commodities and biodiversity - UNEP-WCMC

0 downloads 0 Views 14MB Size Report
for: a) the Watersheds of the Andes, b) the Mekong Basin and c) the Great ... Trends in crop production under GEO-4 Markets First scenario for Burundi, DRC, Ethiopia, .... Protection is modelled for the following: i) PAs off (no ...... Tree cover, broadleaved, deciduous, closed ...... FOGINT cloud forest model developed at King's.
COMMODITIES AND BIODIVERSITY SPATIAL ANALYSIS OF POTENTIAL FUTURE THREATS TO BIODIVERSITY AND ECOSYSTEM SERVICES

Authors Arnout van Soesbergen and Andy Arnell

Prepared for John D. and Catherine T. MacArthur Foundation

Acknowledgements

2

Within UNEP-WCMC, support in preparing this report was provided by Yara Shennan-Farpon, Val Kapos, Rebecca Mant, Marieke Sassen, Cecilia Larrosa and Elizabeth Farmer. Neil Burgess served as an internal reviewer and Jörn Scharlemann (University of Sussex) served as an external reviewer and provided comments on the draft report.

Published March 2015 Copyright 2015 United Nations Environment Programme The United Nations Environment Programme World Conservation Monitoring Centre (UNEP-WCMC) is the specialist biodiversity assessment centre of the United Nations Environment Programme (UNEP), the world’s foremost intergovernmental environmental organization. The Centre has been in operation for over 30 years, combining scientific research with practical policy advice. This publication may be reproduced for educational or non-profit purposes without special permission, provided acknowledgement to the source is made. Reuse of any figures is subject to permission from the original rights holders. No use of this publication may be made for resale or any other commercial purpose without permission in writing from UNEP. Applications for permission, with a statement of purpose and extent of reproduction, should be sent to the Director, UNEP-WCMC, 219 Huntingdon Road, Cambridge, CB3 0DL, UK. The contents of this report do not necessarily reflect the views or policies of UNEP, contributory organizations or editors. The designations employed and the presentations of material in this report do not imply the expression of any opinion whatsoever on the part of UNEP or contributory organizations, editors or publishers concerning the legal status of any country, territory, city area or its authorities, or concerning the delimitation of its frontiers or boundaries or the designation of its name, frontiers or boundaries. The mention of a commercial entity or product in this publication does not imply endorsement by UNEP. Image credits: Willem Tims/shutterstock.com; Marieke Sassen; wong sze yuen/shutterstock.com; AdeleD/shutterstock.com; iofoto/ shutterstock.com; John Wollwerth/shutterstock.com; Dr. Morley Read/shutterstock.com; Christian Vinces/shutterstock.com; vitmark/shutterstock.com; Steffen Foerster/shutterstock.com; Khoroshunova Olga/shutterstock.com; Wasu Watcharadachaphong/shutterstock.com; Vietnam Photography/shutterstock.com; Fotokostic/shutterstock.com; David Vogt/shutterstock.com; Pichugin Dmitry/shutterstock.com; PRILL/shutterstock.com; Alberto Loyo/shutterstock.com; John Bill/shutterstock.com; DuongMinhTien/shutterstock.com; PathomP/shutterstock.com; Trinh Le Nguyen/shutterstock.com; utcon/shutterstock.com; Hai Nv/shutterstock.com; Paul S. Wolf/shutterstock.com; Christian Vinces v/shutterstock.com; Xiong Wei/shutterstock.com; africa924/shutterstock.com; africa924/shutterstock.com; Albie Venter/shutterstock.com; think4photop/shutterstock.com; David Persson/shutterstock.com; Againstar/shutterstock.com; Ostill/shutterstock.com; Dr. Morley Read/shutterstock.com;

UNEP World Conservation Monitoring Centre (UNEP-WCMC) 219 Huntingdon Road, Cambridge CB3 0DL, UK Tel: +44 1223 277314 www.unep-wcmc.org

UNEP promotes environmentally sound practices globally and in its own activities. Our distribution policy aims to reduce UNEP’s carbon footprint

Contents List of Figures

4

List of Tables

7

List of Acronyms

8

1 Executive Summary

9

2 Introduction

13

3 Spatial analysis framework 3.1 Conceptual analysis framework 3.2 Study areas and scale of analysis 3.3 Key criteria 3.4 Global and regional scale analyses 3.5 Focal questions 3.6 Scenarios 3.7 Quantification of scenarios – the IMPACT model 3.8 Landscape changes - the LandSHIFT model 3.9 Biodiversity metric 3.10 Ecosystem function metric 3.11 Pressure and threat

15 15 16 17 17 18 18 24 25 30 34 37

4 Results and discussion 3 4.1 General results 4.2 Great Lakes of Africa region 4.3 The greater Mekong region 4.4 Watersheds of the Andes region

9 39 41 57 73

5 Methods discussion

89

6 Conclusions 6.1 Regional conclusions 6.2 General conclusions

99 99 101

7 References

103

8 Appendices 106 Appendix I. Regional scenario narrative summaries. 106 Appendix II. Commodities modelled in the IMPACT model. 111 Appendix III. Binary links between landscape characteristics and ecosystem function provision. 112 Appendix IV. Cross-walk between IUCN Habitats and LandSHIFT Land-Use Types. 114 Appendix V. The Watershed Exploration Tool: a web-portal for exploring results in the three MacArthur regions. 132 Appendix VI. Watershed boundaries with protected areas and Key Biodiversity Areas (KBAs), 134 for: a) the Watersheds of the Andes, b) the Mekong Basin and c) the Great Lakes of Africa.

3

LIST OF FIGURES

4

No. Title

Page

1

Conceptual analysis framework. Different socio-economic scenarios are characterised through changes in the landscape. Blue boxes represent the analysis of impacts for current and future status, allowing for assessment of changes between these two states.

15

2

MacArthur Foundation focal regions and watersheds within each region. The MacArthur sub 16 regions are three countries in each region for which higher resolution analysis was carried out (see section 2.1.4).

3

Modelling components for two scales of analysis.

17

4

Workflow for mapping biodiversity in watersheds. Data on species range (blue) was combined with modelled land-cover data and watershed polygons. The areas of species X’s range (1) and suitable habitat (green) overlapping with watershed A were extracted. Statistics for all suitable habitat (3) and total range of species X in the region (4) were obtained using similar approaches.

31

5

Workflow for mapping ecosystem functions in watersheds. Modelled data on landcover was combined with ancillary data and watershed polygons. The areas of different land-cover types and ancillary data (e.g. slopes) within a watershed was extracted and processed using a binary link relationship table.

35

6

Screenshot of web-based tool available at: http://macarthur.unep-wcmc.org.

40

7

Trends in crop production under GEO-4 Markets First scenario for Burundi, DRC, Ethiopia, Kenya, Malawi, Mozambique, Rwanda, Tanzania, Uganda and Zambia.

41

8

Recent trends (black based on FAO) and future (colours, based on IMPACT model) projections of maize and beef production under four regional scenarios for the GLR region S1: Industrious Ants, S2: Herd of Zebra, S3: Lone Leopards and S4: Sleeping Lions.

43

9

Difference between domestic production and demand for meat products in Uganda, Rwanda and Burundi between 2005 and 2050 for the Industrious Ants scenario.

44

10

Trends in production and yield for three important crops in the GLR region for the Industrious Ants (S1) scenario with no climate change and with climate change based on the IPSL GCM.

46

11

Land use: Maps of baseline (2005) and potential future (2050) land use in the Great Lakes of Africa MacArthur Region, based on GEO-4 scenarios of change.

47

12

Land-use change: potential change in area (km2) for major land-use categories between 2005 48 and 2050 within the Great Lakes of Africa MacArthur Region, based on GEO-4 scenarios of change.

13

Land use: Maps of baseline (2005) and potential future (2050) land use in Uganda, Rwanda and Burundi, based on regionally-developed scenarios of change.

14

Land-use change: potential change in area (km2) for major land-use categories between 2005 50 and 2050 within Uganda, Rwanda and Burundi, based on regionally-developed scenarios of change.

15

Baseline biodiversity importance and projected changes in biodiversity importance between 2005 and 2050 for watersheds in the Great Lakes of Africa MacArthur Region, based on global (GEO-4) scenarios of change.

51

16

Baseline biodiversity importance and projected changes in biodiversity importance between 2005 and 2050 for watersheds in the Uganda, Rwanda and Burundi, based on regionallydeveloped scenarios of change.

52

17

Ecosystem Function Provision: Baseline provision and projected changes between 2005 and 53 2050 for watersheds in the Great Lakes of Africa MacArthur Region, based on global (GEO-4) scenarios of change.

18

Ecosystem Function Provision: Baseline provision and projected changes between 2005 and 2050 for watersheds in Uganda, Rwanda and Burundi based on regionally-developed scenarios of change.

49

54

No. Title

Page

19

Current and future pressure index for watersheds the Great Lakes of Africa region.

55

20

Current and future pressure index for watersheds in Uganda, Rwanda and Burundi.

55

21

Projected agricultural expansion and baseline biodiversity (0-1) for the Great Lakes of Africa region under the GEO-4 scenarios.

56

22

Trends in crop production under GEO-4 markets first scenario for China, Myanmar, Thailand, Lao PDR/Cambodia and Viet Nam

57

23

Recent (black, based on FAO) and future (colours, based on IMPACT model) projected trends 60 in rice and beef production under the four regional scenarios for the Mekong; S1: Land of the Golden Mekong, S2: Buffalo, Buffalo, S3: DoReki Dragon and S4: Tigers on a Train.

24

Difference between domestic production and demand for meat products in Cambodia, Lao PDR and Viet Nam between 2005 and 2050 for the Land of the Golden Mekong (S1) scenario.

60

25

Trends in production and yield for three important crops in the Mekong region for the land of the Golden Mekong scenario with no climate change and with climate change based on the IPSL GCM.

62

26

Maps of baseline (2005) and potential future (2050) land use in the Greater Mekong MacArthur Region, based on GEO-4 scenarios of change.

63

27

Land-use change: potential change in area (km2) for major land-use categories between 2005 and 2050 within the Greater Mekong MacArthur Region, based on GEO-4 scenarios of change.

64

28

Maps of baseline (2005) and potential future (2050) land use in Viet Nam, Cambodia and Lao PDR, based on regionally-developed scenarios of change.

65

29

Land-use change: potential change in area (km2) for major land-use categories between 2005 and 2050 within Viet Nam, Cambodia and Lao PDR, based on regionally-developed scenarios of change.

66

30

Baseline biodiversity importance and projected changes in biodiversity importance between 2005 and 2050 for watersheds in the Greater Mekong MacArthur Region, based on global (GEO-4) scenarios of change.

67

31

Baseline biodiversity importance and projected changes in biodiversity importance between 2005 and 2050 for watersheds in Viet Nam, Cambodia and Lao PDR, based on regionallydeveloped scenarios of change.

68

32

Ecosystem Function Provision: Baseline provision and projected changes between 2005 and 2050 for watersheds in the Greater Mekong MacArthur Region, based on global (GEO-4) scenarios of change.

69

33

Baseline provision and projected changes between 2005 and 2050 for watersheds in Viet Nam, Cambodia and Lao PDR based on regionally-developed scenarios of change.

70

34

Current and future pressure index for watersheds in the Greater Mekong region.

71

35

Current and future pressure index for watersheds in Lao PDR, Cambodia and Viet Nam

71

36

Projected agricultural expansion and baseline biodiversity (0-1) for the Greater Mekong region under the GEO-4 scenarios.

72

37

Trends in crop production under GEO-4 markets first scenario for Brazil, Colombia, Peru, Ecuador, Bolivia and Venezuela.

73

38

Recent (black, based on FAO) and future (colours, based on IMPACT model) trends in maize and beef production under the four regional scenarios for the Andes region.

75

39

Difference between domestic production and demand for meat products in Colombia, Ecuador and Peru between 2005 and 2050 for the New Dawn (S1) scenario.

76

5

6

No. Title

Page

40

Trends in production and yield for three important crops in the Andes region for the New Dawn scenario with no climate change and with climate change based on the IPSL GCM.

77

41

Maps of baseline (2005) and potential future (2050) land use in the Watersheds of the Andes MacArthur Region, based on GEO-4 scenarios of change.

78

42

Land-use change: potential change in area (km2) for major land-use categories between 2005 79 and 2050 within the Watersheds of the Andes MacArthur Region, based on GEO-4 scenarios of change.

43

Maps of baseline (2005) and potential future (2050) land use in Colombia, Ecuador and Peru, based on regionally-developed scenarios of change.

44

Land-use change: potential change in area (km2) for major land-use categories between 2005 81 and 2050 within Colombia, Ecuador and Peru for regionally-developed scenarios of change.

45

Baseline biodiversity importance and projected changes in biodiversity importance between 2005 and 2050 for watersheds in the Watersheds of the Andes MacArthur Region, based on global (GEO-4) scenarios of change.

82

46

Baseline biodiversity importance and projected changes in biodiversity importance between 2005 and 2050 for watersheds in Colombia, Ecuador and Peru based on regionallydeveloped scenarios of change.

83

47

Ecosystem Function Provision: Baseline provision and projected changes between 2005 and 2050 for watersheds in the Watersheds of the Andes MacArthur Region, based on global (GEO-4) scenarios of change.

84

48

Ecosystem Function Provision: Baseline provision and projected changes between 2005 and 2050 for watersheds in Colombia, Ecuador and Peru based on regionally-developed scenarios of change.

85

49

Current and future pressure index for watersheds in the Andes region.

86

50

Current and future pressure index for watersheds in Colombia, Ecuador and Peru.

86

51

Projected agricultural expansion and baseline biodiversity (0-1) for the Andes region under the GEO-4 scenarios.

87

52

Three different assumptions of species range (squares) overlap with suitable habitat (green) in a watershed (blue circle).

89

53

Scatter plots showing biodiversity importance ranking for watersheds in the GLR region, with different distribution of habitat assumptions: minimum area, maximum area and equal distribution, when plotted against the actual extent of suitable habitat (ESH). The red lines show a one to one relationship. Habitat assumptions with points closest to this line have least effect on biodiversity importance rankings.

90

54

Baseline biodiversity importance for the GLR region aggregated by FAO watersheds (left), 0.5 94 degree grids (middle) and MacArthur watersheds (right).

55

Forest and grassland/shrubland for the baseline (2005) and 2050 under the Industrious Ants scenario without protection, with protected areas and with protected areas and biodiversity areas for part of the GLR region.

80

97

LIST OF TABLES No. Title

Page

1

Scenarios and states for key change factors for Mekong regional scales scenarios selected and, for each of these, the combination of different extreme states for the different change factors selected.

21

2

Scenarios and states for key change factors for Andes regional scale scenarios and the combinations of extreme states for the four key change factors.

22

3

Land-use/land-cover classes in GLC2000 and LandSHIFT.

26

4

Spatial micro-level data used in LandSHIFT for model initialisation.

28

5

Yield (% change) and production (1000 tonnes) changes for four crops under the four GEO-4 scenarios for the GLR region.

42

6

Yield, Area and Production changes (%) between 2005-2050 for seven crops for Burundi, Rwanda and Uganda.

45

7

Yield (% change) and production (1000 tonnes) changes for four crops under the four GEO-4 scenarios for the Mekong region.

58

8

Yield, area and production changes (%) between 2005–2050 for seven crops for Cambodia, Lao PDR and Viet Nam under the RCP 8.5 climate change scenario.

61

9

Yield (% change) and production (1000 tonnes) changes for four crops under the four GEO-4 scenarios for the Andes region.

74

10

Yield, area and production changes (%) between 2005-2050 for five crops for Colombia, Ecuador and Peru under the IPSL RCP 8.5 climate change scenario.

76

11

Confusion matrix comparing presence and absence in watersheds according to ESH based approach and observed presence of trigger species in Important Bird Areas (IBAs).

91

12

Number of species restricted by altitude in each MacArthur region

91

13

LandSHIFT sensitivity run results for technological and climate change for the Andes and Mekong sub regions

92

14

Change in extent for major land-use classes between 2005 and 2050 with varying levels of protection in the LandSHIFT model. Protection is modelled for the following: i) PAs off (no protection), ii) PAs on (protected areas on) or iii) PAs with KBAs (protected areas and Key Biodiversity Areas, KBAs, combined).

95

7

LIST OF ACRONYMS

8

AR4

IPCC Fourth Assessment Report

ASEAN

The Association of Southeast Asian Nations

CCAFS

Climate Change Agriculture and Food Security Research Program of the CGIAR

CEPF

Critical Ecosystem Partnership Fund

EEA

European Environment Agency

ESH

Extent of Suitable Habitat

EPIC

Economics and Policy Innovations for Climate-Smart Agriculture

FAO

Food and Agricultural Organisation of the United Nations

GCM

Global Circulation Model

GDP

Gross Domestic Product

GEO

Global Environment Outlook

GLR

Great Lakes region of Africa

KBA

Key Biodiversity Area

IBA

Important Bird and Biodiversity Area

IFPRI

International Food and Policy Research Institute

IMPACT

International Model for Policy Analysis of Agricultural Commodities and Trade

IPCC

International Panel on Climate Change

IPSL

Institut Pierre Simon Laplace - Climate Model

IWSM

IMPACT Model with Water Simulation Module

KBA

Key Biodiversity Area

LCCS

Land Cover Classification System

OECD

The Organisation for Economic Co-operation and Development

PA

Protected Area

PDR

People’s Democratic Republic

RCP

Representative Concentration Pathway

TNC

The Nature Conservancy

UCI

Universidad para la Cooperación Internacional

UNEP

United Nations Environment Programme

UNEP-WCMC

United Nations Environment Programme World Conservation Monitoring Centre

WDPA

World Database on Protected Areas

WSSD

World Summit on Sustainable Development

1. Executive Summary Over the coming decades, society will have to balance competing needs for land to feed the growing human population, to provide resources and energy to satisfy the ever-accelerating human consumption, to slow global warming and to reduce the rate of loss of ecosystem services and biodiversity. For decision makers to balance these different demands on land, it is crucial that they have access to spatially explicit information and analyses on the effects of different trajectories of human-induced landscape change on biodiversity and ecosystem services.

9

A NOVEL ANALYTICAL FRAMEWORK

10

This report presents a novel analytical framework that is able to provide such information. This framework can be implemented at multiple geographic scales to evaluate priorities for conservation or other action. It considers spatially explicit drivers of land-use change, including changes in human population, commodity markets and agricultural production.

futures of potential landscape change due to likely changes in these drivers. Models were run with projections of climate change based on the Representative Concentration Pathway (RCP) 8.5 emission pathway. Furthermore, a strict conservation policy was assumed where no land within protected areas was converted to other land uses.

The framework has been applied in the context of UNEP-WCMC’s project on Commodities and Biodiversity, funded by the John D. and Catherine T. MacArthur Foundation and implemented in the three MacArthur focal regions which are the Great Lakes of East and Central Africa (henceforth, GLR region) the Greater Mekong and its Headwaters (henceforth, Mekong region) and the Watersheds of the Andes (henceforth, Andes region).

The potential impacts of projected land-use change and within-region variability of these impacts were assessed at the watershed scale for biodiversity and ecosystem function provision. Biodiversity was assessed using a novel index of biodiversity importance which is a metric based on the distribution of suitable habitat for a species in a watershed and in a region. The metric combines known species distribution data for birds, mammals and amphibians from the IUCN Red List spatial dataset of species’ Extent of Occurrence and an established method of linking species’ habitat preferences to land-use/ cover types. The potential of a watershed to provide ecosystem functions was assessed using expert and literature-driven binary links between specific land uses and other environmental properties and the ecosystem functions these properties can provide. Potential changes were analysed for commodity provisioning, wild provisioning and regulating ecosystem functions.

For each of these study regions, the framework was applied for two streams of analysis: ●A  nalysis

based on the Global Environment Outlook (GEO-4) scenarios of change up to 2050 with land-use change modelling at 5 arc minute resolution (~9km) and impacts on biodiversity and ecosystem function assessed for all watersheds within each of the three MacArthur focal regions.

●A 

regional-scale analysis based on regionallydeveloped scenarios for three countries in each MacArthur focal region with land-use change modelling at higher resolution (~1km) and impacts on biodiversity and ecosystem function assessed for all watersheds within these three countries.

All scenarios were quantified with the IMPACT agricultural economic model, producing national projections of agricultural demand and production, which were then applied within the LandSHIFT land-use model to project plausible

Results presented in this report aim to support decision makers in assessing and visualising likely future impacts on biodiversity and ecosystem functions and trade-offs, and in ultimately making more informed choices balancing conservation and development needs. A freely available web-based tool (http:// macarthur.unep-wcmc.org) was developed that can be used to explore all the results for the different scenarios and analyses within each study region.

COMMODITIES AND BIODIVERSITY IN THE MACARTHUR FOCAL REGIONS GLR region

Mekong region

Crop production in this region is projected to consistently increase across all scenarios in the period up to 2050 with production of some crops more than doubling for most countries. Scenarios have very different levels of production in this region, mostly related to large differences in crop yields between scenarios. Climate change under the RCP 8.5 scenario has a positive impact on yields for most crops in this region.

Modelled increases in crop production for this region are much lower than in the GLR. This can partly be explained by lower projected population increases in this region. For example, rice production in Viet Nam is projected to only marginally increase by 2050, and even decrease in China. Meat production, on the other hand, is likely to triple across the region with strong agreement between scenarios. This increase in production is projected to surpass the demand in Viet Nam by 2050 while Cambodia and Lao PDR will need to import meat products to satisfy domestic demand. Climate change impacts on crop yields are variable but typically point to a decrease in yields.

A key change, underpinning much of the projected land-use change in this region is the large increase in meat production. While crop increases are in part the result of changes in yields due to technological and climate changes, production of meat drives large increases in pasture areas. For the full MacArthur region under the global scenarios, these pasture increases lead to the loss of forest and natural grass/shrublands in equal measure. However, under the regional scenarios in Uganda, Rwanda and Burundi, relatively more forest is lost to satisfy the need for crop and grazing lands. Despite the sharp increase in production, domestic demand for all meat products is much higher than production by 2050, necessitating import of these products. The high demand for these products is driven by large projected population increases for all three countries. The resulting impacts of these losses in natural land cover are mainly visible in western Uganda and towards the north of Lake Victoria where current biodiversity and ecosystem function provision are highest and therefore projected to have highest losses. GEO-4 and regionallydeveloped scenarios both show similar results in terms of areas most impacted.

Overall, area of cropland is projected to decrease for all scenarios, while pasture areas are expanding throughout the region but primarily visible in eastern Thailand and surrounding Lake Tonle Sap in Cambodia. Apart from southern China for two of the GEO-4 scenarios, the expansion of pasture areas does not lead to largescale conversion of forest, as it concerns mostly a conversion of land already under agricultural use, or a conversion of natural grassland and shrubland. This agricultural expansion leads to consistent declines in biodiversity importance, particularly in the Mekong Delta in Viet Nam and eastern Thailand. North of Lake Tonle sap is also an area of particular concern, as this area is highly suitable for pasture leading to large projected declines in biodiversity if converted. Biodiversity and ecosystem services are already under pressure from many existing hydroelectric dams in the region, but more dams are under development or planned, particularly in Lao PDR and China. Some of these are planned on the main stem of the river Mekong with potentially devastating impacts on downstream biodiversity and food security.

11

12

Andes region

Finally

Production of most crops is projected to rise throughout the 2005-2050 period in this region with only small differences between scenarios and countries. Similar to the other two regions, meat production continues to rise, with production of beef tripled or even quadrupled by 2050. The production of beef products surpasses demand in Colombia, Peru and Ecuador throughout the regional scenarios by more than 25% of domestic demand in 2050. Climate change has variable impacts on crops in this region with some large projected yield decreases for crops such as maize and sugar cane, but yield increases for potatoes and some other crops. Crop areas for vegetables and plantains are projected to increase in Colombia, Peru and Ecuador under the regional scenarios while areas of rice and some oilseed crops decrease.

For each of the three regions, different scenarios lead to broadly similar changes in land use and impacts on biodiversity and ecosystem function provision. This is the result of similar trends in the underlying drivers of agricultural expansion (trade, population growth). In all three regions, more land is projected to be converted to pasture than to cropland. This is particularly the case in the Mekong and Andes region, where the scenarios point to at least a threefold production of meat products by 2050 due to increased (global) demand reflecting increased wealth and changes in food preferences.

The increased production of meat leads to largescale conversion of forest and other natural lands to pasture under the global scenarios, mostly in Peru and Colombia, while crop area expansion is limited. The regional scenarios for Colombia, Peru and Ecuador lead to only small areas of forest loss although there are large decreases in areas classified as mosaic land cover, which includes natural shrublands as well as tree cover. Large increases in agricultural areas are found along the pacific coast, mostly in Peru. These areas are already under considerable agricultural use but are heavily dependent on irrigation water coming from the Andes. Changes in precipitation patterns under climate change can therefore have serious consequences for the agricultural production in these areas and ultimately food security. Loss of biodiversity importance is most pronounced in watersheds with high baseline biodiversity, with largest decreases in eastern Peru and south Colombia with high agreement between all scenarios. Biodiversity and ecosystem services in the whole of the Andean MacArthur region are under considerable threat from mining, oil and gas developments and large dam projects. In most cases these operations are located in different areas than where agricultural expansion is projected to take place but they are likely to pose a more immediate threat to pristine habitats found in the Amazon basin.

Impacts of climate change on crop yields play a key role in determining the amount of land needed to be converted for agriculture. Climate change has different impacts in the three regions, with mostly increasing yields projected in the GLR and some decreasing yields in the Mekong and Andes regions. Climate change projections are extremely uncertain though and potential impacts on crop yields are hard to predict. Lower than projected yield increases would be particularly devastating for the GLR region where large amounts of forest are already projected to be lost to agricultural uses and even more land would be required if yields were to increase at a lower rate. For national or regional-level policy making that seeks to balance different demands on land or for an initial targeting of conservation investments, the watersheds approach provides valuable information. Yet, it is important that additional analysis is carried out to assess potential impacts at scales that are closer to those at which land-use planning decisions are made. Spatially explicit information and analyses on the effects of different trajectories of human-induced landscape change as well as climate change on biodiversity and ecosystem services are crucial to inform sustainable land-use planning. It is therefore urgent to increase the access to and the capacity to use such information and analyses by decision makers and other stakeholders so that they are better able to manage and plan for tradeoffs and synergies among different demands on land.

2. Introduction The global human population is projected to reach 9 billion by around 2050 and already one in six people go to bed hungry (Godfray et al. 2010). The increased need for food (a projected rise of 70% by 2050) will be exacerbated by increasing prosperity in some regions, which will be associated with increased demand for protein (Alexandratos 2009). Meanwhile, to reduce their carbon emissions, many countries are aiming to meet an increasing percentage of their energy needs from renewable sources, including bio-energy, leading to a projected 90% rise in demand for some bio-energy feed stocks by 2018 (OECD-FAO 2008). Together these rising demands represent an enormous need for increased agricultural production. In the past 40 years, similar increases in production have been achieved with only a 12% increase in global cropland area, largely through improved crop breeding and agricultural intensification (Foley et al. 2005). However, this intensification may not be sustainable over the long term, because of its impacts on the environment and natural resources. Around 30% of agricultural lands are now degraded and annual increases in cereal crop yields in the major ‘bread-basket’ regions are slowing (Foley et al. 2011).

13

14

Over the coming decades, society will have to balance competing needs for land to feed the growing human population, to provide resources and energy to satisfy the ever-accelerating human consumption, to slow global warming and to reduce the rate of loss of ecosystem services and biodiversity. For decision makers to balance these different demands on land, it is crucial that they have access to spatially explicit information and analyses on the effects of different trajectories of human-induced landscape change on biodiversity and ecosystem services. This report presents a novel framework that can be implemented at multiple spatial scales to evaluate priorities for conservation and other actions. It forms part of a project funded by the John D. and Catherine T. MacArthur Foundation which aims to: “provide analyses and baseline data to support decision-making in relation to the potential future impacts of major commodity developments on biodiversity in short-term and long-term planning in the MacArthur focal regions”.

The MacArthur focal regions are the Great Lakes of East and Central Africa region (henceforth GLR region), the Mekong basin and its headwaters in South East Asia (henceforth Mekong region), and the watersheds of the Andes in South America (henceforth Andes region). The analysis framework described in this report has been applied for these regions and includes a spatially-explicit consideration of the drivers of land-use change, including changes in human population, in commodity markets and in agricultural production. Likely changes in these drivers were captured in scenarios and used in a land-use change model to project plausible futures of landscape change. The drivers of change are based on global scenarios as well as regionally specific scenarios, which were developed in consultation with local stakeholders. The potential impacts of projected land-use change, and within-region variability of these impacts, were assessed for biodiversity, using a metric based on the distribution of suitable habitat for species in the region, and for the potential of a watershed to provide ecosystem functions. Results presented in this report aim to support decision makers in assessing and visualising likely future impacts on biodiversity and ecosystem functions and trade-offs, and in ultimately making more informed choices, balancing conservation and development needs.

3. Spatial analysis framework 3.1 CONCEPTUAL ANALYSIS FRAMEWORK The overarching question the spatial analysis framework tries to answer is: Where are current and future potential priorities for biodiversity and ecosystem services in relation to the impact of major commodity markets within the Great Lakes Region of East and Central Africa, the Greater Mekong and its headwaters and the watersheds of the Andes? In order to answer this question, a conceptual analysis framework was designed that can be

used to explore the impacts of plausible futures on biodiversity and ecosystem services based on scenarios. A key component of this framework is the assessment of changes in the landscape that are the result of drivers (i.e. agricultural commodity trends) in the scenarios. These landscape changes were modelled using a landuse change model and used to assess current and future states of biodiversity and ecosystem services, which can then be compared to assess changes (Figure 1).

Plausible Futures..... LU Plans Planned infrastructure Commodity trends Future 1

Current

Scenario

Future 2

Future n

Landscape Change Current Status: Biodiversity, Pressure Ecosystem Function

Future Status: Biodiversity, Pressure Ecosystem Function Projected: Change in Biodiversity Change in Pressure on Biodiversity Change in Ecosystem Function

Figure 1: Conceptual analysis framework. Different socio-economic scenarios are characterised through changes in the landscape. Blue boxes represent the analysis of impacts for current and future status, allowing for assessment of changes between these two states.

15

3.2 STUDY AREAS AND SCALE OF ANALYSIS

16

All analyses for this study are carried out for the three “MacArthur regions” (Figure 2) as identified in scoping studies and expert consultations carried out by BirdLife International, Nature Serve, TNC, WWF and the Critical Ecosystem Partnership Fund (CEPF) for the MacArthur Foundation. The Great Lakes Region of East and Central Africa covers parts of ten countries: Tanzania, Zambia, Malawi, Kenya, Democratic Republic of Congo, Burundi, Rwanda, Uganda, South Sudan and Ethiopia. The Mekong region covers parts of China, Myanmar, Lao PDR, Thailand, Cambodia and Viet Nam, and the Watersheds of the Andes region covers parts of Venezuela, Colombia, Ecuador, Peru, Brazil and Bolivia. Each study region is subdivided into

watersheds and these watersheds are used as the units of analysis, i.e. all results show single values for each watershed. This approach was used as the watershed units are an appropriate scale for the assessment of impacts at the regional level, as well as for the identification of areas for action. Furthermore, this study utilises scenarios in combination with spatially explicit land-use change modelling and an analysis of impacts, each of which has some uncertainty attached, and this is propagated throughout the analysis framework. These uncertainties are further explored in the Methods discussion (Chapter 4). By aggregating the impact results to watersheds some of the spatial uncertainty deriving from the land-use change modelling is accounted for.

Figure 2: MacArthur Foundation focal regions and watersheds within each region. The MacArthur sub regions are three countries in each region for which higher resolution analysis was carried out (see section 2.4).

3.3 KEY CRITERIA The key criteria explored in the analysis framework are biodiversity, pressure on biodiversity and ecosystem functions, and these are addressed through the following key questions: ●B  iodiversity:

what is the estimated importance of the watershed, relative to the MacArthur region for biodiversity?

17

●P  ressure on

biodiversity: what is the pressure on biodiversity in the watershed relative to the pressure on biodiversity in the region?

●E  cosystem

function: what is the estimated importance of the watershed, relative to the MacArthur region, for potential ecosystem function provision?

3.4 GLOBAL AND REGIONAL SCALE ANALYSES The study has been carried out at two different scales (Figure 3): 1) an analysis based on the Global Environment Outlook (GEO-4) (UNEP, 2007) scenarios of change up to 2050 with land-use change modelling done at 5 arc minute resolution (~9km) and impacts on biodiversity and ecosystem function calculated for all watersheds within each of the three MacArthur

regions, and 2) a regional-scale analysis based on regionally-developed scenarios for three countries in each MacArthur region with landuse change modelling done at higher resolution (~1km) and impacts on biodiversity and ecosystem function assessed for all watersheds within these three countries.

Analysis based on global GEO-4 scenarios Global GEO-4 scenarios to 2050

IMPACT model

LandSHIFT land use change model at ~ 9 km resolution

Biodiversity and ES importance in watershed

Analysis based on regional scale scenarios Regional scenarios to 2050

IMPACT model

Figure 3: Modelling components for two scales of analysis.

LandSHIFT land use change model at ~ 1 km resolution

Biodiversity and ES importance in watershed

3.5 FOCAL QUESTIONS

18

The analysis framework generates a range of outputs. For the biodiversity metric, four types of biodiversity subsets are considered: birds, mammals, amphibians and threatened species. All these biodiversity subsets can be compared for each watershed for the baseline situation as well as for four global and four regional future scenarios (described in section 2.6.). Similarly, for the ecosystem function metric there are three main groups of ecosystem functions (cultivated products, wild products and regulating functions) and a further four sub-groups within regulating ecosystem functions. Again, these are assessed for each watershed for all scenarios. While all these criteria are useful to assess current and future conditions from different perspectives, six focal questions were identified that would be helpful to guide the analyses and navigate the results. These focal questions were used in the design of the web-based tool described in the results and Appendix V. The six focal questions are:

1) How do watersheds within a region compare in their current importance for biodiversity? 2) How do watersheds compare in their current importance for ecosystem function provision? 3) How do watersheds compare in their loss of suitable habitat for biodiversity as a result of projected land-use change? 4) How do watersheds compare in their loss of suitable habitat for ecosystem function provision as a result of projected land-use change? 5) How do watersheds compare in terms of future threat from agriculture and their current importance for biodiversity under a given future? 6) How do watersheds compare in terms of future threat from agriculture and their current importance for ecosystem function provision under a given future?

3.6 SCENARIOS The analyses have been carried out at two scales using different sets of socio-economic scenarios and derived drivers of change. The global scale scenarios used the Global Environment Outlook 4 scenarios (UNEP 2007); these were developed as part of the fourth Global Environment Outlook commissioned by UNEP, which assesses the state of the global atmosphere, land, water and biodiversity, changes since 1987 and priorities for action as well as plausible futures for the year 2050.

The regional-scale scenarios were developed in collaboration with the Climate Change Agriculture and Food Security research program of the CGIAR (CCAFS). This research program aims to develop plausible qualitative and quantitative future scenarios that can be used to explore consequences of socio-economic as well as governance assumptions and resulting impacts on food security, environments and livelihoods. These scenarios are developed for a number of regions, each covering several countries (i.e. East Africa, South Asia and South America). Both sets of scenarios are discussed in the sections below.

GEO-4 scenarios The GEO-4 scenarios consist of four plausible trajectories of change between 2007 and 2050. While all scenarios are global, they were developed at regional and sub-regional levels according to UNEP GEO regions. The scenarios consist of a narrative describing the current status and trends, drivers and a storyline into the future as well as a vision for that future. All scenarios were subsequently quantified using multi-model exercises. There are four scenarios named after the key underlying societal focus. The scenarios can be summarised as follows (UNEP 2014): Markets First: the private sector, with active government support, pursues maximum economic growth as the best path to improve the environment and human well-being. Lip service is paid to the ideals of the Brundtland Commission, Agenda 21 and other major policy decisions on sustainable development. There is a narrow focus on the sustainability of markets rather than on the broader human-environment system. Technological fixes to environmental challenges are emphasised at the expense of other policy interventions and some tried-andtested solutions. Policy First: government, with active private and civil sector support, initiates and implements strong policies to improve the environment and human well-being, while still emphasizing economic development. Policy First introduces some measures aimed at promoting sustainable development, but the tensions between environment and economic policies are biased towards social and economic considerations. Still, it brings the idealism of the Brundtland Commission to overhauling the environmental policy process at different levels, including efforts to implement the recommendations and agreements of the Rio Earth Summit, the World Summit on Sustainable Development (WSSD), and the Millennium Summit. The emphasis is on more top-down approaches, due in part to desires to make rapid progress on key targets.

Security First: government and private sector compete for control in efforts to improve, or at least maintain, human well-being for mainly the rich and powerful in society. Security First, which could also be described as Me First, has as its focus a minority: rich, national and regional. It emphasises sustainable development only in the context of maximizing access to and use of the environment by the powerful. Contrary to the Brundtland doctrine of interconnected crises, responses under Security First reinforce the silos of management, and the UN role is viewed with suspicion, particularly by some rich and powerful segments of society. Sustainability First: government, civil society and the private sector work collaboratively to improve the environment and human well-being, with a strong emphasis on equity. Equal weight is given to environmental and socio-economic policies, and accountability, transparency and legitimacy are stressed across all actors. As in Policy First, it brings the idealism of the Brundtland Commission to overhauling the environmental policy process at different levels, including strong efforts to implement the recommendations and agreements of the Rio Earth Summit, WSSD, and the Millennium Summit. Emphasis is placed on developing effective public-private sector partnerships, not only in the context of projects but also that of governance, ensuring that stakeholders across the spectrum of the environment development discourse provide strategic input to policymaking and implementation. There is an acknowledgement that these processes take time, and that their impacts are likely to be more long term than short term.

19

Regional scenarios

20

East Africa region The CCAFS scenarios for the East Africa region were developed in 2010 and 2011 through a total of four multiple-stakeholder workshops with participants from Kenya, Tanzania, Ethiopia, Uganda, Rwanda and Burundi as well as regional actors. Four scenarios were developed starting with a vision for 2030, with the scenario narratives describing the trends and events since 2010. The scenarios were subsequently quantified using the IMPACT model (Rosegrant et al. 1995) which is described in more detail in section 2.7. Initially, the quantification with the IMPACT model was done up to 2030, but later extended to 2050 for this study to be consistent with scenarios and analyses for the two other MacArthur regions. Drivers used as inputs into both models were based on interactions with diverse regional stakeholders involved in the CCAFS scenarios process, who provided semi-quantitative assessments of these drivers of change and the assumptions behind those assessments. The following drivers were included in this exercise: population, gross domestic product, production costs, crop yields, crop production systems, livestock numbers, yields and production systems and finally land-use change emissions tax. The most important drivers of change for the 2010-2030 period in the region were perceived to be population and climate change. Population projections in the scenarios up to 2050 follow those of the IPCC Fifth Assessment Report (AR5) global socioeconomic scenario SSP2 (O’Neill et al. 2012). Climate change is based on the Representative Concentration Pathways (RCP) 8.5 scenario (Riahi et al. 2012). Furthermore, increased drought periods were assumed in all scenarios with a severe drought occurring in the 2020–2022 period. The final four scenarios are summarised below with extended narratives in Appendix I. These scenarios are described in more detail in Vervoort et al. (2013).

Industrious Ants. This scenario is characterised by proactive governance, and high regional integration with a wide range of benefits for food security, environments and livelihoods. However, there are difficult international relations, a costly battle with corruption and challenges posed by being competitive with crops and products aimed at domestic markets. Herd of Zebra. In this scenario, there is an economic boom where regions reach out to international markets. However, the scenario is not economically sustainable, with trade-offs between food security and the environment, dependency on service and industrial markets, and new vehicles for corruption weakening effectiveness. Lone Leopards. This scenario is characterised by visionary actions carried out by individual organisations and initiatives facilitated by governments. It is a world of winners and losers, with uncoordinated trade and shared resources, instability, selfish behaviors and corruption preventing coordination. Sleeping Lions. This scenario is characterised by massive public mobilisations, international investments, informal trade, a personal sense of community and psychological resilience. Governments in 2030 act in self-interest, allowing rein of foreign interests and making money through crises. It is a scenario with no win-win situations, latent capacity and wasted opportunity. Revolutions are common and lead nowhere.

Mekong region The regional scale scenarios for the Mekong region were developed through a single multistakeholder workshop held in late 2013 convened by CCAFS, UNEP-WCMC and the Economics and Policy Innovations for Climate-smart Agriculture (EPIC) program of FAO. The workshop focused on three countries within the Greater Mekong Region: Lao PDR, Viet Nam and Cambodia for the time-period 2013 (now) until 2050. Similar to the East Africa scenarios, the qualitative and semi-quantitative scenarios were quantified with the IMPACT model (see section 2.7 for details). Scenarios were developed according to a set of change factors that related to the key elements of the workshop scope: agriculture, food security, livelihoods and environmental change. This exercise yielded a number of change factors such as GDP, population, agricultural yield, deforestation, pollution etc. These change factors were then clustered and ranked on a scale of 1-3

for both relevance and uncertainty and plotted. The top four factors that were selected using this process are: markets, enforcement capacity and regional collaboration, agricultural investment, and land degradation through land-use change. For each of the four change factors, participants then identified two or three “extreme” states. For example for the factor “markets” two extreme states would be: “unregulated market” and “common regulated market”. A factor-state compatibility matrix was created that showed all combinations of factor states. Each combination of factor states was then ranked according to their compatibility using a scale of 0-2, where 0 meant ‘not possible’, 1 meant ‘uncertain/ disagreement’ and 2 meant ‘possible’. Finally, a software program especially developed for this purpose (OLDFAR, Mason-d’Cruz 2015) was used to select the six most diverse scenarios, and participants then selected the four most useful of these (Table 1).

Table 1: Scenarios and states for key change factors for Mekong regional scales scenarios selected and, for each of these, the combination of different extreme states for the different change factors selected. Enforcement capacity and regional collaboration

Agricultural investment

Land degradation through land-use change

Factors

Markets

Land of the Golden Mekong

Common regulated market

Strong enforcement High public and and strong regional private collaboration

Low

Buffalo, Buffalo

Unregulated

Weak enforcement and weak regional collaboration

Unbalanced: high private investment in business and research

High

The Doreki Dragon

Common regulated market

Strong enforcement Unbalanced: high and strong regional private investment collaboration in business and research

High

Tigers on the Train

Protectionism and closed market

Strong enforcement Low public and and strong regional private collaboration

Low

21

22

For each of these four scenarios, the change factor states were taken as the end state for that scenario in 2050. Participants then considered what each state would look like concretely and then worked backwards using narrative flowcharts to explain the pathway from 2050 to the present day (explorative back-casting). This process generated the four scenario narratives summarised below: Land of the Golden Mekong. In this scenario, Southeast Asia becomes more unified in terms of political, economic and environmental concerns. There is a common regulated market, with high public and private agricultural investment. There is strong enforcement and regional collaboration; however migration is a challenge, with migrants becoming vulnerable and abused. Climate resilience is strong and land degradation levels are low. Buffalo, Buffalo. In this scenario, the ASEAN agreements move forward from 2013, but more problems appear nearing 2020. Markets are unregulated and investment is unbalanced, with high private investment in business and research. Environmental challenges cause regional tensions and weak regional collaborations. Agricultural intensification is unsustainable, with high levels of land degradation due to land-use change. DoReKi Dragon. This scenario is characterised by significant changes driven by ASEANfacilitated development and strong political focus. There is a common regulated market and strong enforcement capacity and regional collaboration. However, investments in agriculture are unbalanced – smallholder farmers struggle – and there are high levels of urbanisation and environmental degradation. Tigers on a Train. This scenario is characterised by strong regional collaborations and a protectionist and closed market. Agricultural investments are not entirely effective and investments by the public and private sector are low. Economic fragility threatens food security, but land degradation is low.

Andes region Similar to the Mekong region, the regional scenarios for the Andes region were developed during a single multi-stakeholder workshop held in Cali, Colombia at the end of 2013. The scenario workshop was jointly organised between CCAFS, UNEP-WCMC and the Universidad para la Cooperación Internacional (UCI) based in Costa Rica. The scenarios were developed for three countries in the Andean region: Colombia, Ecuador and Peru for the period up to 2050. The method to derive the scenarios was similar to that that used in the Mekong region. A set of change factors was defined and scored in terms of relevance and uncertainty, resulting in four key change factors: the state’s level of political power, market, consumption patterns, and economic development. These were then assigned two to three extreme conditions (i.e. political power is centralised or decentralised) and from this set of combinations, the most diverse combinations were identified using the OLDFAR software tool. This resulted in six scenarios from which the participants chose four scenarios to further develop in storylines.

Table 2: Scenarios and states for key change factors for Andes regional scale scenarios and the combinations of extreme states for the four key change factors. Scenarios

Level of Political Power of the State

Consumption Patterns

Economic Development

New Dawn

Centralised

Sustainable and regulated

Sustainable

Low economic development and diversification

Andean Autumn

Centralised

Unsustainable and unregulated

Subsistence

Low economic development and specialisation

Overcoming Obstacles

Decentralised

Sustainable and regulated

Responsible

High economic development and specialisation

Flipping Burgers

Decentralised

Sustainable and regulated

Sustainable

High economic development and diversification

Market

These change factor states were taken as the end state for the given scenario in 2050 and explorative back-casting was used to explain the pathway from 2050 to the present day. This process generated four scenario narratives summarised below: New Dawn. This scenario is characterised by a regulated economy and centralised State government which tends to guarantee food security and regulatory environmental frameworks. Consumption patterns are sustainable, and economic development and diversification is low. Unmet social needs lead to strong conflicts and more active citizen participation that favours economic, political and social change. Andean Autumn. This scenario is characterised by a centralised state, but political growth is not paired with increased efficiency or reduced corruption. Markets are unsustainable and unregulated, and subsistence consumption patterns prevail. The region is isolated from the international community, leading to low economic development. Political and economic inequality are extreme.

Overcoming Obstacles. Early in this scenario the Andean region is in conflict with trends towards decentralisation, regulation and sustainability and hence political and economic conflicts. However, by 2050, the Andes Region has become part of the Pacific Economic Community, with a low carbon economy, high economic development and specialisation and responsible consumption patterns. Flipping Burgers. This scenario is characterised by the promotion of decentralised power in highly centralised States. Environmental controls are weak by 2020, but markets become sustainable and regulated. By 2030, economic integration is stronger, leading to high economic development and diversification. Food security is consolidated and consumption patterns are sustainable by the end of the scenario.

23

24

3.7 Q UANTIFICATION OF SCENARIOS - THE IMPACT MODEL The International Model for Policy Analysis of Agricultural Commodities and Trade (IMPACT) is a global partial equilibrium model focusing on the agricultural sector only and is developed to study the impacts of alternative futures on food security by modelling agricultural commodities supply, demand, trade and prices. The model was first developed in the early 1990s by the International Food and Policy Research Institute (IFPRI). The first significant results of the model were published as a 2020 vision discussion paper, Global Food Projections to 2020: Implications for Investment (Rosegrant et al. 1995), soon followed by a number of other studies looking at effects of population, investment and trade on food security with a particular focus on developing countries. The model can be used to examine the linkage between production of key food commodities and food demand and security at the national scale in the context of scenarios of future change. Since the model was first developed, a number of major improvements have been made. A key improvement was the expansion of the number of agricultural commodities to a total of 45, which include oilseeds (groundnuts, soybeans and rapeseed), cotton and major dry land grains and pulses. These commodities are key to understand the drivers behind the projected growth of global oil, meat and milk demand. A full list of commodities modelled by IMPACT is given in Appendix II.

The current model which was also used for this study is the IMPACT-WATER model. This model combines the original IMPACT model with a water simulation module (IWSM) that balances water availability and use within various economic sectors. In order to incorporate water availability at the basin level, the model uses socalled ‘food producing units’ or FPUs as its unit of analysis. These FPUs are a combination of 115 economic, geo-political regions with 126 water basins thus ensuring that climatic and hydrologic variations within regions are accounted for within hydrologically defined basins. Within the model, the supply and demand of water and crop production are first assessed at the basin scale and crop production is then summed to the national level at which food demand and trade are modelled. Climate change is incorporated in the model through a link with the crop-simulation model DSSAT (Decision Support System for Agrotechnology Transfer) which is able to assess climate change effects and CO2 fertilisation for wheat, rice, soybeans, groundnuts (all C3 pathway crops) and maize (C4 pathway). Other crops modelled in IMPACT are assumed to react similarly to climate change effects within the same geographical region and metabolic pathway (C3 or C4).

Drivers

Assumptions and limitations

Population projections in the model follow the medium variant growth projections from the UN population database 2010 revision (United Nations Population Division 2011). Gross Domestic Product (GDP) projections follow the narrative assumptions of the scenarios and are based on historic GDP with future projections constrained in a plausibility envelope derived from a study on Food Security, Farming and Climate change to 2050 (Nelson et al. 2010). Crop yields in the model follow technical improvements in the scenarios and in terms of cropping production system the model distinguishes between irrigated and rain fed systems. Livestock yields are driven by an exogenous yield trend.

Since IMPACT is a partial equilibrium model, it only models the agricultural subset of the market and therefore needs to parameterise the remaining markets. This approach lacks feedbacks with other sectors and thus that the markets of interest are negligible for the rest of the economy. A key assumption of this type of economic-based model is that people will seek to maximise utility, either in financial or commodity gains (Evans et al. 2001) which is not always the main factor affecting land allocation decisions (Walker 2004).

3.8 LANDSCAPE CHANGES - THE LANDSHIFT MODEL The LandSHIFT model framework (Schaldach et al. 2011) is a tool for medium-term scenario analysis (20-50 years) and assessment of environmental impacts of land-use change and is developed by the Centre for Environmental Systems Analysis of Kassel University, Germany. The model simulates spatial-temporal dynamics of settlement, crop cultivation and livestock grazing. LandSHIFT is based on the concept of “land-use systems” (Mather 2006) as it couples model components representing anthropogenic and environmental systems. Crop yields and net primary productivity (NPP) of grassland are simulated in a productivity module and land-use change is then simulated in a LUC module using demand for land intensive commodities and supply defined by the local biomass productivity in a specific cell. Productivity is influenced by climate change and technological changes.

The model is built around two main components, a land-use change module and a productivity module, which calculates crop yields and net primary productivity (NPP) which are important inputs to the land-use decisions of the LUC module. The model requires input from driving variables that describe the socio-economic and agricultural development of a country as well as micro-level variables (grid-scale) that describe local landscape characteristics (i.e. baseline land use, urban areas, protected areas, roads etc.). Outputs from the model consists of time-series of maps of land-use type as well as population and livestock densities. All output maps are produced for each time-step which is fixed at 5 years. Macro-level input data (i.e. socioeconomic data and agricultural production data) has to be at country scale while micro-level data can be specified for sub-regions.

25

26

LandSHIFT can be run at different resolutions and with different base land-use data; the choice of input land-use data determines the resolution of the output data. For this study, the global land cover dataset for the year 2000 (GLC2000) was used (Bartholomé & Belward 2005). This dataset provides high-resolution (30 arc seconds or ~ 1km), harmonised land cover for the globe based on satellite remote sensing data from the SPOT-4 VEGETATION sensor. GLC2000 uses

the FAO land-cover classification system (LCCS) with a total of 23 main land-use types classified. GLC2000 is widely used in studies requiring spatially explicit land-use information and is regarded to be a good representation of land use in the year 2000 (Fritz et al. 2011). LandSHIFT outputs all land-cover types from the base landuse dataset and sub-divides arable land classes into 20 different crop types (Table 3).

Table 3: Land-use/cover classes in GLC2000 and LandSHIFT. GLC2000 code GLC 2000 description

LandSHIFT code

1

Tree cover, broadleaved, evergreen

1

Tree cover, broadleaved, evergreen

2

Tree cover, broadleaved, deciduous, closed

2

Tree cover, broadleaved, deciduous, closed

3

Tree cover, broadleaved, deciduous, open

3

Tree cover, broadleaved, deciduous, open

4

Tree cover, needle-leaved, evergreen

4

Tree cover, needle-leaved, evergreen

5

Tree cover, needle-leaved, evergreen

5

Tree cover, needle-leaved, evergreen

6

Tree cover, mixed leaf type

6

Tree cover, mixed leaf type

7

Tree cover, regularly flooded, fresh and brackish water

7

Tree cover, regularly flooded, fresh and brackish water

8

Tree cover, regularly flooded, saline water

8

Tree cover, regularly flooded, saline water

9

Mosaic: tree cover/other natural vegetation

9

Mosaic: tree cover/other natural vegetation

10

Tree cover, burnt

10

Tree cover, burnt

11

Shrub cover, closed-open, evergreen

11

Shrub cover, closed-open, evergreen

12

Shrub cover, closed-open, deciduous

12

Shrub cover, closed-open, deciduous

13

Herbaceous cover, closed-open

13

Herbaceous cover, closedopen

14

Sparse herbaceous or sparse shrub cover

14

Sparse herbaceous or sparse shrub cover

15

Regularly flooded shrub and/or herbaceous cover

15

Regularly flooded shrub and/ or herbaceous cover

16

Cultivated and managed areas

16

Cultivated and managed areas

17

Mosaic: cropland/tree cover/other natural vegetation

17

Mosaic: cropland/tree cover/ other natural vegetation

18

Mosaic: cropland/shrub or grass cover

18

Mosaic: cropland/shrub or grass cover

LandSHIFT description

GLC2000 code GLC 2000 description

LandSHIFT code

LandSHIFT description

19

Bare areas

19

Bare areas

20

Water bodies

20

Water bodies

21

Snow and Ice

21

Snow and Ice

22

Artificial surfaces and associated areas

22

Artificial surfaces and associated areas

23

Irrigated agriculture

23

Irrigated agriculture

-

n/a

99

Set-aside

-

n/a

100

Default crop

-

n/a

101

Cassava

-

n/a

102

Temperate cereals

-

n/a

103

Tropical cereals

-

n/a

104

Cotton

-

n/a

105

Fruits

-

n/a

106

Groundnuts

-

n/a

107

Maize

-

n/a

108

Millet

-

n/a

109

Oil crops annual

-

n/a

110

Oil crops permanent

-

n/a

111

Pulses

-

n/a

112

Rice

-

n/a

113

Temperate roots and tubers

-

n/a

114

Tropical roots and tubers

-

n/a

115

Sorghum

-

n/a

116

Soybeans

-

n/a

117

Stimulants

-

n/a

118

Sugarcane

-

n/a

119

Vegetables

-

n/a

120

Wheat

-

n/a

200

Pasture

-

n/a

201

Range land

27

28

The land-use change module in LandSHIFT uses a transition matrix that defines which land-use types can turn into other land-use types. For instance, urban land can never turn into arable land whereas the reverse is possible. Constraints can also be policy decisions or laws prohibiting the conversion of, for instance, protected areas. All cells are assigned a land-use type on the basis of a preference ranking of suitability for a given land-use type, which is computed using a multicriteria analysis with weighting factors which determine the importance of individual factors (e.g. slope, roads, population density). Suitability factors and constraints are based on values from scientific literature, e.g. thresholds of population density before a cell turns into urban area. A set of key variables in the model are the crop-specific correction factors. These factors match countrylevel crop production (as reported by FAO statistics or from the driving economic model such as IMPACT), with the crop production resulting from a baseline simulation run based on the sum of the total calculated potential production of that crop type as present in the initial land-use/land-cover map. These factors can thus be used to account for agriculture management alternatives such as doublecropping which are not explicitly modelled in LandSHIFT.

Climate change is incorporated in the model through the productivity module. This module consists of the dynamic global vegetation model LPJmL which calculates crop yields for all crops considered in LandSHIFT under rainfed and irrigated conditions as well as NPP for pasture. LPJmL specifically simulates sowing dates, crop phenology, crop growth and carbon allocation at a daily time-step taking into account climate variables such as precipitation and temperature which can be based on different climate alternatives. All model runs in this study were carried out for a climate scenario based on the Representative Concentration Pathways (RCPs, Moss et al. 2010) using a global circulation model based on the 5th assessment report of the IPCC (AR5). The climate scenario used is the RCP 8.5 modelled using the IPSL climate model. RCP 8.5 has the most extreme emission pathway of all RCPs and assumes a continued rise in emissions throughout the 21st century and a mean global warming increase of 2˚C by 2050 (Riahi et al. 2011). For this study, all micro-level landscape characteristics data are based on global datasets and the same data are used for both the GEO-4 scenario modelling and the regional-scale analysis (Table 4).

Table 4: Spatial micro-level data used in LandSHIFT for model initialisation. Model variable

Dataset name

Native resolution

Temporal coverage

Land use/cover

GLC2000

30 arc- seconds

ca 2000

Bartholomé, E., & Belward (2005)

Population density

GRUMP

30 arc-seconds

ca 2000

CIESIN & CIAT (2005)

River network density

HydroSHEDS

15 arc seconds

2004

HydroSHEDS rivers of Africa, Asia and South America. Lehner et al. 2006

Road infrastructure

gROADSv1

30 arc seconds

1980-2010

CIESIN, 2013

Protected Areas

WDPA

shapefile

2013

UNEP-WCMC/ IUCN

Key biodiversity Areas

World Bird and Biodiversity areas database

shapefile

2013

BirdLife International

Source

Assumptions and limitations While the LandSHIFT model can be run at various grid resolutions, for this study it was decided to run the model at 5 arc minutes (~10 km) for the full extent of the MacArthur regions and at 30 arc seconds (~1km) for the regionalscale analyses for three countries in each region. The coarser resolution for the extent of the MacArthur regions was chosen, as running at higher resolutions for these regions would be too computationally demanding. However, this does slightly limit the spatial detail of land-use change. For instance, the GLC2000 basemap had to be resampled to the coarser resolution which was done using a majority resampling. Since this method assigns land-use type in the coarser grid cell based on the dominant landuse types at higher resolution within this grid cell, there is a loss of spatial detail. The regional runs were carried out at the native resolution of the GLC2000 basemap for land use/land cover (1km). Since the spatial resolution of the LPJmL crop model is 5 arc minutes (10km) not all parameters are fully resolved at this resolution. However, in the impact analysis for biodiversity and ecosystem functions, all crop types are aggregated within the watershed unit of analysis and so this does not have an effect on the final impact results. Furthermore, since the model uses dominant crop types the cropping pattern can seem artificial in comparison with approaches that assign fractional shares of different crops to grid cells. Again, while this would be an issue if individual crop areas need to be assessed within watersheds, for the impact analysis in this study the overall expansion of cropland under scenarios is more relevant.

There are very few land-cover products that have been consistently assessed at multiple points in time. The GLC2000 dataset is compiled using remote sensing data for roughly the year 2000. While this provides a good baseline, it means that it is not possible to validate the land-use change model with ‘observed’ data at a later point in time (i.e. after one or two time-steps) using a dataset which is compiled using the same algorithms and methods. The accuracy of GLC2000 land-cover types for different regions is found to be variable. Studies for East Africa have shown a good validation for a number of land-cover types for the East Africa region (Herold et al. 2008). For the South East Asia region, however, some discrepancies were found with regards to the total area reported as cropland in GLC2000 compared with actual crop areas reported by the Food and Agricultural Organisation (FAO). Some of these discrepancies are likely the result of differences in reporting of multi-cropping systems. Overall though the GLC2000 dataset provides adequate accuracy for this study. The protected areas data are used in the model as limitations to conversion. Model runs were carried out for scenarios where a) all current protected areas within the MacArthur region are restricted for conversion of land use, b) there are no restrictions in conversion of protected areas (i.e. the same rules apply within PAs as outside PAs with regards to the model’s preference for allocation of a land-use type) and c) a scenario whereby Key Biodiversity Areas are assumed to be future protected areas (i.e. no conversion allowed within KBAs). This last scenario was run to better understand spatial trade-offs in land conversion under different protected area management regimes.

29

3.9 BIODIVERSITY METRIC

30

Biodiversity was assessed at current (modelled for 2005) and future (2050) time periods, using a novel index for biodiversity. A further index was developed to calculate change between these periods. Biodiversity data were obtained from the IUCN Red List spatial dataset and an established method of linking species’ habitat preferences to land-use/land-cover types was used. Aspects of this approach are based on Buchanan et al. (2011), who used an impact metric based on range-rarity (weighted endemism) and IUCN species range data refined by land cover.

IUCN Red List spatial data: selection criteria All available species range data from the IUCN Red List (IUCN, 2014) were collated for vertebrate classes that have been comprehensively assessed: aves (birds), amphibia (amphibians) and mammalia (mammals). It is important to note that these data provide the extent of occurrence of each species, which is defined as ‘the area contained within the shortest continuous imaginary boundary which can be drawn to encompass all the known, inferred or projected sites of present occurrence of a taxon’ (IUCN 2014). Therefore, unsuitable or unoccupied habitats may be included in these ranges. The dataset was filtered to include those ranges listed as extant, native or reintroduced, and with seasonal attributes listed as either resident, resident breeding or resident non-breeding. This step removed ranges of species that were extinct or probably extinct, where a species has been newly introduced (i.e. invasive), or is likely present for only brief period, such as on a migratory passage. From species that remained, all further analyses were carried out on those whose ranges intersected any of the MacArthur regions and for which IUCN habitat preference data were available. Each range polygon was simplified to remove unnecessary detail (i.e. vertices within 10m of each other), thus improving processing speed.

IUCN habitat preferences Additional non-spatial data on IUCN Habitat preferences were compiled which lists each species with their IUCN habitat classes, based on expert opinion and literature (IUCN 2014a). A small number of species (~0.4%) were lacking such data and were excluded from analysis. Only habitat categories classed as suitable were included in the analysis, thus excluding marginal habitats as these were less relevant to the aims of the study. Within the suitable habitat category no distinction was made for those classed as major habitats, as only half the species had habitats in this sub-category.

IUCN-LandSHIFT Crosswalk From the list of suitable habitats for each species, a corresponding list of suitable LandSHIFT landuse types was derived using a crosswalk table based on Foden et al. (2013). This aims to provide a link between species habitat preferences and modelled land use. This crosswalk was originally created to allow the refinement of IUCN species’ ranges by linking IUCN habitat classes to Land Cover Classification System (LCCS), the classification system used by Global Land Cover 2000. Although GLC2000 was used as a basis for modelling, the LandSHIFT model has additional classes for crop, pasture and set-aside areas, thus a number of additional links were added to these classes (see Appendix IV). The spatial linkage between a species range and modelled land use was based on the total area of a species range overlapping with the watershed and the area of suitable land cover in that watershed (Figure 4). The final equations used to calculate the biodiversity importance indices are outlined in the section below.

Datasets Range (EOO) of Species X

Land cover map

Watersheds

A

31 Spatial analysis Extract statistics: 1. Range of Species X in Watershed A 2. All suitable habitat for Species X in Watershed A

1. Range of Species X in Watershed A

2. All suitable habitat for Species X in Watershed A

Calculate: 3. Proportion of Watershed A covered by range of Species X 4. Predicted utilised suitable habitat for Species X in Watershed A (step 3 multiplied by step 2) 5. Predicted utilised suitable habitat for Species X in the region 6. Proportion of the range of Species X in the region

Figure 4: Workflow for mapping biodiversity in watersheds. Data on species range (blue) was combined with modelled land-cover data and watershed polygons. The areas of species X’s range (1) and suitable habitat (2) overlapping with watershed A were extracted. The proportion of watershed A covered by species X (3) was calculated and multiplied by suitable habitat for the species, to derive the predicted utilised suitable habitat in Watershed A (4). The area of predicted utilised suitable habitat for species X in the region (5) and proportion of the range of species X in the region (6) were also calculated.

Biodiversity importance index The relative biodiversity importance in a given watershed is calculated as the sum of the area of predicted utilised suitable habitat within the watershed, divided by the total area of predicted utilised suitable habitat within the region (i.e. summed over all watersheds) multiplied by the relative overlap of that species range with the region to the total extent of occurrence of that species (Equation 1). The second term in the equation was introduced to give more weight to species that are endemic to a region.

Equation 1:

⎛ ⎜ H wi T0 ⎜ importancewi = ⎜ w ⎜⎜ ∑ H wiT 1 0 ⎝

⎞ ⎟ ⎟ Ri ⎟ × ⎟⎟ EOOi ⎠

With H: Area of predicted utilised suitable habitat for a species (i) in a watershed (w) in the baseline situation (T0) EOOi: The species total extent of occurrence, Ri: Overlap of the EOO with the region. Equation 1 describes how to calculate the importance for a single species in the baseline situation. In order to calculate the importance for all species (s) Equation 1 is summed over all species (Equation 2).

Equation 2:

32

⎛ ⎛ ⎜ ⎜ H wi T0 ⎜ = ∑ ⎜ ⎜⎜ W 1 ∑ ⎜ ⎜⎜ 1 H wiT ⎜ 0 ⎝ ⎝ i

importancews

⎞ ⎞ ⎟ ⎟ ⎟ × Ri ⎟ ⎟ ⎟ ⎟⎟ EOOi ⎟⎟ ⎠ ⎠

The above equations can be used to assess the relative importance of a watershed for a single or multiple species for the baseline or a future situation (states). However, to assess the change in biodiversity importance for a species between the baseline and a future situation (i.e. under a given scenario) the relative change in importance was assessed relative to the importance for that species in the baseline situation and summed over all watersheds (i.e. the region (r)) (Equation 3). This is necessary since the state situations are relative to the total summed importance at that time-period (baseline or future) and thus direct comparisons between these state situations would not be a true reflection of change. Equation 3:

⎛ ⎛ ⎜ ⎜ H wi − H wi T1 T0 ⎜ Δimportanceir = ∑ ⎜ ⎜⎜ w 1 ∑ H wi ⎜ ⎜⎜ 1 T0 ⎜ ⎝ ⎝ w

⎞ ⎞ ⎟ ⎟ ⎟ × Ri ⎟ ⎟ ⎟ ⎟⎟ EOOi ⎟⎟ ⎠ ⎠

With T1 the future period. Finally, to assess the change for all species or group of species in a watershed, the above equation is summed over all species: Equation 4:

⎛ ⎛ ⎜ ⎜ H wi − H wi i T1 T0 ⎜ Δimportancews = ∑ ⎜ ⎜⎜ w 1 ∑ H wi ⎜ ⎜⎜ 1 T0 ⎜ ⎝ ⎝

⎞ ⎞ ⎟ ⎟ ⎟ × Ri ⎟ ⎟ EOOi ⎟⎟ ⎟⎟ ⎟ ⎠ ⎠

Biodiversity importance (Equation 2) and change metrics (Equation 4) were calculated for watersheds in each of the three regions for baseline and future scenarios. These results were based on all species combined and for subsets of i) birds, ii) amphibians, iii) mammals and iv) threatened species (i.e. Critically Endangered, Endangered or Vulnerable species, according to the IUCN Red List) (see Appendix IV).

Changes in biodiversity importance were normalised between scenarios to allow visual comparison of results for different scenarios in a given region. Increases in biodiversity importance were normalised from 0 to 1, with 1 representing the highest increase in all scenarios. Decreases in importance (i.e. negative values) were normalised similarly, from 0 to -1. Zero values represented areas of no change in importance. Normalised values were grouped in 7 classes (1 to >0.5, 0.5 to >0.05, 0.05 to >0, 0,