperature as corresponding to the heat index between 51°F (10°C) and 80°F (26.7°C), - Use the wind chill correction formula on the observed ambient temperature when the ambient temperature is below 51°F (10°C) On the basis of the table generated by Steadman (Annex 1), a first formula has been developed by the US National Climatic Data Center (NCDC) (Equation 1). Indexheat =
with
–42.379 + 2.04901523T + 10.14333127rh – 0.22475541Tarh – (6.83783 × 10–3Ta2) – (5.481717 × 10–2rh2) + (1.22874 × 10–3Ta2rh) + (8.5282 × 10–4Tarh2) – (1.99 × 10–6Ta2rh2)
Equation 1
Ta: ambient temperature (°F) rh: relative humidity (%).
This formula was obtained through a complicated set of measurements, mathematically analysed by multiple regression, and produce a heat index with an error of ±1.3 °F compare to the original table. Calculation of the daily relative humidity rh for each weather station and day of observation can be done using the following formula: rh =
ρW × 100 ρ WS
Equation 2
where:
ρW: actual vapour pressure ρWS: saturated vapour pressure
ρ W = 6.11 × 10
7.5×Td 237.7 +Td
ρWS = 6.11 × 10 11
7.5×T 237.7 +T
Equation 3
Equation 4
with: T = daily air temperature (°C) Td = daily dew point (°C):
The dew point is defined as the temperature at which water vapour in the air becomes saturated and water droplets begins to form. Another simpler formula, also aiming at reproducing the values from Steadman table, was developed by Michelozzi et al. (2007) (Equation 2). AT = –2.653 + 0.994Temp + 0.0153Td2
with
Equation 5
AT: apparent temperature Temp: ambient temperature (°C) Td: dew point (°C).
This formula has for example been applied in the Assessment and prevention of acute health effects and weather conditions in Europe (PHEWE) project (Michelozzi et al., 2007; Baccini et al., 2008). As we have two different formulas for that particular index, a comparison between the results obtained through the implementation of both is presented in Table 1 in order to facilitate the selection of the method to be used in the context of the e-atlas project. Difference between Steadman table and Michelozzi 2.3125 0
Temp: ambient temperature (°C)
Td: dew point (°C)
rh: relative humidity (%)
Michelozzi formula NCDC formula Steadman table 1979
37 37 37
25 25 25
50 50
43.6875 46 46
Michelozzi formula NCDC formula Steadman table 1979
32 32 32
28 28 28
80 80
41.1502 44 44
2.8498 0
Michelozzi formula NCDC formula Steadman table 1979
44 44 44
23 23 23
30 30
49.1767 52 52
2.8233 0
Michelozzi formula NCDC formula Steadman table 1979
50 50 50
10 10 10
10 10
48.577 48 48
0.577 0
HI: heat index (°C)
Table 1. Comparison between the Michelozzi et al., 2007 formula (used in the PHEWE) and the NCDC formula (used in this work) and Steadman’s table (Annex 1)
Table 1 shows that equation 1 is more accurately reproducing the values reported in Steadman original table (Annex 1).
12
2.1.1.2
Humidex
The humidex is an index first used in 1965 by Canadian meteorologists. This index gives a measure of the amount of discomfort felt by the combined effect of temperature and humidity. In general, the Humidex values tend to be higher than the Steadman heat index values, except at the extreme end of the table (Annex 1), where they tended to be slightly lower. The humidex while used since 1965 has been defined as follow in 1979 based on the work of Masterton and Richardson (1979): humidex = Ta + h
Equation 6
With h = 0.5555(e – 10.0) e = 6.11exp(5417.7530 × ((1/273.16) – (1/Td))) e : water vapour pressure (hPa) The complete formula therefore reads as follow: 1 1 humidex = Ta + 0.5555 × 6.11exp(5417.7530 × (( ) - ( ))) − 10 . 273.16 Td Equation 7 With: - Ta: ambient temperature (°C) - Td: dew point (°K). 2.1.1.3
Net effective temperature (NET)
The net effective temperature (NET) was developed by the Hong Kong Observatory (Li and Chan, 2000) and combines the effect of air temperature, wind speed and relative humidity. The following formula is used to calculate the NET: NET = 37 – (37 – Ta)/(0.68 – 0.0014rh + 1/(1.76 + 1.4ws0.75)) – 0.29Ta(1 – 0.01rh)
with
Ta= air temperature (°C) ws = wind speed (m/s) rh = relative humidity (%).
13
Equation 8
2.1.1.4
Apparent temperature
The apparent temperature (AT), invented in the late 1970s, was designed to measure thermal sensation in indoor conditions. It was extended in the early 1980s to include the effect of sun and wind. The Australian Bureau of Meteorology use two forms of formula to calculate the AT based on approximations of the value provided by a mathematical model of heat balance in the human body. The first one includes the effects of temperature, humidity and wind: AT = Ta + 0.33e − 0.70ws − 4.00
Equation 9
The second version includes the effects of temperature, humidity, wind and radiation: AT = Ta + 0.348e − 0.70ws + 0.70Q/(ws + 10) − 4.25
Equation 10
where: Ta = dry bulb temperature (°C) e = water vapour pressure (hPa) [humidity] ws = wind speed (m/s) at an elevation of 10 metres Q = net radiation absorbed per unit area of body surface (W/m2)
2.2
Available daily surface meteorological elements at the weather station level
Daily climatic data collected at the weather station level are necessary for the application of any of the formulas mentioned in section 2.1. The most complete and freely available global dataset providing weather station level daily climatic data is the Global Summary of the Day (GSOD) dataset produced by the National Climatic Data Center (NCDC). Accessible from the internet (http://www7.ncdc.noaa.gov/CDO/cdoselect.cmd?datasetabbv=GSOD&countryabbv=&geore gionabbv= [Accessed December 15, 2010]), this dataset gives access to the following 18 surface meteorological elements for over 9000 stations: • • • • • • • • •
Mean temperature (.1 Fahrenheit) Mean dew point (.1 Fahrenheit) Mean sea level pressure (.1 mb) Mean station pressure (.1 mb) Mean visibility (.1 miles) Mean wind speed (.1 knots) Maximum sustained wind speed (.1 knots) Maximum wind gust (.1 knots) Minimum and Maximum temperature (.1 Fahrenheit) 14
• • •
Precipitation amount (.01 inches) Snow depth (.1 inches) Indicator for occurrence of: fog, rain or drizzle, snow or ice pellets, hail, thunder and tornado/Funnel Cloud
Historical data are generally available for 1929 to the present, but the period 1973-present is the most complete.
2.3
Selected method
Comparing the list of available surface meteorological elements presented in section 2.2 with the formulas presented in section 2.1 we can see that all of the method, except the apparent temperature one (section 2.1.1.4), could be applied in the context of the present work as the necessary climatic elements are available in the GSDO dataset. This being said, it was demonstrated in the previous sections that: - Michelozzi’s formula is less precise than the one developed by the NCDC (Table 1) - The Humidex tends to over or under estimate heat index depending on the observed apparent temperature (section 2.1.2) While both remaining formulas (NCDC and NET) are recommended by the World Meteorological Organization (WMO, 2004) to forecast heat stress, factors such wind and radiations are very much influenced by the immediate surroundings. For example, wind speed is reduced by the sheltering effect of belts of trees and solar radiation is affected by short-term localized phenomena such as cloudiness (Australian Bureau of Meteorology 2011 (ABN 92 637 533 532, http://www.bom.gov.au/info/thermal_stress/). In view of the above, the formula developed by NCDC (Equation 1) has therefore been used to estimate daily heat stress over the all area covered in the 2nd version of the WHO e-atlas of disaster risks. Because of the heterogeneity in definitions presented in chapter 2, passing from a measure of daily heat stress to the notion of heat wave is not an easy task and it has therefore been necessary to decide on: - the number of consecutive days that would be considered in the present work, - the measure that would finally be used to spatially distribute heat wave hazard in the context of the e-atlas. For the first one, the application developed in the context of this work, EatlasClimMod 1.0 (see section 3.1), allows the users to actually decide on the number of consecutive days he would like to consider. For the present work, this has been fixed to 3 consecutive days When it comes to the final measure, taking into account data availability (see section 2.2), it has been decided to spatially distribute the annual maximum heat wave index over a period of 3 consecutive days as the representation of heat wave hazard. This measure is obtained by following these steps: 1. Calculating the daily heat index using Equation 1 on the basis of the mean daily temperature and dew point, 15
2. Calculating the average daily heat index for each period of 3 consecutive days and year of observation (1997-2008) using the EatlasClimMod 1.0 application. This is what we are referring to as the heat wave index, 3. Applying the Gumbel frequency analysis on the measures from point 2 to obtain the annual maximum heat wave index for different return period (two, five and then years) The results from step 3 are then used to define a regression, on the basis of relevant parameters, to spatialize the annual maximum heat wave index as the representation of heat wave hazard. Using the mean temperature and dew point to measure the daily heat index and then the average daily heat index over 3 days as the measure of the heat wave index does potential result in an under estimation of the heat stress, and therefore heat wave hazard, over certain part of the area covered in this version of the e-atlas. Even though, the main purpose of the WHO e-atlas being to raise awareness of the public health sector in countries, the results obtained through this approach help identifying where a heat wave is most likely to take place in the near future and this with the flexibility to adjust the result to the operational definition used in a given country. The following section describes in more details each step of the method. Method which is itself illustrated in Figure 1.
16
Source: NCDC global surface summary of day data
Weather stations Extraction of the meteorological elements and pre-processing Temperature
Dew point Application of the NCDC formula (SHI)
Annual frequency of data
Daily heat index (HI) Specifying the window size of the heat wave in days (3, 4, 5…) Heat wave Index for each period of x consecutive days
Annual maximum heat wave indexes with their frequency of certainty Application of the Gumbel frequency analysis
Application of the threshold filters (annual frequency, number of years)
Annual maximum heat wave index for a two, five, eight and ten years return period Application of the stepwise regression approach Independent variables and regression model Application of the regression model Legend Data
Intermediate products
Final product
Annual maximum heat wave index distribution maps for a two, five and ten years return period Reclassification Spatial distribution of the intensity level of heat wave hazard for two, five and ten years return period
Figure 1. Methodology used for obtaining the spatial distribution of heat wave hazard over the region covered in this version of the e-atlas
17
2.3.1 Extraction, preparation and pre-processing of the meteorological data The daily mean temperature and dew point for the 1997-2008 period were mined for 3458 weather stations located within and around the countries of the study area, from the Global Summary of the Day (GSOD) dataset (see section 2.2). It is important to note that no data were available for Afghanistan, Angola, Burundi, Democratic Republic of the Congo, Djibouti, Eritrea, Ethiopia, Iraq, Islamic Republic of Iran, Malawi, Rwanda and Somalia. The pre-processing consisted then in arranging the data in a way that it could then be used in the EatlasClimMod 1.0 application.
2.3.2 Calculation of the daily heat index using the NCDC formula (SHI) The NCDC formula (Equation 1) has been used to generate daily heat index measures. This formula has been applied on each weather station and day of observation over the 1997-2008 period. In order for this formula to work correctly the relative humidity (rh) should be expressed as a percentage, not a fraction (for example: “65” and not “0.65”); As indicated before, Steadman approach is only effective when the temperature is greater than 80ºF (26.7ºC). In the context of the e-atlas, when the temperature is below this thresholds for a given weather station and day,, the ambient air temperature is being directly considered as the heat index. As a confirmation, using NOAA heat index calculator (http://www.hpc.ncep.noaa.gov/html/heatindex.shtml),we can observe that the different between the heat index and the ambient temperature is very small and leading anyway to the very low category in Table 6 (heat index < 80°F).
2.3.3 Calculation of the heat wave index for each period of 3 consecutive days Once the NCCD formula applied on each weather station and day of observation, the average heat index over a period of 3 consecutive days has been calculated. Table 2 present an example of such a calculation for the Sharorah weather station in Saudi Arabia.
18
Table 2. Mean heat index calculation of triplets of consecutive days for the weather station of Sharorah in Saudi Arabia (number 411360)
Date
19940223 19940224 19940225 19940226 19940227 19940228 19940301 19940302 19940303 19940304 19940305 19940306 19940307 19940308 19940309 19940310 19940311 19940312 19940313
Daily heat index (°F)
78.9626 77.1466 0 0 76.7119 75.2886 77.6049 75.0647 0 75.6268 77.3278 77.991 80.0941 79.368 77.2702 80.3279 85.7159 86.2328 82.06
Average heat index for three consecutive days (heat wave index) (°F)
77.9145 0 0 0 0 76.5351 75.9861 0 0 0 76.9819 78.471 79.151 78.9108 78.9887 81.1047 84.0922 84.6696 83.7727
In this table, the average for each triplet of consecutive days is calculated by scanning the database using a 3 days window. It is important to note that the heat wave index can’t be calculated over a 3 days period when there are gaps in the database. At the end of this process, the maximum heat wave index for each year of observation is stored for the next step in the process. In the case of the weather station of Sharorah (Table 2) the maximum heat wave index for 1994 would be of 84.6696 if the value would be limited to those presented in this table.
19
2.3.4 Calculation of the annual maximum heat wave index for a two, five, and ten years return period The fact that climate is changing has become increasingly clear over the past decades. The Intergovernmental Panel on Climate Change (IPCC) reports that heat waves have been more frequent and intense towards the end of the 20th century (IPCC, 2007), and are projected to continue to increase in frequency, intensity and duration worldwide (WHO, 2009; Baccini et al., 2008; IPCC, 2007; Meehl et al., 2007; Robinson, 2001), and an event such as the European heat wave in 2003 would be more common (Meehl et al., 2007). This event could result in future increases in heat-related morbidity and mortality (Reid et al., 2009; Baccini et al., 2008; Campbell-Lendrum and Woodruff, 2007; WHO, 2003b; Delworth et al., 1999). These climate changes are caused by the combined impact of growing human population and economic activities (WHO, 2003b). The IPCC (2000) has developed a series of 40 scenarios of plausible future trajectories for population growth, economic and technological development. Each scenario gives estimates of the greenhouse gas emission levels, and predicts the changes in in the average temperature for a given period relative to the baseline period (Campbell-Lendrum and Woodruff, 2007). For example, temperature may be estimated to increase by 0.54 °C (scenario B2; low emissions scenario), 0.84 °C (scenario A1B; middle emissions scenario) or 1.02 °C (scenario A2; high emissions scenario) in 2030, relative to the baseline period (WHO, 2009). In this work, it has been decided to not use the IPCC scenarios for spatializing the distribution of heat wave hazard because it gives a constant rate of temperature change relative to the baseline period by adding to each daily temperature (T) the corresponding projected average temperature change (∆T) in a single cell grid (∆T = 0.54 °C for scenario B2, ∆T = 0.84 °C for scenario A1B and ∆T = 1.02 °C for scenario A2, for example) while the estimation made through the present project shows that the evolution of the temperature is not homogenous over the all region covered by this version of the e-atlas. If in some areas the trends are indeed showing an increase in some places the estimations shows no variations and sometime even a decrease of the annual mean temperature in other areas. It has therefore been decided to apply the frequency approach on past trends observed to estimate future meteorological trends at the weather station level. This estimation is based on the use of a probability distribution function as directed by several authors (Fuller, 1914; Foster, 1935; Gumbel, 1941; Gumbel, 1942; Kite, 1977; Stolte and Dumontier, 1977; Gerard and Karpuk, 1979; Condie and Lee, 1982; Moin and Shaw, 1985; Stedinger et al., 1992; USACE, 1993; and Jones et al., 2005). A probability distribution function yields expected meteorological conditions over various time periods (return periods) in the future. This approach does not require a comprehensive understanding of meteorology or meteorological phenomena but examines the relationship between the past magnitude and frequency of occurrence of the phenomena in order to identify some statistical regularity between them. In effect, the past is extrapolated into the future. Frequently used probability distribution functions include Gumbel, lognormal, Pearson type 3, log Pearson type 3 and gamma. Despite the extensive literature on the topic, there is no preferred distribution function for the frequency analysis of meteorological data because each function has a unique set of advantages and disadvantages. The problem is complicated by the 20
necessity to evaluate meteorological data for return periods that exceed the length of the observed record. Because of their respective climatologic characteristics, this process has been applied separately on five zones (see section 2.3.5). The final map was created by aggregating, in a seamless way, the results obtained for each zone. In our context, the Gumbel extreme value distribution function (Gumbel, 1941; Gumbel, 1954; Gumbel, 1960; Landwehr et al., 1979; Vogel, 1986; Sarma, 1999; El Morjani, 2003; Koutsoyiannis, 2004, He et al., 2006, El Morjani et al;, 2007) was the most appropriate function because it seeks to identify the temporal distribution of extreme values for various return periods. Additionally, this probability distribution function (Equation 11) is one of the most widely used for extreme value prediction when analysing hydrological and meteorological data (Meylan and Musy, 1998): −e
−
x−a b
F ( x) = e Equation 11 with: F(x) = cumulative distribution function a and b = adjustment parameters; a is a location parameter and b is a scale parameter.
x−a with the reduced variate u, the cumulative distribution function becomes: b −u 1 Equation 12 F ( x) = e −e → u = − ln[− ln F ( x)] = − ln − ln1 − T 1 F ( X ) = 1 − and T = the return period. T
Replacing
with
Taking the weather station of Safita in Syria as example (Table 3), the extreme-value series for the annual maximum heat wave index is fitted to a Gumbel distribution through the following steps: 1. 2.
The annual maximum heat wave index series obtained in section 2.3.3 is ranked in increasing order. The empirical frequency is computed for each value using the Hazen formula r − 0.5 Equation 13 F= n
with r is the rank for each value and n is number of the years of record. 3. 4.
The Gumbel reduced variate u is calculated by applying equation 12. The mean and the standard deviation of the maximum heat wave and of the Gumbel reduced variate u are then calculated.
21
Table 3. Ranking of the annual maximum heat wave index, calculations of empirical frequency, Gumbel reduced variate and associated statistical parameters for the weather station of Safita (Syria)
Year
Annual maximum Rank heat wave index
Empirical frequency F(x) = (r – 0.5)/n
u = –ln[–lnF(x)]
1997
82.4377
1
0.041667
–1.1563
2005 2004 2003 2006 1999 2000 2008 2001 2002 1998 2007 Mean Standard deviation a b
83.7529 84.6323 85.2341 85.3107 86.2935 87.1099 87.91 88.2756 89.7226 92.7673 93.0261 87.2061
2 3 4 5 6 7 8 9 10 11 12
0.125 0.20833 0.29167 0.375 0.45833 0.54167 0.625 0.70833 0.79167 0.875 0.95833
–0.7321 –0.45019 –0.20876 0.019357 0.24826 0.48922 0.75501 1.0647 1.4541 2.0134 3.1568 0.55446
3.3355
1.2273
85.6992 2.7178
5. A graph is then generate to verify if the plotting of the annual maximum heat wave index versus the Gumbel reduced variate u shows a form in conformity with the Gumbel distribution, and the results are linear. If not, this graph allows the identification of outliers, which can be removed from the dataset. In the case of the Safita weather station, the graph shows an adequate fit to the Gumbel distribution and the absence of any outliers (Figure 2).
Figure 2. The annual maximum heat wave versus the Gumbel reduced variate distribution for the weather station of Safita in Syria
22
6. The statistical adjustments parameters of the distribution a and b are determined using the least-rectangles method (Table 3): Sx Su a = x − bu b=
Equation 14 Equation 15
With: Sx = standard deviation of the annual heat wave Su = Gumbel reduced standard deviation x = mean of the annual heat wave u = Gumbel reduced mean Note: the adjustment parameters will remain the same for each return period. 7. the annual maximum heat wave index value for the desired return period T (two, five and ten years) are calculated using the following statistical model (Table 4): 1 X T = a + bu = a + b − ln − ln1 − T
Equation 16
With XT = value of variate with a return period T. The computing process is divided into the following three steps, using a five year return period as an example: 1 a. Calculation of the non-exceedance probability F ( x) = 1 − . For T = 5 years, T F(x) = 0.8. b. Calculation of the Gumbel reduced variate using equation 12, u = 1.5 c. Computation of the annual maximum heat wave index using the linear relation reported in equation 16 (a and b are determined using equation 14 and equation 15: a = 85.6992 and b= 2.7178 → X5 = 85.6992 + (2.7178 × 1.5) = 89.7759). Table 4. Annual maximum heat wave index for the weather station of Safita (Syria) for two, five, and ten year return periods Return period (years) 2 5 10
Adjustment parameters a b 85.6992 2.7178 85.6992 2.7178 85.6992 2.7178
23
Annual maximum heat wave index 86.6953 89.7759 91.8151
2.3.5 Identification of the independent variables and selection of the regression model Spatial interpolation is widely used for translating irregular scattered meteorological data (data collected at discrete locations [i.e. at points]) into continuous data surfaces (rasters). The choice of interpolation method is especially important in the WHO Regions where meteorological data are sparse, and there are large value changes over short spatial distances. Additionally, the spatial density, distribution and spatial variability of sampling stations influence the choice of interpolation technique (MacEachren and Davidson, 1987). Given a set of meteorological data, researchers are confronted with a variety of stochastic and deterministic spatial interpolation methods to estimate meteorological data values at unsampled locations: •
•
deterministic estimation methods including inverse distance weighting (Legates and Willmont, 1990; Eischeid et al., 1995; Lennon and Turner, 1995; Willmott and Matsuura, 1995; Collins and Bolstad, 1996; Ashraf et al., 1997; Dodson and Marks, 1997) and spline methods (Eckstein, 1989; Hutchinson and Gessler, 1994; Hulme et al., 1995; Lennon and Turner, 1995; Collins and Bolstad, 1996) stochastic techniques including the kriging and cokriging techniques (Matheron, 1963; Hudson and Wackernagel, 1994; Collins and Bolstad, 1996; Hammond and Yarie, 1996; Holdaway, 1996; Ashraf et al., 1997, El Morjani, 2003) and polynomial regression (Myers, 1990; Collins and Bolstad, 1996; Benzi et al., 1997; Chessa and Delitala, 1997; Hargy, 1997; Vogt et al., 1997; Agnew and Palutikof, 2000; El Morjani, 2003; Li et al., 2006).
For a summary description of these methods, refer to Collins and Bolstad (1996) and El Morjani (2003). The characteristics of the data found for the region covered in this version of the e-atlas (low spatial data density, a high spatial variability and the absence of meteorological data for many countries) resulted in implausible outputs when applying the inverse distance weighted and kriging interpolation methods, more specifically as follows. The application of the inverse distance weighting method over a test area (Islamic Republic of Iran and Pakistan) generated specking or “birds eye” effects around the station locations, which was not plausible as the spatial variation for the annual maximum heat wave index was not following a regular trend. The application of the kriging technique over the same test area produced results that were inconsistent with the original data. Whatever the model used (spherical, exponential or Gaussian) the statistical cross-validation was not able to fit the theoretical spatial semivariogram. This might be because the density of weather stations is too low and the study area too large to support the use of the kriging interpolation method. It has therefore been necessary to find another model that produces results of good quality.
24
A literature review was carried out to first identify a set of variables that are significantly correlated to the heat wave index, namely: •
• •
• • •
•
Elevation (Z) is known to be a strong determinant of temperature and consequently heat wave index. On a global average, despite the seasonal and geographical variations, vertical lapse rates average 6.5 °C per 1000 m elevation in the free troposphere (Lutgens and Tarbuck, 1995). Mean elevation within a 3×3 pixel window of each cell (Z9) to measure the wider influence of elevation on temperature at any one location (He et al., 2006). Aspect (Asp) as a measure of the local climate effect (microclimate) that can be generated by the orientation of the slope. For example, as the sun‘s rays are from the west at the hottest time of day, in the afternoon, in most cases a west-facing slope will be warmer than a sheltered east-facing slope. This can have major effects on the distribution of vegetation that requires large quantities of moisture. Slope (Slp) as a factor determining the amount of solar radiation received and could have an impact in increasing the temperature and consequently the heat wave index (Agnew et al., 2000). Relative longitude (d_X) and latitude (d_Y) to account for large-scale gradients of the heat wave that can be observed over the region covered in this version of the e-atlas. Distance to the nearest coastline (d_Coast) to account for maritime influences on the heat wave index (Agnew et al., 2000). Distance to urban area (d_urb_pop) to account for the local urban heat island effect on the heat wave (Tamrazian et al., 2008; WHO, 2008, Patzert et al., 2007; Koppe et al., 2004; Oke,1987).
With the variables identified, to which their squares (i.e. d_urb_pop2, d_X2,… ) were also added, stepwise (back and forth steps) linear regression was used to identify their statistical significance, if any, and their relative contribution to the determination of the dependent variable (annual maximum heat wave index), thereby eliminating any insignificant variables. Because of their respective climatologic characteristics, this process has been applied separately on the following five zones: •
Zone 1: African continent
•
Zone 2: Middle East countries (Bahrain, Iraq, Israel, Jordan, Kuwait, Lebanon, Oman, Qatar, Saudi Arabia, Syrian Arab Republic, United Arab Emirates, West Bank and Gaza Strip, Yemen)
•
Zone 3: Afghanistan, Islamic Republic of Iran, Pakistan
•
Zone 4: Armenia, Azerbaijan, Georgia, Kazakhstan, Kyrgyzstan, Mongolia, Tajikistan, Turkmenistan and Uzbekistan
•
Zone 5: Albania, Austria, Belarus, Bosnia and Herzegovina, Bulgaria, Croatia, Czech Republic, Estonia, Greece, Hungary, Latvia, Lithuania, Moldova, Republic of, Montenegro, Poland, Romania, Serbia, Slovakia, Slovenia, former Yugoslav Republic of Macedonia, Turkey and Ukraine. 25
The final map was created by aggregating, in a seamless way, the results obtained for each zone. A stepwise linear regression analysis was performed separately for each of these zones and return periods using S-Plus software. The validation of each regression was carried out using R2 variance analysis as well as a detailed probability and residual analysis in order to identify the significant variables and therefore select the best regression model possible. As an example, the stepwise regression analysis for a two year return period applied on the 121 weather stations located in zone 2 gives the results shown in Table 5. Table 5. Regression model for the annual maximum heat wave index over zone 2 for a two year return period Variable (Intercept) d_X Z9 d_X2 d_Y2 d_urb_pop_2000002 Residual standard error Degrees of freedom Multiple R-squared F statistic Probability (F statistic)
Regression coefficient –1.1935559917 0.0414963292 –0.0156346046 –0.0000035280 –0.0000008424 –0.0000362393
Standard error 17.25999 0.00707 0.00078 0.00000 0.00000 0.00000
t-value –0.06915 5.86524 –20.03714 –5.01369 –6.38177 –7.55741
Probability Pr(>|t|) 0.94498 0.00000 0.00000 0.00000 0.00000 0.00000
3.36629 115 0.86 145.83142 0.0000
Four of the eight independent variables explain most of the variation of the annual maximum heat wave index for the two year return period (HW_2): The four variables (longitude [d_X] and latitude [d_Y] both expressed in kilometres (see section 3.6.1.2), mean elevation [Z9] in metres and distance to urban populations of more than 200 000 inhabitant [d_urb_pop_200000]) were used to derive the following regression equation for a 2 year return period: HW_2 = 0.0414963292*d_X – 0.0156346046*Z9 – 0.0000035280*d_X2 – 0.0000008424*d_Y2 – 0.0000362393*d_urb_pop_2000002 – 1.1935559917 Equation 15 This equation shows a positive correlation between [d_X] and the annual maximum heat wave index, and an inverse correlation with [Z9] and the square of [d_urb_pop_20000]. This means that the annual maximum heat wave index increases with longitude [d_X] and decreases with elevation [Z9] and distance from urban population [d_urb_pop_200000]. Slope and aspect do not appear in the regression equation because they do not significantly explain the variation of the maximum heat wave in zone 2. Eighty-six percent of the variance in extreme heat wave are explained by the four variables retained in the regression equation (R2 = 0.86). The model is considered valid and reliable 26
because of the strong correlation (R = 0.93) and the high degree of confidence that exists in the selected variables (very small probability [F statistic]). This is clearly statistically significant, so the true slope of the regression line is probably not zero.
2.3.6 Spatialization of the annual maximum heat wave index over the region covered in this version of the e-atlas Once the regression models had been found for each of the zones and return periods, they were applied separately to each zone before aggregating the maps to form one layer for the all region covered in this version of the e-atlas.
2.3.7 Classification of the annual maximum heat wave distribution maps into the final heat wave hazard maps
index
As indicated in the introduction of this document, the perception of heat varies from one country to another. This does not only lead to different operational definitions but also the difficulty to come up with a unique scale that would be relevant for the all region covered in this edition of the eatlas. As we nevertheless needed to come up with a unified scale for representing the result of the method applied in the context of the e-atlas, it has been decided to use the thresholds for heat stress forecasts with reference to human impact developed by the US National Weather Service (NWS). Table 6 provides the link between the different classes considered in the NWS classification and the five heat wave hazard intensity levels selected for this project (very low, low, medium, high and very high). Table 6. Correspondence between the US National Weather Service classification and the five heat wave hazard intensity levels Intensity level
Heat index Dangers
Category
Very high
130 °F or higher (54 °C or higher)
High
105–129 °F Sunstroke, muscle cramps, and/or Danger (41–54 °C) heat exhaustion likely. Heatstroke possible with prolonged exposure and/or physical activity.
Medium
90–105 °F (32–41 °C)
Sunstroke, muscle cramps, and/or Extreme heat exhaustion possible with caution prolonged exposure and/or physical activity.
Low
80–90 °F (27–32 °C)
Exercise more fatiguing than usual Caution
Very low
This level is not part of the table developed by the US National Weather Service
Heat stroke or sunstroke imminent Extreme danger
27
3. Implementation This chapter describes how the methodology presented in section 2 of this document has been implemented using the software listed in section 3.1. At the same time, this chapter serves as a user manual for the EatlasClimMod 1.0 application developed under Matlab 6.0. The names of the files are reported in bold in the text. A “*” following the file name indicates a data layer which is described in the Methodology and implementation process for generating the dataset, document that can also be found on the e-atlas DVD.
3.1
Required software and hardware
The implementation of the methods and processes presented in this document requires the following software under Windows NT 4 (Service Pack 5, or 6a), Windows 2000 (Service Pack 3 or 4) and Windows XP (all versions) and any more recent version of Windows: ¾ ArcView 3.x with the Spatial Analyst 1.1 extension; both developed by the Environmental Systems Research Institute, (ESRI) Inc., for the extraction of the data columns to be processed and the geospatial operations. ¾ The following publicly available scripts and extensions which are accessible directly in the e-atlas DVD (in the tools section) have also been used: • Grid Analyst (GridAnalyst.avx) • Compiled_Table_Tools.avx • XTools (Xtoolsmh.avx) • Grid and Theme Projector v.2 (grid_theme_prj.avx).
These scripts should be saved on the computer in the C:\ESRI\AV_GIS30\ARCVIEW\EXT32 and then uploaded as extensions in ArcView before starting the process presented in the following sections. ¾ Matlab 6.0 (or higher), developed by MathWorks. Matlab has been used for the development of the EatlasClimMod 1.0 because software such as Excel can’t handle the large files used in the context of the present work (tables of approximately 160 000 lines) and there was a need for a platform for programming. ¾ The EatlasClimMod 1.0 application. This application has been developed to calculate and estimate the different climatic variables, including annual maximum heat wave index for different return period, used in the context of this version of the WHO e-atlas. The codes of this application and the instruction file are all available in a zip file named EatlasClimMod.zip located in the tools section of the e-atlas DVD. This zip file contains the following functions which needs to be placed anywhere on your computer:
28
calcul_HeatIndex()
fct_CalculHeatIndexPerDay()
index_return()
calcul_HeatWave()
WaveModelling ()
save_liste_STNs()
delete_errorLine()
fct_MainProgram()
save_one_STN()
delete_null_line()
fct_Preprocessing()
check_succession()
fix_threshold()
Gumbel()
fct_rank()
GUI_HWI_Gumbel ()
All these functions are accessed by the user through a graphic interface which has been developed under Matlab 6.0. This interface as well as the different functions will be presented in the next sections. ¾ A text editor such as WordPad to display the results tables. ¾ S-Plus 6.0, developed by Insightful Corporation, to explore and identify statistically significant parameters and their relative contribution to the spatialization of the heat wave using a stepwise multiple regression.
The minimum and recommended hardware requirements for running Matlab and the EatlasClimMod 1.0 application are as follows: • • • •
Processor: Intel Pentium 3 and above 256 MB of RAM (512 MB or more is recommended) 600 MB of free hard drive space (1 GB is recommended) A colour graphics card and monitor (SVGA is recommended)
As a reference, EatlasClimMod 1.0 application has been used on the following computer configuration in the context of the WHO e-atlas project: • • •
processor Pentium Intel 4 (1.4 GHZ) 512 MB of RAM 2 Go of free hard drive space
It is therefore recommended, if possible to use a computer presenting these characteristics or better.
29
3.2
Extraction, preparation and pre-processing of the meteorological elements
The source of the meteorological elements used in this protocol is the Global Summary of Day (GSOD) dataset produced by the National Climatic Data Center (NCDC) (see section 2.2). The following sections describes how the elements needed for the application of the selected method, daily mean temperature and dew point have been extracted from this dataset to cover the 1997 - 2008 period.
3.2.1 Extraction and preparation of the meteorological data The process used to extract the daily mean temperature and dew point data is outlined by the following steps: 1.
Download and save the global summary of day data in ASCII format from: http://www7.ncdc.noaa.gov/CDO/cdo web site, after choosing the geographic region (Africa, Asia, Europe, Middle East) and the date range from 01/01/1997 to 31/12/2008.
2. Open each meteorological regional file in Microsoft Office Access and save them as a dBase IV (*.dbf) table using a specific naming convention (e.g. africa.dbf). These files will not open in Microsoft Office Excel because they contain more than 65 356 records. The fields names and their description can be found in Annex 2. 3. Only keep the columns that are needed for analysis as follows: a. b.
open africa.dbf, and choose C-Tables Tools>Delete Multi-Fields and select all the fields to be deleted (all except: Field 1: STN, Field 3: YEARMODA, Field 4: TEMP and Field 6: DEWP) repeat this step for the other regions.
4. Create a files for each year of observation (1997…2008) from africa.dbf as follows (these files are needed for the pre-processing presented in section 3.2.2): a. open africa.dbf and add a field called “date_string” in the attribute table by choosing the string type, in order to convert the date into string type b. select the header of the “date_string” column and click on the Calculate button c. type the following formula in the Calculator window: [YEARMODA].AsString d. add another new column on the right and call it “year” e. select the header of the “year” column and click on the Calculate button f. type the following formula in the calculator window: [date_str].Left(4). This field contains only the year of observation for each station
30
g. select the year = 1997 and save the table under year.txt (year = 1997, 1998, … , 2008) by selecting File>Export and Export Format = Delimited Text h. using WordPad, delete the first record containing Field 1, Field 3, Field 4, Field 6 of these files in order to have files with only numerical contents. This operation facilitates the processing in Matlab i. repeat steps a to h for the other years of observation and the other regions j. Put all the files created year.txt (year = 1997, 1998, … , 2008) into a specific folder for example named “Africa/data”, “Europe/data”, etc.
3.2.2 Pre-processing of the meteorological data The pre-processing of the climatic elements takes place using the EatlasClimMod 1.0 application under Matlab 6.0 (or higher). The following steps have to be followed in order to start the EatlasClimMod 1.0 application: 1. Run the MATLAB software. The window presented in Figure 3 will appear.
Figure 3. Matlab interface
2. Specify the path to the EatlasClimMod 1.0 application as the current directory using the browsing button
on the upper right side of the window.
3. Launch the EatlasClimMod 1.0 application, by going to the File>Open, select the GUI_HWI_Gumbel.m file as shown in Figure 4 and click on Open.
31
Figure 4. Open file window in MATLAB to run the EatlasClimMod 1.0 application
The EatlasClimMod main program will then be opened in Matlab as shown in Figure 5.
Figure 5. Window appearing in Matlab once the EatlasClimMod file has been opened
to run the EatlasClimMod 1.0 4. From there, press F5 or click on Run button application. The start up screen of this application will then appear as shown in Figure 6.
32
Figure 6. EatlasClimMod 1.0 application startup screen
This start up screen gives access to two menus: Operation and About. The About menu give the user has access to the Help file or to the summary screen window (Figure 7)
Figure 7. EatlasClimMod 1.0 © summary screen
The Operation menu gives access to five options: ¾ Preprocessing: used for data pre-processing; ¾ Heat index: used for the calculation of the daily heat index; ¾ Wave modelling: used to calculate the annual maximum wave for any variable over a given number of consecutive days ¾ Unique stations files: used to save the data of each weather station in a separated file;
33
¾ Gumbel analysis: used to predict the heat wave index for different return periods; this options contains two sub-options:
a. b.
All stations: used to apply the Gumbel method on all weather stations One station: used to apply the Gumbel method on a single weather station
¾ Exit: used to close the application.
From there, the pre-processing consists in removing the records with no data for air temperature and/or dew point (lines with a value of 9999.9), then sorting the records by station and by date. This is done using the following steps in the EatlasClimMod 1.0 application: 1. Click on Operation>Preprocessing. This will display the next screen of the wizard as shown in Figure 8.
Figure 8. Window used to specify the parameters for the pre-processing of the climatic data
The user can then pre-process the data created in section 3.2.1 for: a. a single year by checking the One year option, b. a period of consecutive years by checking the A period option. It is important to remember that this stage that the data files are named by date year.txt (1997.txt, 1998.txt, etc.) and stored in the “data” folder (see section 3.2.1 in this document). This window is also used to specify the path to the folder in which the generated under section 3.2.1 are located as well as the folder in which the pre-processed data should be saved. This is done by clicking on the browse button 34
next to the respective path.
2. Clik Ok once the information entered in the window reported on Figure 8 has been completed. This will automatically run the two functions behind this process, namely::delete_errorLine() and fct_Preprocessing(). This step can take time depending on the processor characteristics, the size of the data files and the complexity of the sorting operation. A status screen is therefore displayed and indicates which year is being currently processed (Figure 9).
Figure 9. EatlasClimMod 1.0 © status screen for the pre-processing operation
At the end, the program will save the new files under the name sort_year.txt (sort_1997.txt, sort_1998.txt, etc.) in the folder selected by the user under step 1. An example of the content of such files is presented on Figure 10.
Figure 10. Example of file resulting from the pre-processing of the climatic elements, with: column 1: Weather station number, column 2: Date of the measure, column 3: daily air temperature (°F), column 4: daily dew point temperature (°F)
3.3
Calculation of the daily heat index using the NCDC formula (SHI)
The daily heat index is calculated using the fct_CalculHeatIndexPerDay() function in EatlasClimMod 1.0. This is done using the following steps:
35
1.
In EatlasClimMod 1.0, click on Operation>Heat index. This will display the next wizard window as shown in Figure 11
Figure 11. Window used to specify the parameters for calculating the daily heat index
In this window, the user: -
can decide to calculate the daily heat index for a single year by checking the One year radio button, or over a period of consecutive years by checking the A period one. should then specify the path to the files resulting from the pre-processing operation (see section 3.2.2) and the path to the folder where the resulting files will be saved (naming this folder “HI” is recommended, this is done by clicking on the browse button
2.
(Figure 11).
After completing all the fields in the window, click on the OK button to run the heat index process. This step takes few minutes depending on the processor characteristics and the size of the data files. A status screen is therefore displayed and indicates which year is being currently processed (Figure 12).
Figure 12. EatlasClimMod© status screen for the heat index process
At the end of the treatment, EatlasClimMod 1.0 will have produced a new file calcul_year.txt (example calcul_1997.txt), containing the daily heat index for each year and station.
36
3.4
Calculation of the heat wave index for a given period of consecutive days and annual maximum heat wave index for each weather station and year of observation
EatlasClimMod 1.0 has been programmed in such a way that it can directly calculate the heat wave index for a given period of consecutive days and the annual maximum heat wave index for each period of observation and weather station. Furthermore, this function can also be used for any other variable and calculate the sum over a consecutive number of days in addition to the mean.
This particular component of the EatlasClimMod 1.0 application is based on the following three functions: 1. fct_MainProgram() for extracting the block of data records of each station and calculating the data frequency. 2. WaveModelling () for the calculation of the heat wave index by translation of a window of size equal to the heat wave size. This calculation is done only for data in successive dates. This function has been also used to select the maximum heat wave for each weather station. 3. check_succession() for checking the condition of succession of the dates in the window. To use these functions: 1. In EatlasClimMod 1.0 click on Operation>Wave modelling. This will display the next wizard window as shown in Figure 13. In this window, the user: - can decide to calculate the annual maximum heat wave index for a single year by checking the One year radio button, or over a period of consecutive years by checking the A period one. - Specifies the path to the files resulting from the heat index operation (see section 3.3) and the path to the folder where the resulting files will be saved (naming this folder “HW” is recommended, this is done by clicking -
-
on the browse button (Figure 13). Specifies the number of consecutive days to be considered for measuring the heat wave index. This information is to be entered in the Wave window size field. In the context of the e-atlas, a period of 3 consecutive days has been used. Check the Mean option to calculate the average daily heat index for each period of 3 consecutive days and then to calculate the annual maximum heat wave index
37
Figure 13. Window used to specify the parameters for calculating the annual maximum heat wave index
2. After completing all the fields in the window, click on the OK button to run the wave modelling process. This step takes few minutes. A status screen is therefore displayed and indicates which year is being currently processed (Figure 14).
Figure 14. EatlasClimMod© status screen for the wave modelling process
At the end of the treatment, EatlasClimMod 1.0 will have produced new files “final_year.txt” (example final_1997.txt) containing for each station the annual maximum heat wave index and the annual frequency computing for each year of observation using the following formula: =n/365, where n is the total number of days of observations per year. Figure 15 present one example of such file.
38
Figure 15. Example of file resulting from the wave modelling operation in EatlasClimMod 1.0 (final_1998.txt) (with column 1: Station, column 2: maximum mean heat wave, column 3: annual frequency)
3.5
Calculation of the annual maximum heat wave index for a two, five, height and ten years return period
3.5.1 Creation of weather station specific files Applying the Gumbel frequency method requires the creation of weather stations specific files containing the annual maximum heat wave indexes obtained through the process presented in section 3.4. This operation is carried out in the EatlasClimMod 1.0 application using the following steps: 1. Click on Operation>Unique stations files. This leads to the window presented on Figure 16.
39
Figure 16. Window used to specify the parameters for creating the weather station specific files
In this window, the user must specify: a. the period of consecutive years for which he wants the data to be extracted by weather station, b. the path to the folder containing the files resulting from the calculation of the average annual maximum heat wave index (see section 3.4) c. the path to the folder in which he wants the results to be saved For b. and c. the path can be changed by clicking on the browse button 2. Once the parameters entered in this window, click on Ok to start the process. This will automatically run the two functions behind this process, namely: save_list_STNs() and save_one_STN(). The resulting files will be named as follow: -
final_STN_NO_REDON.txt for the file containing all the source data sorted by weather station and year and with all the redundancies deleted station_number.txt (for example, 170700.txt) for the weather station specific files (see Figure 17 for an example of such file).
40
Figure 17. Example of weather station specific file with column 1: year, column 2: annual maximum heat wave index, column 3: annual frequency.
3.5.2 Application of the Gumbel frequency analysis The Gumbel frequency analysis technique has been programmed and included in the EatlasClimMod 1.0 application in order to calculate the annual maximum heat wave index over 3 consecutive days for any given weather station and return period. The use of this application requires the introduction of two thresholds which are used as filters to remove any weather station from the calculation in case these are not respected. While the user can specify these thresholds manually in EatlasClimMod 1.0, they have been fixed as follow in the context of the WHO e-atlas. Namely, a weather station would not be taken into account if: - the dataset for that given station does not contain a daily observation for at least 70% of the days in the year (255 days), - the number of year of observation for that station, after applying the first filter, is lower than 8 years. The threshold at eight years will give a good prediction of the annual maximum heat wave index for return periods that do not exceed eight years. In this work, we have nevertheless also calculated the annual maximum heat wave index for a 10 year return period even if this result should be taken with precaution. In EatlasClimMod 1.0, the Gumbel frequency analysis is applied in two steps: - Application of the Gumbel frequency analysis on all the stations - Correction and/or adjustment of the original dataset for unusual observations (typing mistakes, outliers,…). The steps to be followed for these two steps are described in the coming sections. 41
3.5.2.1
Application of the Gumbel frequency analysis on all the weather stations
Here are the steps to be followed in order to apply the Gumbel frequency analysis on all the weather stations at the same time: 1. In EatlasClimMod 1.0, click on Operation>Gumbel analysis; 2. Choose the All stations option. This will open the window presented in Figure 18.
Figure 18. Interface window of the application of the Gumbel method to all stations
In this dialogue box, the user must specify: a. the path to the folder containing the weather station specific files (see section 3.5.1), b. the path to the folder where the resulting files will be saved c. the two thresholds described in section 3.5.2 (annual frequency and minimum number of years of observations). For a. and b. the path to the selected folder can be changed by clicking on the browse button
42
3. Once all the parameters entered, click on the button to start the process, This will automatically run the Apply_Gumbel_allSTN() function which itself calls the following four functions: • fix_threshold(): used to remove the lines in the dataset for which the annual frequency is lower than the fixed threshold; • Gumbel(): produces the tables containing the values of the parameters involved in the Gumbel frequency method (Gumbel reduced variable, empirical frequency, mean and standard deviation) • fct_rank(): used to rank to the annual maximum heat wave index (the rank will be the same in the case of similar values for different years) • index_return(): computes the annual maximum heat wave index for two, five, eight and ten year return periods. In the window, the graph for each station is appearing one after the other once the analysis completed. The passage from a station graph to another one is done automatically. Figure 19 show one example of such window.
Figure 19. Example of window appearing as the gumbel analysis is completed on each weather station
At the end of the treatment, the EatlasClimMod 1.0 application will have produced a graph, plotting the annual heat wave versus the Gumbel reduced variate, for each of the station and stored this graph as an image file, named numSTN.jpg (with numSTN = the station number, 43
for example 85940.jpg, 600600.jpg, 601410.jpg), in the folder selected previously. Figure 20 present one example of such graph.
Figure 20. The annual maximum heat wave versus the Gumbel reduced variate distribution for weather station 293480 with the annual frequency threshold 70% and eight year return period
In addition to these graphs, the application also generates a summary file WaveModelledVariable _allSTN_return_2-5-8-10_Fq-0.7_NbrY-8.txt for each studied zone and places it in the same folder. This file contains, for each station, the annual maximum heat wave index for two, five, eight and ten year return periods as well as the correlation value between the annual maximum heat wave and a Gumbel reduced variable (see figure 20). Figure 21 presents an example of such a file.
44
Figure 21. Example of summary file resulting from the application of Gumbel analysis on all the stations in a given region
On the basis of the graph and summary file, the user is then identifying any potential errors in the original datasets, potential outliers (points far from the regression line on the graph) and/or correlation value lower than 0.80 for example. In these cases, the user writes down the number of the concerned stations and follows the steps reported in section 3.5.2.2. 3.5.2.2
Correction and/or adjustment of the original dataset for unusual observations
The application of the Gumbel frequency analysis on all the weather stations might reveal some data entry mistakes and/or outliers (see section 3.5.2.1). When identified, such cases needs to be corrected in the original dataset in order to improve the correlation in the analysis and therefore reduce the error on the final values for the annual maximum heat wave index for these stations. The following case illustrates how to modify an error in the original dataset and run again the Gumbel analysis on that particular weather station. The graph generated by the analysis for the weather station n°164290 (Figure 22) shows an isolated point (red circle on the graph). Let’s for example consider that this particular measure correspond to a year (1999) where the measurement instruments at the weather station has 45
been changed and that this resulted in wrong measurement of some of the climatic element needed in the application of the NCDC formula (Equation 1).
Figure 22. Graph resulting from the Gumbel frequency analysis for weather station 164290 with the isolated point indicated by the red circle
It would therefore be appropriate to remove this point from the analysis. To do so: 1. Make a copy of the file for that station (164290.txt) located in the “Stations files” folder 2. Using WordPad, open the version of the file (164290.txt) located in the “Stations files” folder (Figure 23).
Figure 23. Weather station n° 164290 specific file in which the year 1999 record is highlighted in blue
46
3. delete the record (the all line) corresponding to year 1999 (100.8434°F) 4. Save the file in the same location (do not move it); it is always preferable to make a copy before modifying a file Once the record deleted, it is possible to run again the Gumbel frequency analysis but on that weather station only. For that: 1. in EatlasClimMod 1.0, go to Operation>Gumbel analysis>One station (Figure 24).
Figure 24. Interface window of the application of the Gumbel method to one station
In this window, the user must specify: a. the path to the folder containing the weather station specific files (see section 3.5.1), b. the number of the station to be corrected (164290) in this case c. the two thresholds described in section 3.5.2 (annual frequency and minimum number of years of observations). 2. After completing all the boxes, click on the button 47
.
3. The new graph is then displayed (Figure 25). If the result is satisfactory, save the graph and the associated text file by selecting the path to the resulting files with the button
, then by validating the change by clicking on the
button.
Figure 25. Interface result of Gumbel application after correction
This operation will have for result to modify the previous version of the summary file (Figure 21), file which is necessary for the rest of the process.
3.6
Stepwise regression analysis
As described in section 2.3.5 stepwise regression analysis was used to identify the variables explaining the variation of the annual maximum heat wave index for each return period over the region covered in this version of the e-atlas and therefore derive the regression model to be used for its spatialization. This process has been implemented in three steps, which are explained in the following sections: • preparation of the GIS layers containing the spatial distribution of the causal factors and dependant variable (annual maximum heat wave index for two , five and ten year return periods), • preparation of the stepwise regression analysis, 48
•
application of the stepwise regression analysis.
Due to the climatological characteristics of the study area, the stepwise regression analysis has been applied separately to the five zones described in section 2.3.5.
3.6.1 Preparation of the GIS layers containing the spatial distribution of the causal factors and dependant variable As reported in section 2.3.5, the causal factors retained to explain the spatial variability of the annual maximum heat wave index are: 1. 2. 3. 4. 5. 6. 7.
Elevation (Z) Slope in percents (Slp_pr) Distance to the nearest coastline (d_Coast) Distance from the relative longitude (d_X) and latitude (d_Y) Distance to urban areas (d_urb_pop) Mean elevation within a 3×3 pixel window around each cell (Z9) Aspect (Asp).
The spatial distribution of the first 2 causal factors can be directly and easily derived from the dataset generated for the implementation of the different models (see the Methodology and implementation process for generating the dataset document also available from the WHO eatlas for disaster risk DVD). When it comes to the last 5 factors, specific process had to be applied in order to get the appropriate GIS layer for the analysis. Before applying the steps reported in the next sections, each of the layers used in this process first had to be projected into a metric projection system. This has been done using the process presented in Annex 3. In addition to that, and as mentioned in section 2.3.5, the all region covered in this version of the e-atlas had to be cut into 5 zones during the analysis to take their respective climatologic characteristics into account. Each of the layers used in the regression analysis therefore had to be cut according to the extent of the respective zone to which a 300 km buffer has been added in order to ensure a good interpolation at the edge of each of the zone. This process is described in Annex 4. 3.6.1.1
Preparation of the distance to the nearest coastline layers
First, the coastlines have been extracted from the projected version of the international boundaries layer (st_ar_int_bord_km.shp) using the following steps: 1. Make sure that the XTools Extension is uploaded in ArcView. 2. Display in the view both the international boundaries (st_ar_int_bord_km.shp) and the layer containing the global coastline border coming from the SALB project Un_coast_01.shp reprojected into metric projection system (Annex 3) (Un_coast_01_km.shp). 49
3. Use the XTools> Clip with polygon(s) function. 4. Select the Un_coast_01_km.shp as the theme that contains features that you wish to clip. 5. Select st_ar_int_bord_km.shp as the polygon theme that contains the polygons that will be used as the reference for the clipping. 6. Specify the name for the new shapefile to be created as st_ar_coast_km. The resulting projected coastline layer has then cut according to the 5 climatic zones (see Annex 4) to give the following layers: zone1_coast_buffer_300.shp, zone2_coast_buffer_300.shp, zone3_coast_buffer_300.shp, zone4_coast_buffer_300.shp, zone5_coast_buffer_300.shp. From there, the distance from the coastline was computed as follows starting with the first climatic zone: 1. In ArcView, display the zone1_coast_buffer_300.shp 2. Navigate to Analysis>Find Distance to create a grid of distances from the coasts in kilometres: 3. in the next menu, select Output Grid Specification and specify as follow: a. set Output Grid Extent = Same As zone1_int_bord_buffer_300.shp b. set Output Grid Cell Size = As Specified Below c. set Cell Size = 1 km d. use the default number of rows and columns e. save the result as zone1_dist_coast. 4. Repeat steps 1 to 3 on the buffered coastline shapefiles for the other zones changing the name of the resulting files accordingly. 3.6.1.2
Preparation of the distance from the relative latitude/longitude layer
The following process was followed for generating the climatic zone specific distance from the relative latitude layers: 1. Create a line shapefile that will pass by the point located at the extreme South of the of the frist climatic zone as follows: a. b. c. d.
In ArcView, add the zone1_int_bord_buffer_300.shp file in the view (See annex 4) create a new line theme using View>New Theme, manually digitize a straight horizontal line passing by the point located at the extreme South of the of this particular zone. This is going to be the zero degrees relative latitude for this zone, save the editing work as zone1_Y.shp. 50
2. Navigate to Analysis>Find Distance to create a grid of distances from the line generated in the zone1_Y.shp file. 3. in the next menu, select Output Grid Specification and specify as follow: a. set Output Grid Extent = Same As zone1_int_bord_buffer_300.shp b. set Output Grid Cell Size = As Specified Below c. set Cell Size = 1 km d. use the default number of rows and columns e. save the result as zone1_dist_Y. 4. Repeat steps 1 to 3 on the other zones. The process used to create a layer containing the relative longitude for study area is identical to the process described for the relative latitude drawing this time a vertical line passing by the point located at the extreme West of each climatic zone. This would represent the zero degrees relative longitude. The resulting output is then saved as zonen_dist_X (n corresponds to the number associated with each climatic zone). 3.6.1.3
Preparation of the distance to urban areas layer
This layer is necessary in order to take the urban heat island (UHI) effect into account in the regression analysis. The urban population layer was derived from the Global Rural–Urban Mapping Project (GRUMP), produced by the Center for International Earth Science Information Network (CIESIN), using the following steps: 1. Go to the GRUMP web site http://sedac.ciesin.columbia.edu/gpw/global.jsp [Accessed December 15, 2010] and specify the following in the different fields to download the layer in Arc/Info ASCII format: a. Get GRUMP (alpha): Urban Extens Grid b. Format: .ascii c. Resolution: 30’’ Click on the Get data button. This will automatically download two files, stored in a zip file: The first file contains the population counts in 2000 adjusted to match UN totals, the second file contains the geospatial dataset of urban/rural extents. These raster data are at 0.00833 degrees (30 arc-seconds) resolution 2. Unzip the two files that will download automatically on your computer and convert them from the Arc/Info ASCII format file into an ESRI GRID format using the following steps: a. b. c. d.
In ArcView, make sure that the Spatial Analyst extension is active open a new view and then go under "File" and choose "Import Data Source" In this window, select “ASCII raster” as import file type In the next window, Navigate and select the ASCII file that you have just uncompressed
51
e. ArcView then prompts to choose a location and a name for the new grid, save it as pop_00_grump and urban_rural_grump. 3. Add the urban_rural_grump grid in the view and open its attribute table. 4. Select the records with value equal 2. This corresponds to the urban area. 5. Convert this grid into vectot format using Theme>Convert to Shapefile function, and save the result as urban_grump.shp. 6. Display the pop_00_grump grid in the view and activate urban_grump.shp. 7. Select Analysis>Summarize zones function and in the specify the following in the Summarize Zones window: a. select “id” in the list of fields in the zone theme b. select the pop_00_grump grid, containing the population data to be summarized c. the result of the summarize zones operation is a table with the data we want, keyed by the “id” field. 8. Join the urban_grump.shp attribute to the summary table using the common “id” field. This will generate the population total number in each urban polygon 9. use the C-Tables Tools>Make Joins Permanent function to fix all the columns added in the attribute table and save the result as urban_pop.shp Once this done, the urban_pop.shp layer has been projected following the steps reported in Annex 3 before being cut according to the 5 climatic zones to which a buffer of 300 km buffer has been added (See Annex 4). The shape files resulting from these operations are named: zone1_urban_pop.shp, zone2_urban_pop.shp, zone3_urban_pop.shp, zone4_urban_pop.shp and zone5_urban_pop.shp. From there, the distance from urban settings presenting a population grater than 100 000, 200 000 and 500 000 has been spatially calculated for each zone using the following process: 1. Add in the view zone1_urban_pop_buffer_300.shp and open its attribute table. 2. Activate the Query Builder and enter the following formula in the box: ([popl_00]>=100000) to select only the urban areas where the population is more than 100 000. 3. Use the process described previously when creating the distance from the coastline in order to create a grid containing the distance from the urban centres with a population higher than 100 000, and save the result as zone1_dist_urb_100000. 4. Repeat steps 1 to 3 on the other populations: 200 000 and 500 000. 5. Repeat steps 1 to 4 on the other zones. 52
3.6.1.4
Preparation of the mean elevation distribution layer
The mean elevation distribution grid was derived from the DEM using the following steps: 1. In ArcView, make sure that the Grid Analyst extension is. 2. Add the st_ar_dem* grid in the view and activate it. 3. Use the Analysis>Neighborhood Statistics function specifying the following in the windows that appears: a.
statistic = Mean
b.
under neighbourhood, type of neighbourhood for analysis = Rectangle
c.
select the “cell” dial; and set “Width” and “Height” to three cells
d.
save the resulting grid as st_ar_Z9.
The resulting grid has then been projected using the steps reported in Annex 3 before being cut according to the extent of each climatic zones to which a 300 km has been added and this following the process reported in Annex 4. The grids resulting from these operations are named: Zone1_ Z9, Zone2_ Z9, Zone3_ Z9, Zone4_Z9 and Zone5_Z9. 3.6.1.5
Preparation of the aspect layers
The aspect, or slope direction, layer has been included in the regression analysis under the form of a dummy variable. This is being done as we don’t want to consider broad directions (N, NE, E, SE,…) and not any small variations in the direction of the slope. In order the use the aspect distribution layer as a dummy variable in the regression analysis, eight grids named Aspect_X (where X is N, NE, E, SE, S, SW, W, and NW) have been derived according to the following classification: • • • • • • • •
0°–22.5° and 337.5°–360°: North 22.5°–67.5°: North East 67.5°–112.5°: East 112.5°–157.5°: South East 157.5°–202.5°: South 202.5°–247.5°: South West 247.5°–292.5°: West 292.5°–337.5°: North West
For example, in Aspect_N cells which slope is directed towards the North are given a value of 1 while any other cells are given the value 0. The procedure used to create these grids is outlined in the following steps. 1. In ArcView, upload the aspect distribution grid st_ar_aspect* into the view. 2. Select the Analysis>Map Calculator function and enter the following formulas in the Calculator window: [st_ar_aspect]>=67.5 and [st_ar_aspect]Appends Tables Together function to create one unique table from the five created previously and save the result as st_ar_HW_2_5_8_10.dbf. 5. Merge the st_ar_HW_2_5_8_10.dbf with the st_ar_stations.shp* shapefile as follows: a. b. c.
in ArcView add the st_ar_HW_2_5_8_10.dbf table in the table window add the st_ar_stations.shp* in the view and open its attribute table Select the header of the STN column in the st_ar_HW_2_5_8_10.dbf table and the header of the Number column in the attribute table of the st_ar_stations.shp* shapefile
54
e.
keeping the attribute table of the st_ar_stations.shp* file active, join the two tables by clicking the join button on the tool bar f. use the C-Tables Tools>Make Joins Permanent function to fix all the columns added in the attribute table of the st_ar_stations.shp* shapefile g. save the resulting shapefile as st_ar_HW.shp. The resulting shape file has then been projected using the steps reported in Annex 3 before being cut according to the extent of each climatic zones to which a 300 km has been added and this following the process reported in Annex 4. The files resulting from these operations are named as follow: zone_1_HW.shp, zone_2_HW.shp, zone_3_HW.shp, zone_4_HW.shp, zone_5_HW.shp.
3.6.3 Preparation of the stepwise regression analysis Before performing the stepwise regression analysis on each of the five zones it was necessary to prepare a table which contained, for each weather station, the annual maximum heat wave index for each return period as well as the variables extracted from each grid prepared in section 3.6.1. The procedure used to create this table is outlined in the following steps. 1. In ArcView, make sure that the Grid Analyst extension is uploaded, 2. In a view, add the eight causal factor distribution grids for the first climatic zone zone1_dem, zone1_Z9, zone1_slope, zone1_asp_Y, zone1_dist_coast, zone1_dist_urb_Z (Z = 100000, 200000 or 500000), zone1_dist_X, zone1_dist_Y and the shapefile containing the distribution of the weather stations to which the annual maximum heat wave indexs for the four return periods have been associated for the first zone (zone1_HW.shp). 3. Make the zone1_HW.shp shapefile the active theme and use the Grid Analyst>Extract X, Y and Z Values for Point Theme from Grid functions 4. Select the first grids listed in step 2 from the drop list. The function will add and then populates three new fields in the attribute table of zone1_HW.shp (Xval, Yval and Zval), the last one storing the value extracted from the grid layer. 5. Open the attribute table of the zone1_HW.shp shapefile and click on the header of the “Zval” column. 6. Rename this field to correspond to the name of the raster layer (i.e. Z for elevation, Z9 for mean elevation, SLP for slope, ASP_X for aspect (where X is N, NE, E, SE, S, SW, W and NW), d_Coast for the distance from the coastline, d_X for the distance from the relative longitude, d_Y for the distance from the relative latitude) using the C-Tables Tools>Rename/Resize/Copy Field(s) function. 7. Repeat steps 3 to 5 on the other causal factor distribution grids until the Zval for each of them is integrated into the attribute table of the zone_1_HW.shp shapefile. 55
8. Save the resulting table as zone1_HW_regression.dbf. 9. Repeat steps 2 to 8 on the other zones, changing the names of the resulting files accordingly.
3.6.4 Application of the stepwise regression analysis Once the stepwise regression table is ready (see section 3.6.3), it is possible to perform the stepwise regression analysis on each zone and for each return period. This procedure is done using S-Plus software as follows: 1. Launch the S-Plus software. 2. Choose File>Import Data>From File function to import the stepwise regression table created above for the first climatic zone (zone1_HW_regression.dbf). 3. Choose Statistics>Regression>Stepwise function. 4. In the “Stepwise Linear Regression” dialogue box that appears: a. b.
c.
under Data Set scroll down the list and click on zone1_HW_regression click on the Create Formula box for the Upper Model and use HW_2 as the response and add all explanatory variables (Z, Z9, SLP, ASP_X, d_Coast, d_X, d_Y, d_pop_X) as Main Effects and Quadratic with: HW_2 = annual maximum heat wave index for a two year return period Z = elevation Z9 = mean elevation SLP = slope ASP_X = aspect (where X is N, NE, E, SE, S, SW, W and NW) d_Coast = distance from coastline d_X = Distance from the relative longitude d_Y = Distance from the relative latitude d_pop_X = distance from urban population (where X is 100000, 200000 and 500000) click OK to run the procedure.
A report is created that shows the result of this selection procedure with the coefficient of the variables selected and their significance, residual standard error, multiple R2 and probability (F statistic). Table 9 shows the report obtained for a two year return period in zone 2.
56
Table 9. Results of the regression analysis for the annual maximum heat wave index over zone 2 for a two year return period Variable (Intercept) d_X Z9 d_X2 d_Y2 d_urb_pop_2000002 Residual standard error Degrees of freedom Multiple R2 F statistic Probability (F statistic)
Regression coefficient
Standard error
t value
–1.1935559917 0.0414963292 –0.0156346046 –0.0000035280 –0.0000008424 –0.0000362393
17.25999 0.00707 0.00078 0.00000 0.00000 0.00000
–0.06915 5.86524 –20.03714 –5.01369 –6.38177 –7.55741
Probability Pr(>|t|) 0.94498 0.00000 0.00000 0.00000 0.00000 0.00000
3.36629 115 0.86 145.83142 0.0000
The regression equation explaining the maximum heat index for this particular return period (two years) and zone (zone 2) can be read as follows: HW_2 = 0.0414963292*d_X – 0.0156346046*Z9 – 0.0000035280*d_X2 – 0.0000008424*d_Y2 – 0.0000362393*d_urb_pop_2000002 – 1.1935559917 5. Repeat steps 3 to 4 for the other return periods in the same zone. 6. Repeat steps 2 to 5 for the other zones changing the names of the files accordingly. Annex 5 present the regressions obtained for all the return periods and climatic zones.
3.7
Spatialization of the annual maximum heat wave index for each return period
The annual maximum heat wave index distribution map for each return period and zone is created by applying the regressions found in section 3.6.4 on the corresponding grids as follows (example for the first climatic zone and two year return period): 1. Make sure that all the eight causal factor distribution layers for the first zone are uploaded in the view. 2. Select the Analysis>Map Calculator function and enter the following formula in the calculator window: ([zone_1_dist_X]*0.0414963292)–([zone_1_Z9]*0.0156346046)– ([zone_1_dist_X]*[zone_1_dist_X]*0.0000035280)– ([zone_1_dist_Y]*[zone_1_dist_Y]*0.0000008424)– ([dist_urb_200000]*[zone1_dist_urb_200000]*0.0000362393)–1.1935559917 3. Save the output grid as zone1_hw_2. This corresponds to the spatial distribution of the annual maximum heat wave index over the first zone for a two year return period. 57
4. Unproject the zone1_hw_2 layer from the Equal-Area Cylindrical projection to the Geographic one using the following steps: a. click either the function
button or use the Grid Projector>Grid and Theme Projector
b. select zone1_hw_2 from the list as the grid to project c. In the Grid Projector window: •
Specify the parameters for the current projection as follows: - Category = projection of the world - Type = Equal-Area Cylindrical projection - Current Projection Units = kilometres
• specify the parameters for the new projection as follows: - Category = projection of the world - Type = Geographic - New Projection Units = decimal degrees d. In the next window, specify the new cell size = 0.008333. e. Save the output grids as zone1_hw_2 _d. 5.
Repeat steps 2 to 4 for the other return periods using the corresponding regressions.
6.
Repeat steps 1 to 5 on the other zones.
The annual maximum heat wave index distribution maps for each return period and each climatic zone were then merged to generate three grids, one for each return period, covering the study area using the following steps: 1. In ArcView, make sure that the Grid Transformation Tool extension is uploaded. 2. Select the Transform Grid>Mosaic function to create one unique grid from zone1_hw_2_d, zone2_hw_2_d, zone3_hw_2_d, zone4_hw_2_d and zone5_hw_2_d. 3. Save the result as st_ar_hw_2_d. 4. Repeat steps 2 and 3 for the other two return periods and save the output mosaic as st_ar_hw_5_d, st_ar_hw_10_d. Finally the following process was applied to clip the annual maximum heat wave index distribution mosaic map for each return period to the borders of the region covered in this version of the e-atlas. 1. In ArcView, make sure that the Grid Analyst extension is uploaded, 58
2. upload the AFRO international boundary level used for the e-atlas (afro_int_bnd.shp) 3. Activate st_ar_hw_2_d and select the Grid Analyst>Extract Grid Theme Using Polygon function •
Click “yes” to continue.
•
Select the afro_int_bnd.shp from the dropdown list to be used as the layer on which the grid needs to be clipped and click OK
•
Make the output grid the active theme and choose the Theme>Convert to Grid function to create the AFRO annual maximum heat wave index distribution map for a two year return period
•
Save it as afro_hw_2.
4. Repeat steps 2 and 3 for the others WHO Regions (EURO and EMRO) and other return periods (five and ten) and save the outputs as afro_hw_5, afro_hw_10, emro_hw_2, emro_hw_5, emro_hw_10, euro_hw_2, euro_hw_5, euro_hw_10. The metadata associated with these grids are reported in Annex 6.
3.8
Creation of the heat wave hazard distribution maps
The last step of the method consisted in reclassifying the annual maximum heat wave index distribution maps to correspond to the five intensity levels selected for this project using the following process. 1. Upload the three annual maximum heat wave index distribution grids into the view. 2. Make the afro_hw_2 grid the active theme. 3. Use the Analysis>Reclassify function to reclassify the active them according to the following classification: Heat wave index ranges (°F)
Intensity level
< 80
1
80–90
2
90–105
3
105–129
4
≥130
5
4. Save the output grid as afro_hw_2_cl. 5. Select Theme>Edit Legend; in the Legend Editor window; change the legend to: 59
1: very low 2: low 3: medium 4: high 5: very high. 6. Repeat steps 1 to 5 for the for the others WHO Regions (EURO and EMRO) and others return periods (five and ten years) and save the results as afro_hw_5_cl, afro _hw_10_cl, emro_hw_2_cl, emro_hw_5_cl, emro_hw_10_cl, euro_hw_2_cl, euro_hw_5_cl, euro_hw_10_cl. The map resulting from the application of this approach for the European Region is reported in Figure 26. Please refer to the e-atlas DVD itself for the maps covering the other two WHO Regions and the other return period. The associated metadata for these layers can be found in Annex 7.
Figure 26. Heat wave hazard distribution map (Two year return period) for the countries of the European Region covered in this version of the WHO e-atlas of disaster risk
60
References and further reading Agnew MD, Palutikof JP, 2000. GIS-based construction of baseline climatologies for the Mediterranean using terrain variables. Climate research, 14:115–27. André J.C, Déqué M, Rogel P, Planton S, 2004. La vague de chaleur de l’été 2003 et sa prévision saisonnière. C. R. Geoscience, 336: 491–503. Ashraf M, Loftis JC and Hubbard KG, 1997. Application of geostatistics to evaluate partial weather station networks. Agricultural and forest meteorology, 84(3–4):255–71. Baccini M, Biggeri A, Accetta G, Kosatsky T, Katsouyanni K, Analitis A, Anderson HR, Bisanti L, D’Ippoliti D, Danova J, Forsberg B, Medina S, Páldy A, Rabczenko D, Schindler C, Michelozzi P, 2008. Heat effects on mortality in 15 European cities. Epidemiology, 9(5):711–9. Baldi M, Pasqui M, Cesarone F, De Chiara G, 2005. Heat waves in the Mediterranean region: analysis and model results. Paper presented to the 16th Conference on Climate Variability and Change, San Diego, California. Beniston M, 2004. The 2003 heat wave in Europe: a shape of things to come? An analysis based on Swiss climatological data and model simulations. Geophysical research letters, 31. doi:10.1029/2003GL018857. Benzi R, Deidda R, Marrocu M, 1997. Characterization of temperature and precipitation fields over Sardinia with principal component analysis and singular spectrum analysis. International journal of climatology, 17(11):1231–62. Browning P.A. (2009): A climatology of apparent temperature, proceeding of the 21st Conference on Climate Variability and Change Campbell-Lendrum D, Woodruff R, 2007. Climate change: quantifying the health impact at national and local levels. WHO environmental burden of disease series no. 14. http://whqlibdoc.who.int/publications/2007/9789241595674_eng.pdf [Accessed December 15, 2010] Chessa PA, Delitala AM, 1997. Objective analysis of daily extreme temperatures of Sardinia (Italy) using distance from the sea as independent variable. International journal of climatology, 17(13):1467–85. Collins FC Jr, Bolstad PV, 1996. A comparison of spatial interpolation techniques in temperature estimation, Proceedings of the Third International Conference/Workshop on Integrating GIS and Environmental Modeling. January 21–25, 1996, Santa Fe, New Mexico, USA. Condie R, Lee KA, 1982. Flood frequency analysis with historic information. Journal of hydrology, 58:47–61.
61
Delworth TL, Mahlman JD and Knutson TR, 1999. Changes in heat index associated with CO2-induced global warming. Climatic change, 43, 369–86. Dodson R, Marks D, 1997. Daily air temperature interpolated at high spatial resolution over a large mountainous region. Climate research, 8(1):1–20. Driscoll DM, 1992. Thermal comfort indexes: current uses and abuses. National weather digest, 17(4):33–8. Easterling DR, Meeh GA, Parmesan C, Changnon SA, Karl TR, Mearns LO, 2000. Climate extremes: observations, modeling, and impacts. Science, 289:2068–74. Ebi K, Meehl GA, 2007. Heatwaves and global climate change, the heat is on: climate change and heatwaves in the Midwest. In Regional impacts of climate change: four case studies in the United States. Pew Center on Global Climate Change, Arlington, Virginia:8–21. Eckstein BA, 1989. Evaluation of spline and weighted average interpolation algorithms. Computers & geosciences, 15(1):79–94. Eischeid JK, Baker FB, Karl TR, Diaz HF, 1995. The quality control of long-term climatological data using objective data analysis. Journal of applied meteorology, 34(12):2787–95. El Morjani Z, 2003. Conception d’un système d’information à référence spatiale pour la gestion environnementale; application à la sélection de sites potentiels de stockage de déchets ménagers et industriels en région semi-aride (Souss, Maroc). Doctoral thesis, University of Geneva. Terre et environnement, 42. El Morjani Z, Ebener SN, Boos J, Abdel Ghaffar E, Musani A, 2007. Modelling the spatial distribution of five natural hazards in the context of the WHO/EMRO Atlas of Disaster Risk as a step towards the reduction of the health impact related to disasters. International journal of health geographics, 6:8. Environment Canada, 2004. Adjusted Historical Canadian Climate (http://www.cccma.bc.ec.gc.ca/hccd/; Accessed December 15, 2010).
Data.
Foster HA, 1935. Duration curves. Transactions of the American Society of Civil Engineers, 99:1213–35. Frich A, Alexander LV, Della-Marta P, Gleason B, Haylock M,. Klein Tank AMG, Peterson T, 2002. Observed coherent changes in climatic extremes during the second half of the twentieth century. Climate research ,19: 193–212. Fuller WE, 1914. Flood flows. Transactions of the American Society of Civil Engineers, 77:564–617. Garcia-Plazaola JI, Esteban R, Hormaetxe K, Fernandez-Marin B, Becerril JM, 2008. Photoprotective responses of Mediterranean and Atlantic trees to the extreme heatwave of summer 2003 in southwestern Europe. Trees. 22:385–92. Gerard R, Karpuk EW, 1979. Probability analysis of historical flood data. Journal of the Hydraulics Division, ASCE, 105(HY9):1153–65. 62
Glickman Todd S, 2000. Glossary of meteorology. Boston, American Meteorological Society. Gosling SN, McGregor GR, Páldy A, 2007. Climate change and heat-related mortality in six cities. Part 1: model construction and validation. International journal of biometeorology, 51:525–40. Gosling SN, Lowe JA, McGregor GR, Pelling M, Malamud BD, 2008. Associations between elevated atmospheric temperature and human mortality: a critical review of the literature. Climatic change, 92:299–341. Granger K, Berechree M, 2003. Heat wave risks. In: Granger K, Hayme M, eds. Natural hazards and the risk they pose to south-east Queensland. Canberra, Australian Geological Survey Organisation. Gumbel EJ, 1941. The return period of flood flows. American mathematical statistics, 12(2):163–90. Gumbel EJ, 1942. Statistical control curves for flood discharges. Transactions, American Geophysical Union, 23:489–500. Gumbel EJ, 1954. The statistical theory of droughts. Proceedings of the American Society of Civil Engineers, 80:1–19. Gumbel EJ, 1960. Bivariate exponential distribution. Journal of the American Statistical Association, 55:698–707. Hajat S, Kovats RS, Atkinson RW, Haines A, 2002. Impact of hot temperatures on death in London: a time series approach. J Epidemiol Community Health 56:367–372 Hajat S, Armstrong BJ, Baccini M, Biggeri A, Bisanti L, Russo A, Páldy A, Menne B, Kosatsky T, 2006. Impact of high temperatures on mortality:is there an added heat effect. Epidemiology, 17(6):632–638. Hammond T, Yarie J, 1996. Spatial prediction of climatic state factor regions in Alaska. Ecoscience, 3(4):490–501. Hargy VT, 1997. Objectively mapping accumulated temperature for Ireland. International journal of climatology, 17(9):909–27. He Y, Bárdossy A, Brommundt J, Enzenhöfer R, 2006. Non-stationary flood frequency analysis in the context of climate variability. Geophysical research abstracts, 8, 04597. Holdaway MR, 1996. Spatial modelling and interpolation of monthly temperature using kriging. Climate research, 6:215–25. Hudson G, Wackernagel H, 1994. Mapping temperature using kriging with external drift: theory and an example from Scotland. International journal of climatology, 14:77–91. Hulme M, Conway D, Jones PD, Jiang T, Barrow EM, Turney C, 1995. Construction of a 1961–1990 European climatology for climate change modelling and impact applications. International journal of climatology, 15:1333–63. 63
Hutchinson MF, Gessler PE 1994. Splines—more than just a smooth interpolator, Geoderma, 62:45–67. Huth R, Kyselý J, Pokorná L,. 2000. A GCM simulation of heat waves, dry spells, and their relationships to circulation. Climatic change, 46:29–60. Hutter H.P, Moshammer H, Wallner P, Leitner B, Kundi M, 2007. Heatwaves in Vienna: effects on mortality. Middle European journal of medicine, 119(7–8):223–7. Huynen MMTE, Martens P, Schram D,Weijenberg MP, Kunst AE, 2001. The impact of heat waves and cold spells on mortality rates in the Dutch population. Environmental health perspectives, 109:463–70. IPCC, 2000. Emissions scenarios. A special report of IPCC working group III for the Intergovernmental Panel on Climate Change (http://www.ipcc.ch/ipccreports/sres/emission/index.htm; Accessed December 15, 2010). IPCC, 2007. Summary for policymakers. In Climate Change 2007: the physical science basis. Contribution of working group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge and New York, Cambridge University Press. Jones T, Middelmann M, Corby N, 2005. Natural hazard risk in Perth, Western Australia. Canberra, Geoscience Australia. Kite GW, 1977. Frequency and risk analyses in hydrology. Fort Collins, Colorado, Water Resources Publications. Koppe C, Kovats S, Jendritzky G, Menne B, 2004. Heat-waves: risks and responses. Copenhagen, World Health Organization Regional Office for Europe. Health and Global Environmental Change, series no. 2. Koutsoyiannis D, 2004. On the appropriateness of the Gumbel distribution for modelling extreme rainfall In Hydrological risk: recent advances in peak river flow modelling, prediction and real-time forecasting. Assessment of the impacts of land-use and climate changes:303–19. Bologna, Italy, Editoriale Bios, Castrolibero. Kovats SR, Jendritzky G, 2006. Heat-waves and human health. In Menne B, Ebi KL, eds., Climate Change and Adaptation Strategies for Human Health. Darmstadt, Steinkopff:63–97. Landwehr JM, Matalas NC, Wallis JR, 1979. Probability weighted moments compared with some traditional techniques in estimating Gumbel parameters and quantiles. Water resources research, 15:1055–64. Legates DR, Willmott CJ, 1990. Mean seasonal and spatial variability in global surface air temperature. Theoretical and applied climatology, 41:11–21. Lennon JJ, Turner JRG, 1995. Predicting the spatial distribution of climate: temperature in Great Britain. Journal of animal ecology, 64:370–92. 64
Li PW, Chan ST, 2000. Application of a weather stress index for alerting the public to stressful weather in Hong Kong. Meteorological applications, 7:369–75. Li J, Huang JF, Wang XZ, 2006. A GIS-based approach for estimating spatial distribution of seasonal temperature in Zhejiang province, China. Journal of Zhejiang University science A, 7(4):647–56. Lutgens FK, Tarbuck EJ, 1995. The atmosphere, 6th ed. Englewood Cliffs, New Jersey, Prentice-Hall, Inc. MacEachren AM, Davidson JV, 1987. Sampling and isometric mapping of continuous geographic surfaces. American cartographer, 14(4):299–320. Masterton JM and Richardson FA, 1979. Humidex: a method of quantifying human discomfort due to excessive heat and humidity. Ontario, Canada, Environment Canada, Atmospheric Environment Service. Matheron G, 1963. Principles of geostatistics. Economic geology, 58:1246–66. Meehl GA, Tebaldi C, 2004. More intense, more frequent, and longer lasting heat waves in the 21st century. Science, 305(5686):994–7. Meehl GA, Stocker TF, Collins WD, Friedlingstein P, Gaye AT, Gregory JM, Kitoh A, Knutti R, Murphy JM, Noda A, Raper SCB, Watterson IG, Weaver AJ, Zhao ZC, 2007. Global climate projections. In Solomon S, Qin D, Manning M, Chen Z, Marquis M, Averyt KB, Tignor M, Miller HL, eds. Climate change 2007: the physical science basis. Contribution of working group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge and New York, Cambridge University Press. Meylan P, Musy A, 1999. Hydrologie fréquentielle. Bucharest, Editions HGA. Michelozzi P, Kirchmayer U, Katsouyanni K, Biggeri A, McGregor G, Menne B, Kassomenos P, Anderson HR, Baccini M, Accetta G, Analytis A, Kosatsky T, 2007. Assessment and prevention of acute health effects of weather conditions in Europe, the PHEWE project: background, objectives, design. Environmental health, 6:12. Moin SMA, Shaw MA, 1985. Regional flood frequency analysis for Ontario streams: volume 1, single station analysis and index method. Burlington, Ontario, Inland Waters Directorate, Environment Canada. Myers RH, 1990. Classical and modern regression with applications. Boston, PWS-Kent Publishing. Oke TR, 1995. The heat island characteristics of the urban boundary layer: characteristics, causes and effects. In Cermak JE, Davenport AG, Plate EJ, Viegas DX, eds. Wind climate in cities:81–107. Alphen an den Rijn, Netherlands, Kluwer Academic. Palecki MA, Changnon SA, Kunkel KE, 2001. The nature and impacts of the July 1999 heat wave in the midwestern United States: learning from the lessons of 1995. Bulletin of the American Meteorological Society, 82:1353–67. 65
Parsons KC, 2003. Human thermal environments: the effects of hot, moderate, and cold environments on human health, comfort and performance. London and New York, Taylor and Francis. Patzert WC, LaDochy S, Willis JK, Mardirosian T, 2007. Will the real Los Angeles stand up: impacts of a weather station relocation on climatic records (and record weather). 16th Conference on Applied Climatology, American Meteorological Society Annual Meeting. Rebetez M, Dupont O, Giroud M, 2008. An analysis of the July 2006 heatwave extent in Europe compared to the record year of 2003. Theoretical and Applied Climatology. Printed in The Netherlands. Reid CE, O’Neill MS, Gronlund CJ, Brines SJ, Brown DG, Diez-Roux AV, Schwartz J, 2009. Mapping community determinants of heat vulnerability. Environmental health perspectives, 117:1730–36. Robine JM, Cheung SLK, Le Roy S, Van Oyen H, Herrmann FR, 2003. Report on excess mortality in Europe during summer 2003. EU Community Action Programme for Public Health, Grant Agreement 2005114 (http://ec.europa.eu/health/ph_projects/2005/action1/docs/action1_2005_a2_15_en.pdf; Accessed December 15, 2010). Robine JM, Cheung SLK, Le Roy S et al., 2008. Death toll exceeded 70 000 in Europe during the summer of 2003. Comptes rendus biologies, 331(2):171–8. Robinson PJ, 2001. On the definition of a heat wave. Journal of applied meteorology, 40(4):762–75. Rothfusz LP, 1990. The heat index “equation” (or, more than you ever wanted to know about heat index). Fort Worth, Texas, National Oceanic and Atmospheric Administration, National Weather Service, Office of Meteorology. SR/SSD 90-23 (http://www.srh.noaa.gov/images/ffc/pdf/ta_htindx.pdf; Accessed December 15, 2010). Sarma P, 1999. Flood risk zone mapping of Dikrong sub basin in Assam. (http://www.gisdevelopment.net/application/natural_hazards/floods/nhcy0006pf.htm; Accessed December 15, 2010) Schär C, Vidale PL, Lüthi D, Frei C, Häberli C, Liniger MA, Appenzeller C, 2004: The role of increasing temperature variability in European summer heatwaves. Nature, 427:332– 6. Souch C, Grimmond S, 2006. Applied climataology: Urban climate. Progress in physical geography, 30:270–9. Steadman RG, 1979. The assessment of sultriness. Part 1: A temperature-humidity index based on human physiology and clothing science. Journal of applied meteorology, 18:861–73. Steadman RG, 1984. A universal scale of apparent temperature. Journal of climate and applied meteorology, 23:1674–87.
66
Stedinger JR, Vogel RM, Foufoula-Georgiou E, 1992. Frequency analysis of extreme events. In Maidment DA, ed. Handbook of hydrology. New York, McGraw-Hill. Stolte W, Dumontier S, 1977. Flood frequency analysis for mountainous and prairie streams. Edmonton, Canada, University of Saskatchewan and Alberta Environment. Tamrazian A, LaDochy S, Willis JK, Patzert WC, 2008. Heat waves in southern California: are they becoming more frequent and longer lasting? APCG [Association of Pacific Coast Geographers] yearbook, 70:59–69. Tan J, Zheng Y, Song G, Kalkstein LS, Kalkstein AJ, Tang X, 2007. Heatwave impacts on mortality in Shanghai, 1998 and 2003. International journal of biometeorology, 51:193– 200. USACE, 1993. Hydrologic frequency analysis. Engineer Manual 1110-2-1415. Washington DC, US Army Corps of Engineers. Vogel RM, 1986. The probability plot correlation coefficient test for the normal, lognormal, and Gumbel distributional hypotheses. Water resources research, 22(4):587–90; corrections, 23(10):2013. Vogt JV, Viau AA, Paquet F, 1997. Mapping regional air temperature fields using satellitederived surface skin temperatures. International journal of climatology, 17(14):1559– 79. Willmott CJ, Matsuura K, 1995. Smart interpolation of annually averaged air temperature in the United States. Journal of applied meteorology, 34(12):2577–86. World Health Organization (WHO), 2003a. The health impacts of 2003 summer heat waves. Briefing note for the delegations of the fifty-third session of the WHO Regional Committee for Europe. Vienna, Austria, 8–11 September 2003. 12pp. At (http://www.irisknet.cn/riskdb/globalrisk/juzai/10.6.pdf;Accessed December 15, 2010) World Health Organization (WHO), 2003b. Methods of assessing human health vulnerability and public health adaptation to climate change. World Health Organization, Geneva (http://www.euro.who.int/document/e81923.pdf; Accessed December 15, 2010) World Health Organization (WHO), 2009. Improving public health responses to extreme weather/heat-waves: EuroHEAT. Technical summary. Copenhagen, World Health Organization Regional Office for Europe. World Meteorological Organization (WMO), 2004. Guidelines on Biometeorology and air quality forecasts, PWS-10, WMO/TD No. 1184
67
Annex 1. Steadman’s table for heat index °F (°C), Relative humidity (%) Temp. 110 (47) 108 (43) 106 (41) 104 (40) 102 (39) 100 (38) 98 (37) 96 (36) 94 (34) 92 (33) 90 (32) 88 (31) 86 (30) 84 (29) 82 (28) 80 (27)
40 136 (58) 130 (54) 124 (51) 119 (48) 114 (46) 109 (43) 105 (41) 101 (38) 97 (36) 94 (34) 91 (33) 88 (31) 85 (29) 83 (28) 81 (27) 80 (27)
45
50
55
60
65
70
75
80
85
90
95
100
137 (58) 130 (54) 124 (51) 119 (48) 114 (46) 109 (43) 104 (40) 100 (38) 96 (36) 93 (34) 89 (32) 87 (31) 84 (29) 82 (28) 80 (27)
137 (58) 131 (55) 124 (51) 118 (48) 113 (45) 108 (42) 103 (39) 99 (37) 95 (35) 91 (33) 88 (31) 85 (29) 83 (28) 81 (27)
137 (58) 130 (54) 124 (51) 117 (47) 112 (44) 106 (41) 101 (38) 97 (36) 93 (34) 89 (32) 86 (30) 84 (29) 81 (27)
137 (58) 129 (54) 123 (51) 116 (47) 110 (43) 105 (41) 100 (38) 95 (35) 91 (33) 88 (31) 84 (29) 82 (28)
136 (58) 128 (53) 121 (49) 114 (46) 108 (42) 103 (39) 98 (37) 93 (34) 89 (32) 85 (29) 82 (28)
134 (57) 126 (52) 119 (48) 112 (44) 106 (41) 100 (38) 95 (35) 90 (32) 86 (30) 83 (28)
132 (56) 124 (51) 116 (47) 109 (43) 103 (39) 97 (36) 92 (33) 88 (31) 84 (29)
129 (54) 121 (49) 113 (45) 106 (41) 100 (38) 94 (34) 89 (32) 84 (29)
135 (57) 126 (52) 117 (47) 110 (43) 102 (39) 96 (36) 90 (32) 85 (29)
131 (55) 122 (50) 113 (45) 105 (41) 98 (37) 91 (33) 86 (30)
127 (53) 117 (47) 108 (42) 100 (38) 93 (34) 86 (30)
132 (56) 121 (49) 112 (44) 103 (39) 95 (35) 87 (31)
68
Annex 2. Description of the NCDC daily meteorological elements dataset
This annex provides the indication of the type of the data (int[eger], real or char[acter]) as well as a description of each of the fields of the daily meteorological data coming from the global surface summary of the day data produced by the National Climatic Data Center (NCDC). Field
Type
Description
STN
Int
Station number
WBAN
Int
This is the historical “Weather Bureau Air Force Navy” number where applicable
YEARMODA Int
Year, month and day
TEMP
Real
Mean temperature for the day in degrees fahrenheit to tenths. Missing = 9999.9
Count
Int
Number of observations used in calculating mean temperature
DEWP
Real
Mean dew point for the day in degrees fahrenheit to tenths. Missing = 9999.9
Count
Int
Number of observations used in calculating mean dew point
SLP
Real
Real mean sea level pressure for the day in millibars to tenths. Missing = 9999.9
Count
Int
Number of observations used in calculating mean sea level pressure
STP
Real
Mean station pressure for the day in millibars to tenths. Missing = 9999.9
Count
Int
Number of observations used in calculating mean station pressure
VISIB
Real
Mean visibility for the day in miles to tenths. Missing = 999.9
Count
Int
Number of observations used in calculating mean visibility
WDSP
Real
Mean wind speed for the day in knots to tenths. Missing = 999.9
69
Field
Type
Description
Count
Int
Number of observations used in calculating mean wind speed
MXSPD
Real
Maximum sustained wind speed reported for the day in knots to tenths. Missing = 999.9
GUST
Real
Maximum wind gust reported for the day in knots to tenths. Missing = 999.9
MAX
Real
Maximum temperature reported during the day in degrees fahrenheit to tenths. Missing = 9999.9
Flag
Char
Blank indicates maximum temperature was taken from the explicit maximum temperature report and not from the hourly data. * indicates maximum temperature was derived from the hourly data (i.e. highest hourly or synoptic-reported temperature)
MIN
Real
Minimum temperature reported during the day in degrees fahrenheit to tenths—time of minimum temperature report varies by country and region, so this will sometimes not be the minimum for the calendar day. Missing = 9999.9
Flag
Char
Blank indicates minimum temperature was taken from the explicit minimum temperature report and not from the hourly data. * indicates minimum temperature was derived from the hourly data (i.e. lowest hourly or synoptic-reported temperature)
PRCP
Real
Total precipitation (rain and/or melted snow) reported during the day in inches and hundredths; will usually not end with the midnight observation—i.e. may include latter part of previous day. 0.00 indicates no measurable precipitation (includes a trace). Missing = 99.99
70
Field
Type
Description
Flag
Char
A = one report of 6-hour precipitation amount B = summation of two reports of 6-hour precipitation amount C = summation of three reports of 6-hour precipitation amount D = summation of four reports of 6-hour precipitation amount E = one report of 12-hour precipitation amount F = summation of 2 reports of 12-hour precipitation amount G = one report of 24-hour precipitation amount H = station reported 0 as the amount for the day (e.g. from 6-hour reports), but also reported at least one occurrence of precipitation in hourly observations—this could indicate a trace occurred, but should be considered as incomplete data for the day I = station did not report any precipitation data for the day and did not report any occurrences of precipitation in its hourly observations—it is still possible that precipitation occurred but was not reported
SNDP
Real
Snow depth in inches to tenths—last report for the day if reported more than once. Missing = 999.9 Note: most stations do not report 0 on days with no snow on the ground; therefore, 999.9 will often appear on these days
FRSHTT
Int
Indicators (1 = yes, 0 = no/not reported) for the occurrence during the day of: fog (‘F’—1st digit); rain or drizzle (‘R’—2nd digit). Snow or ice pellets (‘S’—3rd digit). Hail (‘H’—4th digit). Thunder (‘T’—5th digit). Tornado or funnel cloud (‘T’—6th digit).
71
Annex 3. Projection of a GIS layers into the metric projection system This operation was used to switch the map units of the different layers used in the analysis from decimal degrees to kilometres in order to be able to measure distances. The operation was performed separately on each the vector and raster layers. Taking the international boundaries as an example (st_ar_int_bord.shp*), all the vector layers have been projected using the following steps: 1. In ArcView, make sure that the Grid and Theme Projector v.2 extension are uploaded. button or use the Grid Projector>Grid and Theme Projector 2. Click either the function from the menu. 3. In the window that open, select the st_ar_int_bord.shp* layer from the list as the theme to project. 4. In the next window that opens: a.
Specify the following parameters for the current projection Category = projection of the world Type = Geographic Current Projection Units = decimal degrees
b.
Specify the following parameters for the new projection Category = projection of the world Type = Equal-Area Cylindrical New Projection Units = kilometers.
5. Save the output theme as st_ar_int_bord_km.shp. Use the same process to reproject the st_ar_HW.shp shapefile, the global coastline border coming from the SALB project Un_coast_01.shp, and the urban population “urban_grump.shp”, and then save the result respectively as st_ar_HW_km.shp, st_ar_coast_km.shp and “urban_grump_km.shp” The raster layers have also been projected. Here is for example the process followed for the Digital Elevation Model (DEM): 1. Click either the
button or select Grid Projector>Grid and Theme Projector.
2. Select the st_ar_dem grid from the list in the next window. 3. In the next window that opens: 72
a.
specify the parameters for the current projection Category = projection of the world Type = Geographic Current Projection Units = decimal degrees
b.
specify the parameters for the new projection Category = projection of the world Type = Equal-Area Cylindrical New Projection Units = kilometres
c.
in the next window, specify the new cell size (in km) = 1
d.
choose Interpolation Method = Bilinear Interpolation and Transformation Order = 4 save the output grid as st_ar_dem_km
e.
73
Annex 4. Creation of a 300 km buffer around each climatic zone and clipping of the different layers for the regression analysis In order to insure good interpolation at the edge of each of the 5 climatic zones considered in this version of the e-atlas to conduct the regression analysis, a 300 km buffer was added to each of these zones using the following process: 1. Extract the five zones (see section 2.3.5) from the projected version of the international boundaries layer (See Annex 3) using the following steps in ArcView: a. b. c. d. e. f.
In ArcView, make sure that the XTools Extension is uploaded add the projected version of the international boundaries layer st_ar_int_bord_km.shp, and open its attribute table, select the countries part of the first zone (see section 2.3.5) In the view, make the st_ar_int_bord_km.shp shapefile the active theme use the Theme>Convert To Shapefile function to create a shapefile containing only the countries part of the first zone and save it as zone1_int_bord_km.shp repeat steps c) to e) for the other four zones.
2. Create the 300 km buffer around each climatic zone as follows: a. b. c. d. e.
f.
from the View>Properties change the Map Units and Distance Units to kilometres select Theme>Create Buffer; the Create Buffers wizard appears in the first wizard dialog box, choose zone1_int_bord_km.shp as the items to buffer in the second dialogue box, select At a Specified Distance as the method used to create the buffer and Width of the Buffer = 300. Make sure that the distance units are set to kilometres in the third dialogue box, create the buffer using Only Inside of the Polygon Parameter. Specify that the buffer be saved as a new theme and save as zone_1_int_bord_buffer_300.shp. The new buffer theme will be added to the current view Repeat steps c) to e) on the other four zones.
Taking the projected version of the weather station location layer (st_ar_HW_km.shp) as an example, the following steps have then been applied to clip each vector layer to the 5 buffered climatic zones. 1. In ArcView, make sure that the XTools Extension is uploaded, 2. In the view, display both the buffered international boundaries of the first zone zone1_int_bord_buffer_300.shp and the projected layer containing the distribution of the weather stations with the associated annual maximum heat wave index st_ar_HW_km.shp. 74
3. Use the XTools> Clip with Polygon(s) function. 4. Select st_ar_HW_km.shp as the theme that contains features that you wish to clip. 5. Select zone1_int_bord_buffer_300.shp as the polygon theme that contains the polygons that will be used as the reference for the clipping. 6. Specify the name for the new shapefile to be created as zone_1_HW.shp. 7. Repeat steps 2 to 6 for the others zones For the raster layers, taking the DEM as an example, the following steps are applied for the clipping. 1. In ArcView, make sure that the Grid Analyst extension is uploaded, 2. In the view, add the st_ar_dem_km grid and zone1_int_bord_buffer_300.shp shape file. 3. Make the first grid st_ar_dem_km active and use the Grid Analyst>Extract Grid Theme Using Polygon function. 4. Select first the zone1_int_bord_buffer_300.shp from the drop list to use in the clip. 5. Make the resulting grid the active theme and select the Theme>Convert to Grid function to save the output grid under zone1_dem. 6. Repeat steps 2 to 5 for the four other zones.
75
Annex 5. Final regression by climatic zone and return period Climatic Zone 1 2 year return period Variable (Intercept) d_X d_Y d_Coast Z9 d_X2 d_Y2 d_Coast2 Residual standard error Degrees of freedom Multiple R2 F statistic Probability (F statistic)
Regression coefficient 95.7206132477 -0.0033862017 0.0035901753 0.0182266865 -0.0123123549 0.0000004905 -0.0000010259 -0.0000118808
Standard error 0.94595 0.00056 0.00103 0.00266 0.00060 0.00000 0.00000 0.00000
t value 101.18960 -5.96629 3.45422 6.84982 -20.29256 4.13806 -3.97147 -4.96879
Probability Pr(>|t|) 0.00000 0.00000 0.00061 0.00000 0.00000 0.00004 0.00008 0.00000
2.36714 376 0.87 102.11671 0.0000
5 year return period Variable (Intercept) d_X d_Y d_Coast Z9 d_X2 d_Y2 d_Coast2 Residual standard error Degrees of freedom Multiple R2 F statistic Probability (F statistic)
Regression coefficient 98.0572978356 -0.0031550047 0.0037923961 0.0178097335 -0.0132616354 0.0000004247 -0.0000010430 -0.0000118238
Standard error 0.97263 0.00058 0.00106 0.00273 0.00062 0.00000 0.00000 0.00000
t value 100.81583 -5.40642 3.54868 6.50950 -21.25745 3.48489 -3.92679 -4.80928
Probability Pr(>|t|) 0.00000 0.00000 0.00043 0.00000 0.00000 0.00055 0.00010 0.00000
2.49033 376 0.86 108.23526 0.0000
10 year return period Variable (Intercept) d_X d_Y d_Coast Z9 d_X2 d_Y2 d_Coast2 Residual standard error Degrees of freedom Multiple R2 F statistic Probability (F statistic)
Regression coefficient 99.6050334239 -0.0030019589 0.0039259987 0.0175286827 -0.0138882329 0.0000003812 -0.0000010543 -0.0000117823 2.67579 376 0.87 107.82327 0.0000
76
Standard error 1.01280 0.00060 0.00111 0.00284 0.00064 0.00000 0.00000 0.00000
t value 98.34536 -4.94013 3.52798 6.15266 -21.37888 3.00390 -3.81191 -4.60234
Probability Pr(>|t|) 0.00000 0.00000 0.00047 0.00000 0.00000 0.00284 0.00016 0.00000
Climatic Zone 2 2 year return period Variable (Intercept) d_X Z9 d_X2 d_Y2 d_urb_pop_2000002 Residual standard error Degrees of freedom Multiple R2 F statistic Probability (F statistic)
Regression coefficient –1.1935559917 0.0414963292 –0.0156346046 –0.0000035280 –0.0000008424 –0.0000362393
Standard error 17.25999 0.00707 0.00078 0.00000 0.00000 0.00000
t value –0.06915 5.86524 –20.03714 –5.01369 –6.38177 –7.55741
Probability Pr(>|t|) 0.94498 0.00000 0.00000 0.00000 0.00000 0.00000
3.36629 115 0.86 145.83142 0.0000
5 year return period Variable
Regression coefficient
(Intercept) d_X Z9 d_X2 d_Y2 d_urb_pop_2000002
-1.4858951684 0.0433285345 -0.0174196720 -0.0000036910 -0.0000009011 -0.0000387645
Residual standard error Degrees of freedom Multiple R2 F statistic Probability (F statistic)
3.998179721 115 0.851852017 132.2501731 0.0000
Standard error 19.75782 0.00089 0.00089 0.00000 0.00000 0.00000
t value -0.07520 5.34997 -19.50249 -4.58216 -5.96325 -7.06201
Probability Pr(>|t|) 0.94018 0.00000 0.00000 0.00001 0.00000 0.00000
10 year return period Variable (Intercept) d_X Z9 d_X2 d_Y2 d_urb_pop_2000002 Residual standard error Degrees of freedom Multiple R2 F statistic Probability (F statistic)
Regression coefficient -1.7218641147 0.0445593627 -0.0186022240 -0.0000038007 -0.0000009400 -0.0000404400 4.23859 115 0.84 119.25738 0.0000
77
Standard error 21.95895 0.00900 0.00099 0.00000 0.00000 0.00000
t value -0.07841 4.95044 18.73883 -4.24535 -5.59713 -6.62876
Probability Pr(>|t|) 0.93763 0.00000 0.00000 0.00004 0.00000 0.00000
Climatic Zone 3 2 year return period Variable (Intercept) d_Y d_Coast Z d_X2 d_Y2 d_Coast2 d_urb_pop_1000002 Residual standard error Degrees of freedom Multiple R2 F statistic Probability (F statistic)
Regression coefficient -53.1781889487 0.0969459665 -0.0205682576 -0.0127255301 0.0000002616 -0.0000148160 0.0000067601 -0.0000193574
Standard error 36.37329 0.01937 0.00574 0.00067 0.00000 0.00000 0.00000 0.00001
t value -1.46201 5.00409 -3.58147 -18.96185 2.93343 -5.67149 3.17242 -1.43094
Probability Pr(>|t|) 0.14656 0.00000 0.00050 0.00000 0.00407 0.00000 0.00195 0.15525
3.0879 111 0.88 115.21083 0.0000
5 year return period Variable (Intercept) d_Y d_Coast Z d_X2 d_Y2 d_Coast2 d_urb_pop_1000002 Residual standard error Degrees of freedom Multiple R2 F statistic Probability (F statistic)
Regression coefficient -71.4881968853 0.1107702050 -0.0220452333 -0.0135653761 0.0000002657 -0.0000169619 0.0000079712 -0.0000164899
Standard error 34.27109 0.01794 0.00599 0.00073 0.00000 0.00000 0.00000 0.00000
t value -2.08596 6.17331 -3.67700 -18.50377 2.73039 -6.96450 3.60031 -2.58500
Probability Pr(>|t|) 0.03927 0.00000 0.00036 0.00000 0.00735 0.00000 0.00047 0.01103
3.3922 111 0.88 114.83167 0.0000
10 year return period Variable (Intercept) d_Y d_Coast Z d_X2 d_Y2 d_Coast2 d_urb_pop_1000002 Residual standard error Degrees of freedom Multiple R2 F statistic Probability (F statistic)
Regression coefficient -80.0184786476 0.1161119758 -0.0233678124 -0.0143146748 0.0000002912 -0.0000176413 0.0000080805 -0.0000172978 3.7352 111 0.87 107.58299 0.0000
78
Standard error 36.94741 0.01934 0.00646 0.00079 0.00000 0.00000 0.00000 0.00000
t value -2.16573 6.00228 -3.61528 -18.11147 2.77631 -6.71880 3.38533 -2.51523
Probability Pr(>|t|) 0.03247 0.00000 0.00045 0.00000 0.00645 0.00000 0.00098 0.01332
Climatic Zone 4 2 year return period Variable (Intercept) d_X d_Y d_Coast Z9 d_urb_pop_200000 d_Y2 Residual standard error Degrees of freedom Multiple R2 F statistic Probability (F statistic)
Regression coefficient 176.4771561885 -0.0004623153 -0.0292519325 0.0032572948 -0.0083252236 0.0032310067 0.0000020952
Standard error 8.86203 0.00007 0.00389 0.00034 0.00025 0.00063 0.00000
t value 19.91380 -6.02430 -7.51875 9.48589 -32.46587 5.05836 4.96814
Probability Pr(>|t|) 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000
1.63869 334 0.92 563.848538 0.0000
5 year return period Variable (Intercept) d_X d_Y d_Coast Z9 d_urb_pop_200000 d_Y2 Residual standard error Degrees of freedom Multiple R2 F statistic Probability (F statistic)
Regression coefficient 170.8936280355 -0.0005394495 -0.0245399413 0.0031430110 -0.0094392499 0.0037300827 0.0000014842
Standard error 9.99389 0.00008 0.00438 0.00038 0.00028 0.00072 0.0000
t value 17.09979 -6.23330 -5.59324 8.11644 -32.64130 5.17833 3.12081
Probability Pr(>|t|) 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00196
1.84798 334 0.91 528.12117 0.0000
10 year return period Variable (Intercept) d_X d_Y d_Coast Z9 d_urb_pop_200000 d_Y2 Residual standard error Degrees of freedom Multiple R2 F statistic Probability (F statistic)
Regression coefficient 167.2099050980 -0.0005901345 -0.0214274631 0.0030665140 -0.0101775349 0.0040590787 0.0000010806 2.08250 334 0.89 464.39677 0.0000
79
Standard error 11.26217 0.00009 0.00494 0.00043 0.00032 0.00081 0.00000
t value 14.84703 -6.05105 -4.33384 7.02713 -31.23096 5.00040 2.01625
Probability Pr(>|t|) 0.00000 0.00000 0.00001 0.00000 0.00000 0.00000 0.04457
Climatic Zone 5 2 year return period Variable (Intercept) d_X d_Y Z9 d_X2 d_Y2 Residual standard error Degrees of freedom Multiple R2 F statistic Probability (F statistic)
Regression coefficient 172.1284473407 -0.0022105458 -0.0277088791 -0.0084252218 0.0000004237 0.0000019549
Standard error 8.71767 0.00034 0.00376 0.00027 0.00000 0.00000
t value 19.74475 -6.48069 -7.36438 -31.10557 8.10837 4.77601
Probability Pr(>|t|) 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000
1.9894 663 0.81 571.17325 0.0000
5 year return period Variable (Intercept) d_X d_Y Z9 d_X2 d_Y2 Residual standard error Degrees of freedom Multiple R2 F statistic Probability (F statistic)
Regression coefficient 180.5607277459 -0.0021701535 -0.0293761400 -0.0099389837 0.0000004311 0.0000020294
Standard error 10.35032 0.00040 0.00446 0.00032 0.00000 0.00000
t value 17.44492 -5.35870 -6.57595 -30.90619 6.94836 4.17590
Probability Pr(>|t|) 0.00000 0.00000 0.00000 0.00000 0.00000 0.00003
2.36199 663 0.79 504.01075 0.0000
10 year return period Variable (Intercept) d_X d_Y Z9 d_X2 d_Y2 Residual standard error Degrees of freedom Multiple R2 F statistic Probability (F statistic)
Regression coefficient 186.1694837095 -0.0021426282 -0.0304918681 -0.0109420068 0.0000004359 0.0000020800 2.69442 663 0.78 442.94389 0.0000
80
Standard error 11.80708 0.00046 0.00509 0.00036 0.00000 0.00000
t value 15.76761 -4.63796 -5.98355 -29.82717 6.15886 3.75196
Probability Pr(>|t|) 0.00000 0.00000 0.00000 0.00000 0.00000 0.00019
Annex 6. Metadata for the annual maximum heat wave index distribution layers (two, five and ten year return periods) Dataset title
Spatial distribution of annual maximum heat wave index for the WHO Regions (Africa, Eastern Mediterranean and part of Europe)
Theme keywords
WHO, Africa, Eastern Mediterranean, Europe, natural disaster, Geographic Information System (GIS), natural hazard, heat wave, heat index, apparent temperature, relative humidity
Dataset topic category
Heat wave
Geographic location
The layer cover a total of 100 countries (22 for the Eastern Mediterranean, 46 for Africa and 32 for Europe)
Publication date
20110401
Data exchange format
ArcView grid
Filename
st_hw_2_ucl , st_ hw _5_ucl, st_ hw _10_ucl
Dataset edition
Second edition
Abstract
This dataset contains the spatial distribution of the annual maximum heat wave index for two, five and ten year return periods over the WHO Regions (Africa, Eastern Mediterranean and part of Europe)
Lineage
The process used to create the annual maximum heat wave index for two, five and ten year return periods distribution layer is described in the Methodology and implementation process for modelling the spatial distribution of heat wave hazard document that can be found in the first volume (2nd edition) of the WHO e-atlas of disaster risk for the WHO Regions (Africa, Eastern Mediterranean and part of Europe)
81
Data quality comments
Please refer to the data specific metadata for more information regarding the quality of the grids used as input for the creation of the annual maximum heat wave index grid Because of the methods and resolution used (1 km) special care should be taken when using this dataset for application below the national level Even though the dataset covers all of the WHO Regions, the lack of historical data for many countries reduces the quality of the dataset for those areas
Distributor
WHO Mediterranean Centre for Health Risk Reduction (WMC)
Spatial representation type
grid
Map projection
Unprojected (Geographic)
Reference system
WGS 84 datum
Geographic box
X min: –25.358747°, X max: 91.8287°
Resolution
Y min: –46.978931°, Y max: 63.459827°
Redistributions constraints
The annual maximum heat wave index distribution layers are copyrighted. The owner of the data agrees to the use, reproduction, distribution, display, publication and dissemination at no cost to third parties of the annual maximum heat wave index layer, in any manner and in any form whatsoever, subject to the copyright and acknowledgement mentioned in these metadata
Access and use constraints
These layers may not be reproduced, changed, adapted, translated, stored in a retrieval system or transmitted in any form or by any means without prior permission of the copyright holder, except to make a security backup. Requests for permissions, with a statement of purpose and extent, should be address to the VRAM programme at the WHO Mediterranean Centre for Health Risk Reduction (
[email protected])
Acknowledgement
WHO e-atlas of disaster risk for the WHO Regions (Africa, Eastern Mediterranean and part of Europe) 2nd edition. Copyright © WHO 2011. All rights reserved
Disclaimer
All reasonable precautions have been taken by WHO to produce these layers. However these layers are being distributed without warranty of any kind, either express or 82
implied regarding their content. The responsibility for their interpretation and use lies with the user. In no event shall the World Health Organization be liable for damages arising from their use Dataset language
English
Dataset character set
ASCII
Metadata provider
WHO Mediterranean Center for Health Risk Reduction (WMC)
Metadata contact
El Morjani Zine El Abidine BP 3566 Poste Talborjt 80000 Agadir Morocco Telephone: +212 528 28 55 30 email:
[email protected]
Metadata date
20110401
Metadata language
English
Metadata character set
ASCII
Metadata standard
ISO 19115
83
Annex 7. Metadata for the heat wave hazard distribution layers (two, five and ten year return periods)
Dataset title
Spatial distribution of the intensity level of heat wave hazard for the WHO Regions (Africa, Eastern Mediterranean and part of Europe)
Theme keywords
WHO, Africa, Eastern Mediterranean, Europe, natural disaster, Geographic Information System (GIS), natural hazard, heat wave hazard, apparent temperature, relative humidity
Dataset topic category
Heat wave hazard
Geographic location
The layer covers a total of 100 countries (22 for the Eastern Mediterranean, 46 for Africa and 32 for Europe)
Publication date
20110401
Data exchange format
ArcView grid
Filename
st_hw_2_cl , st_ hw _5_cl, st_ hw _10_cl
Dataset edition
Second edition
Abstract
This dataset contains the spatial distribution of the intensity level of heat wave hazard for two, five and ten year return periods over the WHO Regions (Africa, Eastern Mediterranean and part of Europe) according to five intensity levels (very low, low, medium, high and very high)
Lineage
The process used to create the intensity level of heat wave hazard distribution maps is described in the Methodology and implementation process for modelling the spatial distribution of heat wave hazard document that can be found in the first volume (2nd edition) of the WHO e-atlas of disaster risk for the WHO Regions (Africa, Eastern Mediterranean and part of Europe)
84
Data quality comments
Please refer to the data specific metadata for more information regarding the quality of the grids used as input for the creation of the heat wave hazard grid Because of the methods and resolution used (1 km) special care should be taken when using this dataset for application below the national level Even though the dataset covers all of the WHO Regions, the lack of historical data for many countries reduces the quality of the dataset for those areas
Distributor
WHO Mediterranean Center for Health Risk Reduction (WMC)
Spatial representation type
grid
Map projection
Unprojected (Geographic)
Reference system
WGS 84 datum
Geographic box
X min: –25.358747°, X max: 91.8287° Y min: –46.978931°, Y max: 63.459827°
Resolution
30 arc-seconds (0.008333°)
Redistributions constraints
The intensity level of heat wave hazard distribution layer is copyrighted. The owner of the data agrees to the use, reproduction, distribution, display, publication and dissemination at no cost to third parties of the intensity level of heat hazard distribution layer distribution layer, in any manner and in any form whatsoever, subject to the copyright and acknowledgement mentioned in this metadata
Access and use constraints
These layers may not be reproduced, changed, adapted, translated, stored in a retrieval system or transmitted in any form or by any means without prior permission of the copyright holder, except to make a security backup. Requests for permission, with a statement of purpose and extent, should be address to the VRAM programme at the WHO Mediterranean Centre for Health Risk Reduction (
[email protected])
Acknowledgement
WHO e-atlas of disaster risk for the WHO Regions (Africa, Eastern Mediterranean and part of Europe) 2nd edition. Copyright © WHO 2011. All rights reserved
85
Disclaimer
All reasonable precautions have been taken by WHO to produce these layers. However these layers are being distributed without warranty of any kind, either express or implied regarding their content. The responsibility for their interpretation and use lies with the user. In no event shall the World Health Organization be liable for damages arising from their use.
Dataset language
English
Dataset character set
ASCII
Metadata provider
WHO Mediterranean Centre for Health Risk Reduction (WMC)
Metadata contact
El Morjani Zine El Abidine BP 3566 Poste Talborjt 80000 Agadir Morocco Telephone: +212 528 28 55 30 email:
[email protected]
Metadata date
20110401
Metadata language
English
Metadata character set
ASCII
Metadata standard
ISO 19115
86
