How climate seasonality modifies drought duration ... - AGU Publications

Journal of Geophysical Research: Atmospheres RESEARCH ARTICLE

How climate seasonality modifies drought duration and deficit

10.1002/2013JD020383

A. F. Van Loon1 , E. Tijdeman1 , N. Wanders2 , H. A. J. Van Lanen1 , A. J. Teuling1 , and R. Uijlenhoet1

Key Points: • Meteorological drought: no seasonality effects on drought characteristics • Soil moisture drought: divergent pattern in semiarid and monsoon climates • Hydrological drought: bimodal pattern in savannah and cold climates

1 Hydrology and Quantitative Water Management Group, Wageningen University, Wageningen, the Netherlands,

Correspondence to: A. F. Van Loon, [email protected]

Citation: Van Loon, A. F., E. Tijdeman, N. Wanders, H. A. J. Van Lanen, A. J. Teuling, and R. Uijlenhoet (2014), How climate seasonality modifies drought duration and deficit, J. Geophys. Res. Atmos., 119, 4640–4656, doi:10.1002/2013JD020383.

Received 17 JUN 2013 Accepted 18 JAN 2014 Accepted article online 23 JAN 2014 Published online 30 APR 2014

2 Department of Physical Geography, Faculty of Geosciences, Utrecht University, Utrecht, the Netherlands

Abstract

Drought propagation through the terrestrial hydrological cycle is associated with a change in drought characteristics (duration and deficit), moving from precipitation via soil moisture to discharge. Here we investigate climate controls on drought propagation with a modeling experiment in 1271 virtual catchments that differ only in climate type. For these virtual catchments we studied the bivariate distribution of drought duration and standardized deficit for the variables precipitation, soil moisture, and discharge. We found that for meteorological drought (below-normal precipitation), the bivariate distributions of drought characteristics have a linear shape in all climates and are thus not affected by seasonality in climate. Despite the linear shape of meteorological drought, soil moisture drought (below-normal storage in the unsaturated zone) and hydrological drought (below-normal water availability in aquifers, lakes, and/or streams) show strongly nonlinear shapes in drought characteristics in climates with a pronounced seasonal cycle in precipitation and/or temperature. These seasonality effects on drought propagation are found in monsoonal, savannah, and Mediterranean climate zones. In these regions, both soil moisture and discharge show deviating shapes in drought characteristics. The effect of seasonality on drought propagation is even stronger in cold seasonal climates (i.e., at high latitudes and altitudes), where snow accumulation during winter prevents recovery from summer hydrological drought, and deficit increases strongly with duration. This has important implications for water resources management in seasonal climates, which cannot solely rely on meteorology-based indices as proxies for hydrological drought duration and deficit and need to include seasonal variation in both precipitation and temperature in hydrological drought forecasting.

1. Introduction Drought is a natural disaster, resulting in severe economic and societal problems around the world [Seneviratne et al., 2012]. It has negative impacts on, amongst others, navigation, crop production, ecosystems, hydropower production, water supply for drinking water, irrigation, and cooling water for industry. These impacts are diverse, but have in common that they are characterized by a water deficit compared to normal conditions, albeit in different parts of the terrestrial hydrological cycle. Drought is a complex phenomenon and has therefore been defined in many ways [Tallaksen and Lanen, 2004; Mishra and Singh, 2010; Sheffield and Wood, 2011]. In this research we define drought as a sustained period of below-normal water availability. We generally distinguish three types of drought, namely, meteorological drought (a precipitation deficit), agricultural or soil moisture drought (a below-normal storage in the unsaturated zone), and hydrological drought (a below-normal water availability in aquifers, lakes, and/or streams). Hydrological drought has a variety of causes ranging from precipitation deficiency to prolonged frost conditions [Sheffield and Wood, 2011] to increased evapotranspiration [Teuling et al., 2013]. The translation of a drought signal from deviating meteorological conditions into soil moisture and/or hydrological drought is called drought propagation [Tallaksen and Lanen, 2004].

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.

VAN LOON ET AL.

For water resources management this drought propagation is of paramount importance as the impacts of drought (see above) are related to soil moisture drought and hydrological drought and are not caused directly by meteorological drought. Meteorological drought has, however, been most studied [Dai, 2013], mainly because meteorological information is most easy to obtain, especially on larger scales (e.g., global scale). The relation between meteorological conditions and the drought types that cause the impacts mentioned above is, however, often not straightforward [Sheffield et al., 2012]. Drought propagation strongly depends on climate and catchment characteristics [Van Loon and Van Lanen, 2012], and consequently, hydrological drought characteristics show a pronounced variation around the globe [Van Lanen et al., 2013]. Many studies investigated drought propagation on the catchment scale [Eltahir and Yeh, 1999; Peters et al., 2006; Tallaksen et al., 2009; Teuling et al., 2013] and the regional scale [Vidal et al., 2010; Hannaford et al., ©2014. The Authors.

4640

Journal of Geophysical Research: Atmospheres

10.1002/2013JD020383

2011; Stewart et al., 2011]. These studies have resulted in a more thorough understanding of the processes underlying drought propagation. However, the geographical reference is limited to the catchment or region under study, and a further generalization and extension of process knowledge to the global scale is needed, as is advocated by Mishra and Singh [2010] and in the recent Intergovernmental Panel on Climate Change report on extremes by Seneviratne et al. [2012]. From the catchment-scale studies we learnt that seasonality in climate can lead to very severe drought events. Severity of drought events is commonly measured in terms of duration of the event and the missing water or water deficit [Tallaksen et al., 2009]. These two characteristics are related [Hisdal et al., 2004], but this relation has not yet been systematically analyzed considering the diversity in climate around the globe. In the selected catchments in Van Loon and Van Lanen [2012], durations of seasonal droughts were typically between 200 and 300 days, and deficits depended on average catchment wetness (e.g., deficits were around 3 mm for a Spanish catchment and around 50 mm for a Norwegian catchment). For comparison on the global scale standardization of the deficit is therefore needed. In this study, we investigated the role of seasonal drought types on the duration and standardized deficit of droughts in climates with different degrees of seasonality. We intended to find out how duration and standardized deficit of drought events are related for the different drought types meteorological drought, soil moisture drought, and hydrological drought. The aim of this study is to investigate the effects of seasonality in climate on drought propagation. We performed a modeling experiment in a large number of virtual catchments that differ only in climate so that we can study differences in drought duration and deficit between climates with different degrees of seasonality.

2. Methodology 2.1. Model We studied the relation between drought duration and deficit through a controlled modeling experiment in virtual catchments. The use of a synthetic model is well suited for this purpose, because it allows for the isolation of climate effects from effects of catchment properties and easy sensitivity analyses. The basis of the modeling experiment was a conceptual hydrological model that combines a degree day snow accumulation model, a soil water balance model, and a groundwater model based on linear reservoir theory [Van Lanen et al., 2013]. The snow accumulation model uses the snow calculation approach of the well-known conceptual hydrological model Hydrologiska Byråns Vattenbalansavdelning (HBV) [Seibert and Vis, 2012]. Snow accumulation takes place when the temperature is lower than a threshold temperature. Snow melt starts when the temperature is higher than the threshold temperature and is calculated using a degree day model, i.e., the daily temperature difference (temperature minus threshold temperature) is multiplied by the degree day factor to yield the daily amount of snow melt. The snow cover retains melt water until a certain level and refreezing of this water in the snow cover can take place. Snow melt and rainfall are directly added to the soil water balance model to simulate water storage in and fluxes to and from the unsaturated zone. Parameters for minimum and maximum soil moisture storage (wilting point and field capacity) and a critical storage are assigned based on soil type. Evapotranspiration from the soil is at potential level as long as the soil moisture storage is between field capacity and the critical level. Below the critical level, actual evapotranspiration is calculated by multiplying potential evapotranspiration with a reduction factor based on actual soil moisture storage. Actual evapotranspiration becomes zero at wilting point. More information on the soil water balance model and the parameters involved can be found in Van Lanen et al. [2013]. Percolation from the soil water balance model to the groundwater model is a function of soil moisture levels. If soil moisture is at field capacity, all excess water is drained to the groundwater. If soil moisture is between field capacity and critical level, the recharge flux is reduced in a way similar to the HBV model [Seibert and Vis, 2012]. If soil moisture is below critical level, recharge is zero. The groundwater model uses the De Zeeuw-Hellinga approach [Kraijenhoff van de Leur, 1962; Ritzema, 1994] to calculate groundwater discharge (Q) with a daily time step: Qt = Qt−1 ⋅ e

VAN LOON ET AL.

©2014. The Authors.

− 1j

+ Rcht ⋅ (1 − e

− 1j

)

(1) 4641


10.1002/2013JD020383

in which Q is in mm d−1 , Rcht is recharge (in mm d−1 ), and j is a response parameter (in day). For naturally drained aquifers, a general interpretation of the response parameter is [Kraijenhoff van de Leur, 1958; Birtles and Wilkinson, 1975] j∼

𝜇 ⋅ L2 kD

(2)

in which kD is transmissivity of the groundwater system (in m2 d−1 ), 𝜇 is the storage coefficient of the groundwater system (–), and L is distance between streams (in m). These three parameters allow the calculation of the response parameter for a given groundwater system so that fast-responding systems have a low j value, and slow-responding systems have a high j value. More information on the groundwater model and the parameters involved can be found in Van Lanen et al. [2013]. No channel routing was included in the hydrological model, so the discharge simulated by this model can be characterized as subsurface discharge. 2.2. Data and Parametrization The combined hydrological model was forced by daily time series of precipitation, temperature, and potential evapotranspiration taken from a global reanalysis data set (Water and Global Change (WATCH) Forcing Data (WFD)) [Harding et al., 2011; Weedon et al., 2011; WATCH, 2012]. The WFD data set consists of gridded time series of meteorological variables. The data have a spatial resolution of 0.5◦ based on the Climatic Research Unit (CRU) land mask. The WFD originate from modification of the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA-40 reanalysis data [Uppala et al., 2005]. The data have been interpolated and corrected for the elevation differences between the grids of ERA-40 and CRU. For precipitation, the ERA-40 data were first adjusted to have the same number of wet days as CRU [Brohan et al., 2006]. Next, the data were bias-corrected using monthly Global Precipitation Climatology Centre (GPCC) precipitation totals [Schneider et al., 2008], and finally, gauge-catch corrections were applied. For temperature, the ERA-40 data were bias-corrected using CRU monthly average temperatures and temperature ranges. For more information the reader is referred to Weedon et al. [2011]. The WFD was chosen in this study because it is considered to provide a good representation of “real meteorological events, synoptic activity, seasonal cycles, and climate trends” [Weedon et al., 2011]. The WFD represent state-of-the-art bias correction, not only for precipitation but also for other variables that are important in hydrological modeling, e.g., temperature and variables needed for the calculation of potential evapotranspiration. Some studies using the WFD have shown that the bias correction method is robust [Piani et al., 2010; Hagemann et al., 2011] and that it improves simulations of river flow in different regions compared to other large-scale data sets (not bias-corrected climate data in Hagemann et al. [2011] and Tropical Rainfall Measuring Mission (TRMM) in Li et al. [2013]). The hydrological model was run for the period 1958–2001 for grid cells of 0.5◦ × 0.5◦ across the world, which represent virtual catchments. To ensure that different climates are well represented, we randomly selected a number of grid cells within each Köppen-Geiger climate type [Geiger, 1961; Peel et al., 2007] on the basis of the relative area of each climate type on the globe. Excluding extremely dry (desert, BW) and extremely cold (glacier, EF) climates, this resulted in a final selection of 1271 grid cells distributed across the remaining 27 Köppen-Geiger climate types (Figure 1). Van Lanen et al. [2013] demonstrated that this number is more than sufficient as a representative sample to study drought characteristics on a global scale. In order to isolate climate effects, which are the focus of this study, from effects of catchment properties, all grid cells were assigned the same catchment characteristics based on average conditions for land use, soil, and groundwater. The grid cells were in this way treated as spatially lumped virtual catchments. The chosen combination of climate type and catchment properties can occur in the selected grid cell, but it is not georeferenced to that particular cell. Parameters of the model were chosen in accordance with recommendations for an average situation, e.g., a threshold temperature for snowfall and snow melt of 0◦ C and a groundwater response parameter (j value) of 250 days [Seibert, 2000; Van Lanen et al., 2013]. Land use was also assumed to be identical in all virtual catchments, i.e., permanent grassland with a rooting depth of 50 cm. Soil parameters were derived from a standard series of existing soils that predominantly differ in soil texture [Wösten et al., 2001a,2001b]. We used moisture retention data to select a soil type with a medium soil moisture supply capacity, i.e., a light silty loam soil. Soil moisture retention characteristics for this soil type were applied to all selected grid cells. To check the sensitivity of our conclusions to these assumptions of average conditions for land properties we performed a sensitivity analysis (section 2.4). VAN LOON ET AL.


4642


10.1002/2013JD020383

Figure 1. Koppen-Geiger subclimate zones including selected grid cells, representing 1271 virtual catchments (for a description of the climate types, see Peel et al. [2007], adapted from Wanders et al. [2010]).

The hydrological model was run once for the entire modeling period to get good initial values for the stores (soil moisture and especially groundwater). After these 43 years the stores adapted to the imposed climate, providing representative initial conditions for the real model run. No maximum groundwater storage was applied, which assumes rather deep water tables in the virtual catchments (see for discussion section 4.3). The adopted approach generated possible time series for soil moisture storage (SM) and subsurface discharge (Q; further called discharge for brevity) for 1271 virtual catchments with different climate types. Additionally, we used time series of precipitation (P) from the WATCH Forcing Data for further analysis. 2.3. Drought Analysis and Density Fields From the time series of P, SM, and Q we identified droughts with the widely used threshold level approach [Hisdal et al., 2004; Fleig et al., 2006; Sheffield and Wood, 2007], defining a drought when a variable is below a predefined threshold. This threshold could be fixed throughout the year, but that would imply that yearly recurring low-flow periods (summer in warm climates and winter in cold climates) [Laaha and Blöschl, 2006; Burn et al., 2008] would be marked as drought. Since these events do not correspond to the definition of “below-normal water availability,” a variable threshold is more suitable for drought studies on global scale. A seasonal threshold usually has too little variation to capture the normal hydrological regime, and a daily threshold is too much influenced by daily variations. Therefore, a monthly threshold is regarded as a good intermediate option. In this study we used a smoothed monthly threshold based on the 80th percentile of monthly duration curves of P (after applying a 10 day moving average), SM, and Q (see Figure 2). The smoothing of the monthly threshold (by a 30 day moving average) was needed to remove the stepwise pattern and to prevent artifact droughts at the beginning or end of a month. This procedure is comparable to that applied in many other drought studies [e.g., Hisdal et al., 2004; Vidal et al., 2010] and is assumed to be robust for the analysis of drought propagation, especially when taking into account seasonality [Van Loon and Van Lanen, 2012]. Drought characteristics that are most important to water managers and water users are drought duration and a measure of the severity of the drought. For fluxes (i.e., precipitation and discharge) the most commonly used severity measure is deficit (D), calculated by adding up the deviations between the actual flux and the threshold level over the drought period (Figure 2) [Hisdal et al., 2004; Fleig et al., 2006]. The deviation (d) from the threshold on time t is calculated according to { d(t) =

VAN LOON ET AL.


𝜏(t) − x(t) if x(t) < 𝜏(t) 0 if x(t) ≥ 𝜏(t)

(3)

4643


10.1002/2013JD020383

discharge

deficit

duration

drought events

time (e.g. one year) Figure 2. Example of the application of the threshold level method to a time series of discharge, including an illustration of drought characteristics duration and deficit (adapted from Van Loon [2013]; FDC = flow duration curve).

in which x(t) is the hydrometeorological variable on time t and 𝜏(t) is the threshold level of that hydrometeorological variable on time t (all in mm d−1 , t is measured in days). The deficit of drought event i (Di ) is then given by Di =

T ∑

d(t) ⋅ Δt

(4)

t=1

in which Di is in mm. This deficit is standardized as follows: Dsi = Di ∕ x

(5)

in which Dsi is the standardized deficit of drought event i (in days) and x is the mean of hydrometeorological variable x(t). The physical interpretation of standardized deficit is the number of days with mean flow required to reduce the deficit to zero. Standardized deficit is further called deficit for brevity. For state variables (i.e., soil moisture and groundwater storage) deficit is physically meaningless [Tallaksen et al., 2009], but it still gives a good indication of the severity of a drought event. In this study we use deficit also for soil moisture drought because it enables comparison with other hydrological variables, precipitation and discharge.

100 25 5 1

P =1.46 SM=1.02 Q =0.72 0

Standardized deficit (days)

250

We transformed the data by taking the fourth root of both duration and deficit for purposes of visualization. A logarithmic transformation would exclude zero deficit values. Similar to Kim et al. [2003], nonparametric kernel density estimators [Wand and Jones, 1995] were Cfa−climate adopted to determine smoothed bivariate probability density fields of the drought characteristics. To exclude outliers (in the high, low, and intermediate ranges of the drought characteristics), the area contained within the 90% contour of the probability density field was selected.

1

30

120

360

900

Duration (days) Figure 3. Contour plot of the 90% probability density fields of transformed drought duration and standardized deficit in precipitation (blue), soil moisture (red), and discharge (green) for the humid subtropical climate (Cfa climate), including linear regression lines and the slope of these lines (𝛼 ).

VAN LOON ET AL.


As an example, the contour lines of the 90% probability density fields of droughts in P, SM, and Q of the humid subtropical climate (Cfa climate) are presented in Figure 3. This subclimate type was chosen because climate seasonality is assumed to be minimal in this climate type. Temperature is relatively constant and does not drop below zero for longer periods so that no snow accumulation and snow melt seasons can be distinguished. Precipitation is equally distributed over the year pointing at a climate without monsoon influence [Peel et al., 2007]. In this Cfa climate, the density fields of all variables have an elongated shape and a clear orientation to one direction, which we henceforth call a linear shape (Figure 3). This shape shows a positive relation between deficit and 4644


10.1002/2013JD020383

duration in all three variables. The variable Q exhibits a longer density field than P and SM, indicating that many droughts in Q have a longer duration and a higher standardized deficit. As reported in the literature [Peters et al., 2006; Tallaksen et al., 2009], this is a result of propagation of drought through the hydrological cycle and is caused by pooling and lengthening. Pooling is the merging of meteorological droughts into a prolonged hydrological drought, and lengthening is the increase in drought duration moving from meteorological via soil moisture to hydrological drought [Van Loon and Van Lanen, 2012; Van Loon, 2013]. The linear shape indicates that one governing process is dominant, namely, a rainfall deficit propagating into a soil moisture drought and hydrological drought (termed classical rainfall deficit drought in Van Loon and Van Lanen [2012]). To quantify the orientation of the density fields, we fitted a linear regression line through the drought events within the density field (visualized by the dashed lines in Figure 3). We used a fixed origin at (duration, deficit) = (1,0), which is the theoretical minimum for a drought with the applied drought identification method. The slope of the regression line (𝛼 ) was used for comparison between hydrometeorological variables and climate types. The smaller this 𝛼 is, the smaller the increase of standardized deficit with duration. In the example in Figure 3, 𝛼 decreases from 1.46 for P to 1.02 for SM and 0.72 for Q. This means a reduced increase of deficit with duration when moving through the hydrological cycle, which is also a result of drought propagation [Peters, 2003]. It is caused by attenuation, i.e., the damping effect of stores (soil moisture and groundwater) on the drought signal [Van Loon and Van Lanen, 2012; Van Loon, 2013]. Van Lanen et al. [2013] studied similar bivariate probability density fields of Q drought characteristics in the five major climate types for a number of realizations of catchment characteristics (among which a similar configuration as used in this study). They found that the Q density fields of the A and C climates had the most narrow elongated shape, while the Q density fields of the B and E climates were slightly wider. The D climate showed an intermediate behavior. Based on the outcomes of Van Lanen et al. [2013], the shape of the Cfa climate in Figure 3 can be regarded as the reference shape for this research. In this study, we investigate the effect of seasonality in climate on drought characteristics by comparing density fields of subclimate types with the density fields in Figure 3. More detailed information about the methodology can be found in Van Lanen et al. [2013], Tijdeman et al. [2012], and Van Loon [2013]. 2.4. Sensitivity Analysis In this study we assigned the same catchment characteristics to all grid cells with the purpose of studying only climate effects in virtual catchments. To investigate how sensitive our conclusions are to changes in catchment properties we performed a sensitivity analysis. Because Van Lanen et al. [2013] showed that hydrological drought duration and deficit were highly sensitive to the parameters of the groundwater model and almost insensitive to the parameters of the soil water balance model, we decided to focus only on the groundwater model for this sensitivity analysis. Ten values of the response parameter ( j) were tested ranging from extremely fast responding ( j = 20 days) to extremely slow responding ( j = 1500 days). This selection of j values is expected to represent the variability of groundwater system responses that can be found in reality. Van Lanen et al. [2013] selected a response parameter of 100 days for a fast-responding system and a response parameter of 1000 days for a slow-responding system, so in this study we considered even more extreme cases. Other model parameters (snow accumulation model and soil water balance model) were fixed.

3. Results 3.1. Seasonality Effects on Precipitation Drought Figure 4 displays density fields of a selection of subclimate types. The orientation of the precipitation density fields in these subclimates is very similar, indicating no apparent climate effect on P drought characteristics. This is not according to our expectations. We expected that regions that are governed by completely different climate systems, e.g., tropical convective rainfall areas versus temperate western climates dominated by low-pressure systems, would have different drought duration and deficit. But apparently, these effects are minor when averaging over 10 days and looking at relative deviations from a variable threshold. So, P drought duration and deficit are a measure of long-term variability in precipitation, which is comparable across the globe. This shows that for droughts, small-scale variations in precipitation characteristics are not nearly as important as for floods [Gaál et al., 2012]. VAN LOON ET AL.


4645

1 5 25 100

0

1 5 25 100

10.1002/2013JD020383

0



1

30 120 360 900 1

30 120 360 900 1

30 120 360 900 1

30 120 360 900 1

30 120 360 900

Duration (days) Figure 4. Contour plot of the 90% probability density fields of drought duration and deficit in P, SM, and Q for a selection of subclimates, including linear regression lines, and their slope (𝛼 ). Plots of all subclimates are included in Appendix A.

In Figure 4, only a slight difference in shape is visible in climate types with a strong seasonality in precipitation (e.g., BSh and Cwa climates). In these climates, high variability in precipitation between wet and dry seasons causes more variation in standardized deficit and, therefore, a slightly wider density field of P drought characteristics. 3.2. Seasonality Effects on Soil Moisture Drought The soil moisture density fields of some subclimates (e.g., Af, Cfb, and Dfb climates in Figure 4) exhibit a linear shape, comparable to that of the reference shape in Figure 3. Some other subclimates (e.g., Aw climate) show a divergent shape in the SM density field, i.e., at longer durations the density field becomes wider, with even a separate part of the SM density field near the x axis in extreme cases (e.g., BSh and Cwa climates). This results in a large variability in the slope of the regression line between different subclimate types (𝛼 of SM in Figure 4). This difference in drought characteristics of soil moisture is related to seasonality in precipitation, as indicated by the second letter in the climate type name, and how that interacts with soil moisture storage. The letter “f” in the climate type name refers to significant precipitation in all seasons [Geiger, 1961]. In these climate types meteorological droughts can cause soil moisture droughts in all seasons. As the soil moisture threshold is relatively constant (like in the Cfa climate in Figure 5, fourth row), SM droughts in these climate types have similar drought characteristics, i.e., drought duration and deficit are strongly linearly related. This results in a narrow density field, comparable to the one of SM in Figure 3. The letters “s” and “w” in the climate type name denote climates with a dry summer or winter, respectively [Geiger, 1961]. In these climate types, the threshold has a strong seasonality (like in the Aw climate in Figure 5, fourth row), which has implications for SM drought development. In the example Aw climate (Figure 5), a meteorological drought in the wet season (e.g., June–July 1959) results in a SM drought with a large increase of deficit with duration, because the threshold level is high enough to ensure that SM droughts are not limited by the wilting point but can develop freely (e.g., June–August 1959). In the dry season, however, the threshold level is low, and SM droughts are bounded by the wilting point (e.g., November–March 1959). This causes limited deviation from the threshold and smaller standardized deficit values than expected based on drought duration. Occurrence of drought events in wet and dry seasons results in a divergent shape of drought characteristics for these dry seasonal climate types. To distinguish these drought types in the density fields, we subdivided SM droughts for which soil moisture storage reaches wilting point and those for which it did not. In Figure 6 (first row, Aw climate), the droughts for which SM reaches wilting point comprise the lower half of the cloud of points, showing that in the dry season droughts with the same duration have a lower deficit than in the wet season. In some climate types (e.g., the BSh climate and climate types with the letter “a” as third letter, denoting a hot summer [Geiger, 1961], such as the Cwa climate and Dsa climate), a separate part of the SM density field near the x axis is visible (Figure 4). In these climates seasonality in precipitation is complemented with VAN LOON ET AL.


4646

Jul63

Jan64

Jul64

Jan65

80

Dec58

Jun59

Dec59

Jun60

Dec58

Jun59

Dec59

Jun60

0

Jul63

Jan64

Jul62

Jan63

Jul63

Jan64

Jul62

Jan63

Jul63

Jan64

Jul62

Jan63

Jul63

Jan64

0

soil moisture [mm]

Jan62

Jan62

discharge [mm/d]

Jun58

Jan62

precipitation [mm/d]

15 10 5


0 140

soil moisture [mm]

40 80 4 3 2 1 0

discharge [mm/d]

Jun58

−20

temperature [°C] Jun60

−50

35 25

Dec59

0

40 6 4 2 0 1.2 0.6 Jan63

Jun59

Jan63

80

Jan65

Jun58

Jul62

40

Jul64

Dec58

snow [mm]

Jan64

Jun60

Jan64

3.0

Jul63

0.0

discharge [mm/d]

Jan63

Dec59

Jul63

2.0

Jan65

Jun59

Jan63

1.0

Jul64

Jun58

Jul62

0.0

Jan64

Dec58

Jan62

100 150

Jul63

140

Jan63

Jun60

50

Jan65

Dec59

0

Jul64

Jun59

0.30

Jan64

40 80

soil moisture [mm]


Jan63

Dec58

Dwd

0.15

Jul63

Jun58

10.1002/2013JD020383

0.00

Jan65

15

temperature [°C] Jul64

snow [mm]

Jan64

80

40 20

Jul63

0

snow [mm]

Jan63

Aw

40

Cfa

0

temperature [°C]


Jan62

Figure 5. Time series of hydrometeorological variables of the reference humid subtropical climate (Cfa), a typical tropical savanna climate (Aw), and a typical continental subarctic climate with extremely severe winters (Dwd).

a strong seasonality in temperature, resulting in high evapotranspiration, even lower threshold values, and consequently lower standardized deficit of SM droughts in the dry season. The shape of the SM density field is quantified by the correlation (R2 ) of the SM drought events so that a high R2 reflects a more clustered SM density field, and a low R2 reflects a more divergent density field. The spatial distribution of this R2 is plotted in Figure 7. We clearly see highest correlations in climates with significant precipitation in all seasons, especially at higher latitudes on the Northern Hemisphere, and lowest correlations in climates with strong seasonality, e.g., the savannah (Aw), steppe (BSh), monsoonal (Cwa), and Mediterranean (Csa) climates in most of Africa, southern Asia, Brazil, and Middle America. The warmer and more seasonal the subclimate, the more limited drought deficit development in SM in the dry season and the more divergent the shape of the SM density field. 3.3. Seasonality Effects on Hydrological Drought The density field of discharge of many subclimates exhibits a linear shape (e.g., Af and Cfb climates in Figure 4), indicating regular hydrological drought development like in the reference shape of the Cfa climate in Figure 3. Although all subclimate types have similar values for the slope of the regression line (𝛼 of Q in Figure 4), a striking feature in some other subclimates is a change in direction of the Q density field around 120–150 days (Aw, BSh, Cwa, Dwd, and ET climates). Two distinct modes can be distinguished with different sensitivity of deficit with duration, i.e., longer Q droughts had a larger increase of deficit with duration than shorter Q droughts. This bimodal shape in Q drought characteristics is related to seasonality in both precipitation and temperature, influencing drought propagation. In Figure 5 (lower row) we can visually compare hydrological droughts in a warm seasonal climate (Aw climate) and a cold seasonal climate (Dwd climate) with the VAN LOON ET AL.


4647

250

Aw-climate

Dwd-climate

1

5

25

100

Cfa-climate

10.1002/2013JD020383

SM Droughts(NWPL) Droughts(WPL)

0



1

30

120

360

900

1

30

120

360

900

1

Duration (days)

30

120

360

900

Duration (days)

Aw-climate

Dwd-climate

1

5

25

100

Cfa-climate

Q Droughts(CRD) Droughts(RTS) Droughts(WTD)

0


250

Duration (days)

1

30

120

360

900

1

30

120

360

900

1

Duration (days)

30

120

360

900

Duration (days)

Aw-climate

Dwd-climate

1

5

25

100

Cfa-climate

Q

Q

Q

SD=0.73 LD=0.77

SD=0.77 LD=1.13

SD=0.86 LD=2.26

0


250

Duration (days)

1

30

120

360

900

1

Duration (days)

30

120

360

Duration (days)

900

1

30

120

360

900

Duration (days)

Figure 6. SM and Q density fields of all three climate types (Cfa, Aw, and Dwd) with drought types indicated (WPL = Wilting Point Limited, NWPL = Non Wilting Point Limited, CRD = classical rainfall deficit drought, RTS = rain-to-snow-season drought, and WTD = wet-to-dry-season drought based on Van Loon and Van Lanen [2012]), and Q density fields including regression lines for short-duration (SD) and long-duration (LD) droughts.

reference climate (Cfa climate). Hydrological droughts in the Cfa climate develop as a result of low rainfall and end due to high rainfall and are therefore all classified as classical rainfall deficit droughts (see section 2). In the Dwd climate, however, some hydrological droughts that develop in summer also end in summer (e.g., July 1962), but some continue throughout the winter (e.g., August 1962 to June 1963). In winter, chances of recovery from a hydrological drought that developed during the previous summer are extremely low, because all precipitation falls as snow, and no recharge takes place. Therefore, the recession of discharge stays below the threshold until the snow melt peak in spring (June 1963). These long multiseason droughts (termed rain-to-snow-season drought in Van Loon and Van Lanen [2012]) have a larger increase of deficit with duration, especially when snow melt is delayed. A comparable process is observed in warm subclimates. In warm subclimates with a distinct dry season (i.e., Aw, BSh, and Cwa climates), a hydrological drought in the wet season that ends by a precipitation peak in that same season results in a short drought with drought characteristics exhibiting a linear shape. A hydrological drought that does not end in the wet season continues throughout the dry season VAN LOON ET AL.


4648


10.1002/2013JD020383

90°N

60°N

30°N

0°N

30°S

60°S 180°W

90°W

0°E

90°E

180°E

R2 of soil moisture drought characteristics: < 0.68

0.68 - 0.88

0.88 - 0.92

> 0.92

Figure 7. Spatial distribution of the correlation between duration and deficit of SM drought events within the density fields (quantified by R2 ); subclimate types are distributed evenly over the four classes.

(July 1958 to September 1959 in the example Aw climate in Figure 5), because chances of recovery during the dry season are extremely low. The on-average low precipitation and high evapotranspiration in the dry season rarely result in recharge to the groundwater and Q stays below the threshold until a precipitation peak in the next wet season (e.g., October 1959). Even when a high precipitation peak occurs in the dry season this water is immediately lost to evapotranspiration and soil moisture replenishment (September–October 1958 and May 1959, Figure 5 (third and fourth rows)). The resulting long multiseason droughts (termed wet-to-dry-season drought in Van Loon and Van Lanen [2012]) have a larger increase of deficit with duration. These conclusions, which were based on visual inspection of the time series, were tested quantitatively on the larger scale by selecting drought events independently and plotting the resulting drought types in Figure 6 (middle). The criteria for selecting drought events of a certain type were based on Van Loon and Van Lanen [2012] and were only taken from weather variables to ensure an independent selection process. The rain-to-snow-season drought was defined as starting in summer (temperature above 1◦ C for at least 10 days) and continuing into winter (temperature below −1◦ C for at least 10 days), and the wet-to-dry-season drought was defined as starting in the wet season (long-term average precipitation higher than 50 mm for at least 10 days) and continuing into the dry season (long-term average precipitation lower than 10 mm for at least 10 days). In the Cfa climate (Figure 6, middle), only a few events are classified as such seasonal drought types, and they do not have much influence of the shape of the density field. In the Aw climate, wet-to-dry-season droughts make up the top right part of the graph, exhibiting a steeper relationship between deficit and duration. In the Dwd climate, rain-to-snow-season droughts also cluster on the top right part of the graph, causing an even steeper relationship between deficit and duration. This proves that in cold seasonal climates rain-to-snow-season droughts and in warm seasonal climates wet-to-dry-season droughts are responsible for the bimodal shape in drought characteristics. The strength of the bimodal shape is quantified by the difference between 𝛼 of short-duration droughts (6 months; with arbitrary origin), as displayed in Figure 6, third row. The difference between these slopes is a measure of the strength of the bimodal shape so that in subclimates with stronger bimodality, this 𝛼 difference is larger (e.g., 1.4 for the Dwd climate) than in sublimates with a less pronounced bimodality (e.g., 0.36 for the Aw climate) and than the reference climate (e.g., 0.05 for the Cfa climate). The spatial distribution of the 𝛼 difference is plotted in Figure 8. Highest values are found in the cold regions in the high northern latitudes and in mountainous areas, like the Himalayas, the Andes, and the Rocky Mountains. The steppe (BSh), savannah (Aw), and Mediterranean (Csb) subclimates (mainly on the Southern Hemisphere) show intermediate values, and the temperate subclimates (Cfa, Cfb, and Cfc) of Europe and the eastern part of the United States have lowest values. In cold seasonal subclimates, the 𝛼 difference between short- and long-duration droughts is larger than in dry seasonal subclimates, indicating a more pronounced bimodal shape in the Q density field in cold seasonal VAN LOON ET AL.


4649


10.1002/2013JD020383

90°N

60°N

30°N

0°N

30°S

60°S 180°W

90°W

0°E

90°E

180°E

Slope difference between short and long Qsub droughts: < 0.12

0.12 - 0.27

0.27 - 0.53

> 0.53

Figure 8. Spatial distribution of the differences between the slope of the linear regression line (𝛼 ) of long- and short-duration droughts; subclimate types are distributed evenly over the four classes.

subclimates than in dry seasonal subclimates. This leads to the conclusion that there is a lower chance of recharge during winter in cold climates than during the dry season in warm climates. 3.4. Sensitivity to Aquifer Properties To test if these patterns caused by climate seasonality remain when changing the catchment properties of our virtual catchments, we changed the parameters of the groundwater model from average conditions to both faster and slower-responding systems, i.e., lower and higher j values, respectively. For the Aw climate

Figure 9. Contour plot of the 90% probability density fields of drought duration and deficit in Q for the (a) Aw climate for a selection of groundwater response parameters ( j = 20–1500 days) and (b) the Dwd climate for a range of groundwater response parameters ( j = 20–1500 days). The response parameter used in the previous analyses ( j = 250 days) is indicated.

VAN LOON ET AL.


4650


10.1002/2013JD020383

(Figure 9a), density fields of discharge exhibited the expected bimodal shape for j values higher than 100 days. For faster-responding systems discharge drought characteristics are more and more similar to precipitation drought characteristics (see Figure 4). In the Dwd climate (Figure 9b), the bimodal shape in discharge density fields is present for all j values, ranging from fast to slow-responding systems. The density fields do have different shapes, but the bimodality is still clearly visible. We also recognize the expected flattening of the density fields with higher j values, both in the Aw and Dwd climate. This is in accordance with Van Lanen et al. [2013], who found that in slow-responding systems hydrological droughts have a longer Figure 10. The 𝛼 difference of discharge droughts for a range of duration and lower standardized deficit, groundwater response parameters (j = 100–1500 days) for the resulting in a less tilted density field. This selection of subclimate types. Note that the x axis is not equally flattening is the reason that the 𝛼 difference spaced, points represent separate runs, and lines are added just for is decreasing for higher j values (Figure 10). visualization purposes. But, even though the different selected subclimates get closer together with increasing j values, the order remains the same: no bimodal shape for the nonseasonal climate types Cfb and Dsa, a slight bimodal shape for the warm seasonal climates Aw, BSh, and Cwa, and a strong bimodal shape for the cold seasonal climates Dwd and ET.

4. Discussion In this section we compare our results with documented historical drought events, and we discuss the choices made for the modeling experiment in virtual catchments. 4.1. Documented Drought Events Maps quantifying the patterns of droughts in soil moisture and discharge show that the nonlinear effects found in the bivariate density plots of drought duration and deficit are more pronounced when the climate is more seasonal (Figures 7 and 8). The effect of seasonality on soil moisture drought is mostly visible in the warm seasonal climates (like monsoonal, Mediterranean, and semiarid climates), and the effect of seasonality on hydrological drought is most pronounced in the cold seasonal climates (such as Boreal climates). These findings are consistent with smaller-scale drought studies [Vidal et al., 2010; Van Loon and Van Lanen, 2012] and with observational evidence from recent severe drought events, like the 2011 drought in the Horn of Africa [Viste et al., 2012] and the 2009–2010 winter drought in Europe [Cattiaux et al., 2010] and Central Asia [Davi et al., 2010]. The 2009–2010 winter drought in Europe was most pronounced in Scandinavia, where a rain-to-snow-season drought occurred because reservoirs that had low levels as a result of a summer drought were not replenished before the start of winter. This caused problems with drinking water supply and electricity production at the end of winter. In the same winter, Mongolia suffered severe problems with feeding cattle and one third of the population was affected by famine [Humanitarian Appeal, 2010; UNICEF, 2010]. These problems started when the soil moisture and groundwater storage that were low due to summer drought were not replenished before the start of winter. Similar examples can be found for droughts in warm seasonal climates where problems occur with irrigation, energy generation, and drinking water supply. For example, a precipitation deficit during the monsoon seasons of 2002 and 2009 in India caused postmonsoon groundwater levels to drop dramatically and reservoirs that dried up during the monsoon season did not refill until next years’ wet season [Bhuiyan et al., 2006, 2009]. In 2002 this wet-to-dry-season drought caused large-scale ecological damage, mass migration, and death [UNDP, 2002]. VAN LOON ET AL.


4651


10.1002/2013JD020383

There is a difference in persistence between seasonal droughts in warm seasonal climates and in cold seasonal climates. In cold climates (D/E climate) winter precipitation can be very high, but no infiltration to soil moisture or recharge to the groundwater occurs because all water is stored as snow. In warm climates (A/B/C climate) in the dry season, however, the situation is not binary but more gradual. This means that if there is some precipitation in the dry season, this will be lost to soil moisture storage and evapotranspiration, but if there is a high precipitation peak, some recharge to the groundwater system can occur. This makes the variability of hydrological drought characteristics higher for warm seasonal climates than for cold climates, and the predictability of the persistence of the hydrological drought lower. 4.2. Virtual Catchments In this study we have investigated only the effect of variability in climate on drought characteristics, and we fixed vegetation, soil, and groundwater response in the virtual catchments. Variability in catchment characteristics (e.g., geology and land cover), however, also contributes to a nonlinear response of hydrological drought to the meteorological situation and is just as important for hydrological drought development as variability in climate [Van Lanen et al., 2013; Van Loon, 2013]. Therefore, we tested whether our conclusions would hold under changing catchment properties by changing the groundwater response characteristic. In a slower-responding system droughts clearly are attenuated so that the deficit increases less with duration, but the effects of seasonality in climate leading to a bimodal shape in drought characteristics are still present. In this study climate and geology were assumed to be independent on short timescales. Surely, vegetation and, to a lesser extent, soils are related to climate, but geology is expected to be influenced by climate on much longer timescales. On geological timescales climate has changed significantly, influencing geology mainly through variations in sea level and therefore in deposition environment. This resulted in aquifer properties that are not typical for the current climate. As we only considered short timescales in this study, we neglected this dependency between climate and geology. We did not consider the dependency between vegetation and soil characteristics and climate either, because Van Lanen et al. [2013] showed that aquifer characteristics had much more significant effect on drought duration and deficit than vegetation and soil characteristics. 4.3. Model Structure and Parametrization The modeling approach used in this study is conceptual, but it includes the processes that are most important for drought propagation, i.e., snow accumulation, evapotranspiration, soil moisture depletion, groundwater recharge, and discharge, and the model concepts are similar to those in other conceptual hydrological models, which have proven to represent hydrological processes well [Wagener et al., 2004]. One important omission in our model approach is surface runoff. As all rainfall and snow melt are processed through the soil model and groundwater model, high precipitation peaks or snow melt peaks are dampened. The implication of this omission for hydrological drought development and recovery is that in reality the discharge time series show more short-lived small peaks, and therefore, real hydrological droughts are more often interrupted, and minor droughts develop. In most drought analysis pooling of the discharge droughts is needed to get rid of these minor droughts or small interruptions of a series of dependent droughts [Fleig et al., 2006]. The setup of our model makes pooling unnecessary. There is no upper limit to groundwater storage in the model, which potentially can lead to problems in the simulation. However, an upper limit is only needed in lowland areas with shallow water tables where excess recharge to an aquifer is drained by the surface water system. The omission of routing from the model is regarded as not so important in the study of drought because the traveltime of water in the river system mainly acts as another delay, which in our model is included in the groundwater model. Tijdeman et al. [2012] compared simulations by our model approach with observations and with model results of a calibrated HBV model for four contrasting catchments in Europe. The conclusion of this comparison was that the drought characteristics of our model approach are comparable to those of the HBV model, when low values for the response parameter (j value) are chosen because the lack of surface runoff is compensated by low response parameters. VAN LOON ET AL.


4652


10.1002/2013JD020383

5. Conclusions and Implications 5.1. Conclusions This research shows that in climates with strong seasonality in precipitation and/or temperature, 1. Precipitation drought characteristics are comparable to those in climates without seasonality. 2. Soil moisture droughts in the wet season can develop freely and have a large deficit, whereas soil moisture droughts in the dry season are limited by the wilting point and have a small deficit (Figure 5). This results in a divergent shape in the density field of soil moisture drought characteristics (Figure 6, top row). 3. Droughts in discharge are prolonged in cold seasonal climates by snow accumulation in winter and in dry seasonal climates by low precipitation and high evapotranspiration in the dry season (Figure 5). This results in a larger increase of deficit with duration and therefore in a bimodal shape in the density field of hydrological drought characteristics (Figure 6, middle row). 5.2. Implications With this study, we aimed to increase process understanding of hydrological drought development on the global scale, which has some important implications for water management in seasonal climates. It is, for example, known that the drought characteristics duration and deficit are related (as reported by, e.g., Dracup et al. [1980], Woo and Tariiule [1994], and Hisdal et al. [2004] and shown in Figure 3), but this research has shown that in seasonal climates the relation is not linear, and sensitivity of drought deficit changes with duration. For operational drought monitoring and forecasting this information is very valuable. The forecasting of the deficit of soil moisture droughts should consider the season. A certain precipitation drought in the wet season results in much higher deficits than that of the same precipitation drought during the dry season. Water managers can also use the information on the development of soil moisture droughts to see when wilting point is reached. Low deficits indicate that the soil moisture is bounded by the wilting point. For hydrological droughts it is important to monitor the drought condition at the onset of the dry season in warm seasonal climates and at the beginning of winter in cold seasonal climates, because that is the most reliable predictor of the drought situation during and at the end of the dry season (winter). For the prediction of hydrological droughts in seasonal climates, not only precipitation forecasts are needed but also information about the seasonal cycle of temperature and precipitation is essential. Including this knowledge in hydrological drought forecasting could increase forecasting skill considerably, as it makes the forecast less dependent on the forecast skill of actual precipitation and temperature. A reliable forecast of seasonal droughts is important as summer droughts (wet season droughts) that continue into winter (the dry season) have higher increase of deficit with duration. Based on the results presented in this paper, we provide an argument for taking into account terrestrial processes on hydrological drought development in water management. We found that seasonality effects on soil moisture and discharge drought characteristics cannot be explained by meteorological processes alone. Frequently, however, meteorology-based drought indices (e.g., Standardized Precipitation Index, SPI, or Palmer Drought Severity Index, PDSI) are used as proxy for hydrological drought [e.g., Nalbantis and Tsakiris, 2009; Zhai et al., 2010; Dai, 2011]. The effects of seasonality in climate, for example, by high evapotranspiration or snow accumulation on hydrological drought development are not accounted for in these indices. Due to the nonlinear response of soil moisture, groundwater, and streamflow to the meteorological situation in climates with strong seasonality, hydrological drought characteristics cannot be derived straightforwardly from meteorological drought characteristics. If we compare our conclusions with results of the catchment-scale studies mentioned in section 1, similar drought propagation processes were found, but this is the first time that we could systematically investigate the effects of increasing climate seasonality on drought duration and deficit at global scale, including ungauged parts of the world.

Appendix A: All Subclimate Types In this appendix we extended Figure 4 to include all subclimate types (Figure A1). VAN LOON ET AL.


4653

10.1002/2013JD020383

25 1 5 0 100 25 1 5 0

1 5

25

100

0

1 5

25

100

0


100

0

1 5

25

100

0

1 5

25

100


1

30 120 360 900 1

30 120 360 900 1

30 120 360 900 1

30 120 360 900 1

30 120 360 900

Duration (days) Figure A1. Contour plot of the 90% probability density fields of drought duration and deficit in P, SM, and Q for all subclimates, including linear regression lines and their slope (𝛼 ).

Acknowledgments The authors thank the financial contribution of the EU projects WATCH (contract 036946) and DROUGHT-R and SPI (contract 282769), the NWO grant (contract NWO GO-AO/30), and the support of the UNESCO-IHP VII FRIEND programme and research school WIMEK-SENSE. Furthermore, we thank three anonymous reviewers for their helpful comments.

VAN LOON ET AL.

References Bhuiyan, C., R. Singh, and F. Kogan (2006), Monitoring drought dynamics in the Aravalli region (India) using different indices based on ground and remote sensing data, Int. J. Appl. Earth Obs. Geoinf., 8(4), 289 – 302, doi:10.1016/j.jag.2006.03.002. Bhuiyan, C., W. Flügel, and R. Singh (2009), Erratic monsoon, growing water demand, and declining water table, J. Spatial Hydrol., 9(1), 1–20. Birtles, A. B., and W. B. Wilkinson (1975), Mathematical simulation of groundwater abstraction from confined aquifers for river regulation, Water Resour. Res., 11, 571–580. Brohan, P., J. Kennedy, I. Harris, S. Tett, and P. Jones (2006), Uncertainty estimates in regional and global observed temperature changes: A new data set from 1850, J. Geophys. Res., 111, D12106, doi:10.1029/2005JD006548. Burn, D. H., J. M. Buttle, D. Caissie, G. MacCulloch, C. Spence, and K. Stahl (2008), The processes, patterns and impacts of low flows across Canada, Can. Water Resour. J., 33(2), 107–124, doi:10.4296/cwrj3302107. Cattiaux, J., R. Vautard, C. Cassou, P. Yiou, V. Masson-Delmotte, and F. Codron (2010), Winter 2010 in Europe: A cold extreme in a warming climate, Geophys. Res. Lett., 37, L20704, doi:10.1029/2010GL044613.


4654


10.1002/2013JD020383

Dai, A. (2011), Drought under global warming: A review, Wiley Interdiscip. Rev. Clim. Change, 2(1), 45–65, doi:10.1002/wcc.81. Dai, A. (2013), Increasing drought under global warming in observations and models, Nature Clim. Change, 3(1), 52–58, doi:10.1038/nclimate1633. Davi, N., G. Jacoby, K. Fang, J. Li, R. D’Arrigo, N. Baatarbileg, and D. Robinson (2010), Reconstructing drought variability for Mongolia based on a large-scale tree ring network: 1520–1993, J. Geophys. Res., 115, D22103, doi:10.1029/2010JD013907. Dracup, J. A., L. Kil Seong, and E. G. Paulson Jr. (1980), On the statistical characteristics of drought events, Water Resour. Res., 16(2), 289–296. Eltahir, E. A. B., and P. J.-F. Yeh (1999), On the asymmetric response of aquifer water level to floods and droughts in Illinois, Water Resour. Res., 35(4), 1199–1217, doi:10.1029/1998WR900071. Fleig, A. K., L. M. Tallaksen, H. Hisdal, and S. Demuth (2006), A global evaluation of streamflow drought characteristics, Hydrol. Earth Syst. Sci., 10(4), 535–552, doi:10.5194/hess-10-535-2006. Gaál, L., J. Szolgay, S. Kohnová, J. Parajka, R. Merz, A. Viglione, and G. Blöschl (2012), Flood timescales: Understanding the interplay of climate and catchment processes through comparative hydrology, Water Resour. Res., 48(4), W04511, doi:10.1029/2011WR011509. Geiger, R. (1961), Uberarbeitete Neuausgabe von Geiger, R.: Koppen-Geiger / Klima der Erde., (Wandkarte 1:16 Mill), Klett-Perthes, Gotha. Hagemann, S., C. Chen, J. O. Haerter, J. Heinke, D. Gerten, and C. Piani (2011), Impact of a statistical bias correction on the projected hydrological changes obtained from three GCMS and two hydrology models, J. Hydrometeor, 12(4), 556–578, doi:10.1175/2011JHM1336.1. Hannaford, J., B. Lloyd-Hughes, C. Keef, S. Parry, and C. Prudhomme (2011), Examining the large-scale spatial coherence of European drought using regional indicators of precipitation and streamflow deficit, Hydrol. Processes, 25(7), 1146–1162, doi:10.1002/hyp.7725. Harding, R., et al. (2011), Preface to the “water and global change (WATCH) special collection: Current knowledge of the terrestrial global water cycle”, J. Hydrometeorology, 12(6), 1149–1156, doi:10.1175/JHM-D-11-024.1. Hisdal, H., L. M. Tallaksen, B. Clausen, E. Peters, and A. Gustard (2004), Hydrological Drought Characteristics, Developments in Water Science, vol. 48, chap. 5, pp. 139–198, Elsevier Science B.V., Amsterdam, The Netherlands. Humanitarian Appeal (2010), http://ochaonline.un.org/cap2006/webpage.asp?MenuID=15693™Page=1856(lastaccess:28december2012). Kim, T., J. Valdés, and C. Yoo (2003), Nonparametric approach for estimating return periods of droughts in arid regions, J. Hydrol. Eng., 8(5), 237–246, doi:10.1061/(ASCE)1084-0699(2003)8:5(237). Kraijenhoff van de Leur, D. A. (1958), A study of non-steady groundwater flow with special reference to a reservoir-coefficient, De Ingénieur, 70, 87–94. Kraijenhoff van de Leur, D. A (1962), Some effects of the unsaturated zone on nonsteady free-surface groundwater flow as studied in a sealed granular model, J. Geophys. Res., 67, 4347–4362. Laaha, G., and G. Blöschl (2006), Seasonality indices for regionalizing low flows, Hydrol. Processes, 20(18), 3851–3878. Li, L., C. Ngongondo, C. Xu, and L. Gong (2013), Comparison of the global TRMM and WFD precipitation datasets in driving a large-scale hydrological model in Southern Africa, Hydrol Res., 44.5, 770, doi:10.2166/nh.2012.175. Mishra, A. K., and V. P. Singh (2010), A review of drought concepts, J. Hydrol., 391(1-2), 202–216, doi:10.1016/j.jhydrol.2010.07.012. Nalbantis, I., and G. Tsakiris (2009), Assessment of hydrological drought revisited, Water Resour. Manage., 23, 881–897, doi:10.1007/s11269-008-9305-1. Peel, M. C., B. L. Finlayson, and T. A. Mcmahon (2007), Updated world map of the Koppen-Geiger climate classification, Hydrol. Earth Syst. Sci., 11, 1633–1644. Peters, E. (2003), Propagation of drought through groundwater systems: Illustrated in the Pang (UK) and Upper-Guadiana (ES) catchments, PhD thesis. Peters, E., G. Bier, H. A. J. van Lanen, and P. J. J. F. Torfs (2006), Propagation and spatial distribution of drought in a groundwater catchment, J. Hydrol., 321(1/4), 257–275, doi:10.1016/j.jhydrol.2005.08.004. Piani, C., G. Weedon, M. Best, S. Gomes, P. Viterbo, S. Hagemann, and J. Haerter (2010), Statistical bias correction of global simulated daily precipitation and temperature for the application of hydrological models, J. Hydrol., 395(3–4), 199 – 215, doi:10.1016/j.jhydrol.2010.10.024. Ritzema, H. P. (1994), Subsurface flow to drains, in Drainage Principles and Applications, 2nd ed., edited by H. P. Ritzema, pp. 263–304, International Institute for Land Reclamation and Improvement: Wageningen, The Netherlands. Schneider, U., T. Fuchs, A. Meyer-Christoffer, and B. Rudolf (2008), Global Precipitation Analysis Products of the GPCC, Deutscher Wetterdienst, Offenbach a. M., Germany. Seibert, J. (2000), Multi-criteria calibration of a conceptual runoff model using a genetic algorithm, Hydrol. Earth Syst. Sci., 4(2), 215–224. Seibert, J., and M. J. P. Vis (2012), Teaching hydrological modeling with a user-friendly catchment-runoff-model software package, Hydrol. Earth Syst. Sci., 16(9), 3315–3325, doi:10.5194/hess-16-3315-2012. Seneviratne, S. I., et al. (2012), Changes in climate extremes and their impacts on the natural physical environment, in Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation, A Special Report of Working Groups I and II of the Intergovernmental Panel on Climate Change (IPCC), edited by C. B. Field et al., pp. 109–230, Cambridge Univ. Press, Cambridge, U. K., and New York, NY, USA. Sheffield, J., and E. Wood (2011), Drought: Past Problems and Future Scenarios, 192 pp., Earthscan, London, Washington, D. C. Sheffield, J., and E. F. Wood (2007), Characteristics of global and regional drought, 1950–2000: Analysis of soil moisture data from off-line simulation of the terrestrial hydrologic cycle, J. Geophys. Res., 112, D17115, doi:10.1029/2006JD008288. Sheffield, J., E. F. Wood, and M. L. Roderick (2012), Little change in global drought over the past 60 years, Nature, 491(7424), 435–438, doi:10.1038/nature11575. Stewart, R., J. Pomeroy, and R. Lawford (2011), The drought research initiative: A comprehensive examination of drought over the Canadian Prairies, Atmos. Ocean, 49(4), 298–302, doi:10.1080/070559002011.622574. Tallaksen, L. M., and H. A. J. V. Lanen (eds.) (2004), Hydrological Drought: Processes and Estimation Methods for Streamflow and Groundwater, Developments in Water Science; 48, The Netherlands, Elsevier Science B. V., Amsterdam, The Netherlands. Tallaksen, L. M., H. Hisdal, and H. A. J. van Lanen (2009), Space-time modelling of catchment scale drought characteristics, J. Hydrol., 375(3–4), 363–372, doi:10.1016/j.jhydrol.2009.06.032. Teuling, A. J., A. F. Van Loon, S. I. Seneviratne, I. Lehner, M. Aubinet, B. Heinesch, C. Bernhofer, T. Grünwald, H. Prasse, and U. Spank (2013), Evapotranspiration amplifies European summer drought, Geophys. Res. Lett., 40, 2071–2075, doi:10.1002/grl.50495. Tijdeman, E., A. F. Van Loon, N. Wanders, and H. A. J. Van Lanen, (2012), The effect of climate on droughts and their propagation in different parts of the hydrological cycle, DROUGHT-R&SPI Tech. Rep. 2, [Available at http://www.eu-drought.org/, last access: 9 November 2012], Wageningen University, The Netherlands.

VAN LOON ET AL.


4655


10.1002/2013JD020383

UNDP, (2002), Drought damage assessment and agricultural rehabilitation for drought-affected districts of Rajasthan, Tech. Rep., UNDP Mission Report. UNICEF (2010), http://www.unicef.org/eapro/media_12706.html(last access: 28 december 2012). Uppala, S. M., et al. (2005), The ERA-40 re-analysis, Q. J. R. Meteorolog. Soc., 131(612), 2961–3012, doi:10.1256/qj.04.176. Van Lanen, H. A. J., N. Wanders, L. M. Tallaksen, and A. F. Van Loon (2013), Hydrological drought across the world: Impact of climate and physical catchment structure, Hydrol. Earth Syst. Sci., 17, 1715–1732, doi:10.5194/hess-17-1715-2013. Van Loon, A. F. (2013), On the propagation of drought. How climate and catchment characteristics influence hydrological drought development and recovery, PhD thesis, Wageningen University [http://edepot.wur.nl/249786, last access: 28 May 2013]. Van Loon, A. F., and H. A. J. Van Lanen (2012), A process-based typology of hydrological drought, Hydrol. Earth Syst. Sci., 16(7), 1915–1946, doi:10.5194/hess-16-1915-2012. Vidal, J.-P., E. Martin, L. Franchistéguy, F. Habets, J.-M. Soubeyroux, M. Blanchard, and M. Baillon (2010), Multilevel and multiscale drought reanalysis over France with the Safran-Isba-Modcou hydrometeorological suite, Hydrol. Earth Syst. Sci., 14(3), 459–478, doi:10.5194/hess-14-459-2010. Viste, E., D. Korecha, and A. Sorteberg (2012), Recent drought and precipitation tendencies in Ethiopia, Theor. Appl. Climatol., 1–17, doi:10.1007/s00704-012-0746-3, (to appear in print). Wagener, T., H. S. Wheater, and H. V. Gupta (2004), Rainfall-Runoff Modelling in Gauged and Ungauged Catchments, Imperial College Press, London, U. K. Wand, M., and M. Jones (1995), Kernel Smoothing, Chapman and Hall, London. Wanders, N., H. A. J. Van Lanen, and A. F. Van Loon, (2010), Indicators for drought characterization on a global scale, WATCH Tech. Rep. 24, [Available at: http://www.eu-watch.org/publications/technical-reports] (last access: 5 July 2012), Wageningen University, The Netherlands. WATCH (2012), the European Union funded 6th Framework Project “WATer and global CHange” [http://eu-watch.org and http://www. waterandclimatechange.eu/about/watch-forcing-data-20th-century, last access: 1 March 2013]. Weedon, G. P., S. Gomes, P. Viterbo, W. J. Shuttleworth, E. Blyth, H. Österle, J. C. Adam, N. Bellouin, O. Boucher, and M. Best (2011), Creation of the WATCH Forcing Data and its use to assess global and regional reference crop evaporation over land during the twentieth century, J. Hydrometeorology, 12, 823–848, doi:10.1175/2011JHM1369.1. Woo, M.-K., and A. Tariiule (1994), Streamflow droughts of northern Nigerian rivers, Hydrol. Sci. J., 39(1), 19–34. Wösten, J. H. M., G. J. Veerman, W. J. M. De Groot, and J. Stolte, (2001a), Waterretentie – en doorlatendheidskarakteristieken van boven – en ondergronden in Nederland: De Staringreeks, Technisch Rapport 153, Alterra, Wageningen, available at: www2.alterra.wur.nl/Webdocs/PDFFiles/Alterrarapporten/AlterraRapport153.pdf (last access: 22 October 2013). Wösten, J. H. M., Y. A. Pachepsky, and W. J. Rawls (2001b), Pedotranfer functions: Bridging the gap between available basic soil data and missing soil hydraulic characteristics, J. Hydrol., 251, 123–150. Zhai, J., B. Su, V. Krysanova, T. Vetter, C. Gao, and T. Jiang (2010), Spatial variation and trends in PDSI and SPI indices and their relation to streamflow in 10 large regions of China, J. Climate, 23(3), 649–663, doi:10.1175/2009JCLI2968.1.

VAN LOON ET AL.


4656