This presentation takes a close look at the data and methodology behind WRI’s brand new Aqueduct water risk mapping tool (http://aqueduct.wri.org/) which includes 12 new indicators of water-related risk. Through a step by step description of how the Aqueduct water risk maps were created, it covers the hydrological modeling and data sources used to compute all 12 indicators of water-related risk, as well as the methodology used to weigh and aggregate each indicator into physical, regulatory, reputational and overall water risk scores.

1. AQUEDUCT
DATA & METHODOLOGY
PAUL REIG, TOM PARRIS, AND FRANCIS GASSERT February 20, 2013

2. AGENDA
 Introduction
 Modeling Water Supply and Demand – Tom Parris
 Aqueduct Indicators – Paul Reig
 Aggregation and scoring – Francis Gassert
 Q&A

3. RISKS TO:
GOVERNMENTS
COMPANIES
INVESTORS
Photo: NOAA

4. DETAILED, COMPARABLE, GLOBAL
WATER RISK INFORMATION
Photo: NASA

5. UNDERSTANDING RISK: AQUEDUCT FRAMEWORK
Overall Water Risk
Physical Risk: Physical Risk: Regulatory &
QUANTITY QUALITY Reputational Risk
 Baseline water stress
 Inter-annual variability
 Seasonal variability  Media coverage
 Return flow ratio
 Flood occurrence  Access to water
 Upstream protected land
 Drought severity  Threatened amphibians
 Upstream storage
 Groundwater stress

6. WHERE WE PILOT TESTED: KEY BASINS

7. HOW WE EVOLVED: GLOBAL MAPPING

8. BUILDING ON SCIENCE: EXPERT REVIEWERS
 CDP Water Disclosure Project  The World Bank
 Ceres  US Environmental Protection Agency
 Columbia University  University of Michigan at Ann Arbor
 Deloitte Consulting LLP  University of North Carolina Chapel Hill
 Global Adaptation Institute  University of Virginia
 Global Water Strategies  Water Footprint Network
 Nanjing University  World Business Council for Sustainable
Development
 National Geographic
 Yale University
 Pacific Institute
 The Nature Conservancy

9. DATA: SELECTION CRITERIA
Step 1: Literature review
Step 2: Public domain & global coverage
Step 3: Comparative analysis to evaluate:
 granularity,
 time coverage
 source
Step 4: Selection of data source

10. Tom Parris | Vice President | ISciences
MODELING WATER SUPPLY AND DEMAND

11. UNDERSTANDING RISK: AQUEDUCT FRAMEWORK
Overall Water Risk
Physical Risk: Physical Risk: Regulatory &
QUANTITY QUALITY Reputational Risk
 Baseline water stress
 Inter-annual variability
 Seasonal variability  Media coverage
 Return flow ratio
 Flood occurrence  Access to water
 Upstream protected land
 Drought severity  Threatened amphibians
 Upstream storage
 Groundwater stress

12. OVERVIEW
MODELING: WATER SUPPLY AND DEMAND
Hydrologically
Withdrawals &
connected Runoff
consumptive use
catchments
 Catchments
 Demand: Withdrawals & Flow accumulator
consumptive use
 Source: Runoff
 Flow accumulation
Total blue Available blue
 Supply: Total and available water (Bt) water (Ba)
blue water

13. Catchments
CATCHMENTS
 Global Drainage Basin Database (GDBD; n=73074)
 Aggregated to <100,000 km2 target
 Mean area = 8804 km2 (n=14998)

14. WATER DEMAND: BASELINE 2010
1. Source data: withdrawals by sector (agricultural, domestic, and
industrial) reported by country (FAO), baselined to 2010
2. Spatially disaggregate by sector
 Agricultural withdrawals disaggregated by Global Map of Irrigated Areas
(2000)
 Domestic withdrawals disaggregated by Gridded Population of the World
(2010) and Nighttime Lights (2010)
 Industrial withdrawals disaggregated by Nighttime Lights (2010)
3. Multiply each sector by consumptive use ratio
4. Re-aggregate to basins and sum sectoral consumptive and total
withdrawals

15. Withdrawal Disaggregation Methodology
WATER DEMAND: INPUT DATA
Sector Variable Source
Water withdrawals FAO Aquastat
All sectors
Pacific Institute
GDP World Bank
Population World Bank
Average annual precipitation FAO Aquastat
Total renewable water supply FAO Aquastat
Sectoral water withdrawal ratio Calculated
Irrigated area FAOSTAT
Agricultural FAO Aquastat
Freydank & Siebert 2008
Agricultural land area World Bank
Industrial CO2 emissions World Bank
Electricity, total net generation Energy Information Administration
Coal production Energy Information Administration
Refinery Processing Gain Energy Information Administration
Domestic Urban population World Bank

16. WATER DEMAND: BASELINE TO 2010
 Start with reported national withdrawals (FAO
Aquastat) by sector (domestic, industrial,
agricultural) (black points)
 Project using two random and two fixed effects
regression models (light points)
 Withdrawals reported 2008-2010 were not
modeled (vertical dashed line)
 Average four models to estimate national
withdrawals for 2010 (dark red points)
 Regressions explained 94-99% of the total
variation.

17. Withdrawal Disaggregation Methodology
WATER DEMAND: DISAGGREGATION
1. Input data: withdrawals by sector (agricultural, domestic, and industrial)
reported by country (FAO), baselined to 2010
2. Spatially disaggregate by sector
 Agricultural withdrawals disaggregated by Global Map of Irrigated
Areas (2000)
 Domestic withdrawals disaggregated by Gridded Population of the
World (2010) and Nighttime Lights (2010)
 Industrial withdrawals disaggregated by Nighttime Lights (2010)
3. Multiply each sector by consumptive use ratio
4. Re-aggregate to basins and sum sectoral consumptive and total
withdrawals

18. WATER DEMAND: AGRICULTURE
 Global Map of Irrigation Area

19. WATER DEMAND: INDUSTRY
 Nighttime Lights (2010)

20. WATER DEMAND: DOMESTIC
 Gridded Population of the World
(2010)
 Nighttime Lights (2010)

21. WATER DEMAND: TOTAL WITHDRAWALS
Total withdrawals (Ut) are the sum of agricultural,
𝑈 𝑡 = 𝑈 𝑎𝑎𝑎 + 𝑈 𝑑𝑑𝑑 + 𝑈 𝑖𝑖𝑖
domestic, and industrial withdrawals

22. WATER DEMAND: CONSUMPTIVE USE
1. Input data: withdrawals by sector (agricultural, domestic, and industrial)
reported by country (FAO), baselined to 2010
2. Spatially disaggregate by sector
 Agricultural withdrawals disaggregated by Global Map of Irrigated Areas
(2000)
 Domestic withdrawals disaggregated by Gridded Population of the World
(2010) and Nighttime Lights (2010)
 Industrial withdrawals disaggregated by Nighttime Lights (2010)
3. Multiply each sector by consumptive use ratio
4. Re-aggregate to basins and sum sectoral consumptive and total
withdrawals

23. WATER DEMAND: CONSUMPTIVE USE
Consumptive use (Ct) is the sum of sectoral use times the sectoral
consumptive use ratio (cr) (Shiklomanov and Rodda 2003)
𝐶𝑡 = 𝑈 𝑎𝑎𝑎 × 𝑐𝑐 𝑎𝑎𝑎 + 𝑈 𝑑𝑑𝑑 × 𝑐𝑐 𝑑𝑑𝑑 + 𝑈 𝑖𝑖𝑖 × 𝑐𝑐𝑖𝑖𝑖

24. WATER SOURCE: RUNOFF
 evaluated 4 datasets (UNH/GRDC, CFSR, MERRA-Land, GLDAS-2)
 GLDAS-2, 1°, NOAH monthly (summed to annual)
 1948-2008 (used 1950-2008)
 directly resampled to 1km without interpolation

25. MODELING: FLOW ACCUMULATION
runoffa
 Runoff
 Precipitation minus
evapotranspiration, loss to
groundwater, and increase in soil runoffb
Bta
moisture.
 Total Blue Water (Bt)
 Accumulated runoff
 Equivalent to naturalized flow.
Btb

26. MODELING: FLOW ACCUMULATION
runoffa
 Runoff
 Precipitation minus evaporation,
transpiration, deep groundwater
recharge, and change in soil moisture
Baa
runoffb
 Total Blue Water
 Accumulated runoff consumptive
usea
 Equivalent to naturalized flow
 Available Blue Water (Ba)
Bab
 Upstream runoff minus consumptive
use, plus runoff
 Loosely equivalent to surface water consumptive useb
and shallow groundwater

27. WATER SUPPLY: TOTAL BLUE WATER
Total blue water (Bt) is the sum of
naturalized (uninhibited) runoff

28. WATER SUPPLY: AVAILABLE BLUE WATER
Available blue water (Ba) is the sum of
upstream runoff minus consumptive use

29. Paul Reig | Aqueduct Project | WRI
AQUEDUCT’S WATER RISK INDICATORS

30. UNDERSTANDING RISK : AQUEDUCT FRAMEWORK
Overall Water Risk
Physical Risk: Physical Risk: Regulatory &
QUANTITY QUALITY Reputational Risk
 Baseline water stress
 Inter-annual variability
 Seasonal variability  Media coverage
 Return flow ratio
 Flood occurrence  Access to water
 Upstream protected land
 Drought severity  Threatened amphibians
 Upstream storage
 Groundwater stress

31. SUPPLY AND DEMAND INDICATORS
𝑇𝑇𝑇𝑇𝑇 𝑤𝑤𝑤𝑤𝑤 𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 (2010)
𝐁𝐁𝐁𝐁𝐁𝐁𝐁𝐁 𝐰𝐰𝐰𝐰𝐰 𝐬𝐬𝐬𝐬𝐬𝐬 =
𝑀𝑀𝑀𝑀 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑏𝑏𝑏𝑏 𝑤𝑤𝑤𝑤𝑤 (1950 − 2008)

32. SUPPLY AND DEMAND INDICATORS
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 𝑜𝑜 𝑡𝑡𝑡𝑡𝑡 𝑏𝑏𝑏𝑏 𝑤𝑤𝑤𝑤𝑤 (1950 − 2008)
𝐈𝐈𝐈𝐈𝐈𝐈𝐈𝐈𝐈𝐈𝐈 𝐯𝐯𝐯𝐯𝐯𝐯𝐯𝐯𝐯𝐯𝐯 =
𝑀𝑀𝑀𝑀 𝑡𝑡𝑡𝑡𝑡 𝑏𝑏𝑏𝑏 𝑤𝑤𝑤𝑤𝑤 (1950 − 2008)

33. SUPPLY AND DEMAND INDICATORS
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 𝑡𝑡𝑡𝑡𝑡 𝑏𝑏𝑏𝑏 𝑤𝑤𝑤𝑤𝑤 (1950 − 2008)
𝐒𝐒𝐒𝐒𝐒𝐒𝐒𝐒 𝐯𝐯𝐯𝐯𝐯𝐯𝐯𝐯𝐯𝐯𝐯 =
𝑀𝑀𝑀𝑀 𝑚𝑚𝑚𝑚𝑚𝑚𝑚 𝑡𝑡𝑡𝑡𝑡 𝑏𝑏𝑏𝑏 𝑤𝑤𝑤𝑤𝑤 (1950 − 2008)

34. SUPPLY AND DEMAND INDICATORS
𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈 𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐
𝐔𝐔𝐔𝐔𝐔𝐔𝐔𝐔 𝐬𝐬𝐬𝐬𝐬𝐬𝐬 =
𝑀𝑀𝑀𝑀 𝑡𝑡𝑡𝑡𝑡 𝑏𝑏𝑏𝑏 𝑤𝑤𝑤𝑤𝑤 (1950 − 2008)
Data: Lehner et al. GRanD

35. SUPPLY AND DEMAND INDICATORS
𝑇𝑇𝑇𝑇𝑇 𝑏𝑏𝑏𝑏 𝑤𝑤𝑤𝑤𝑤 𝑓𝑓𝑓𝑓 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑙𝑙𝑙𝑙𝑙 1950 − 2008
𝐔𝐔𝐔𝐔𝐔𝐔𝐔𝐔 𝐩𝐩𝐩𝐩𝐩𝐩𝐩𝐩𝐩 𝐥𝐥𝐥𝐥 =
𝑀𝑀𝑀𝑀 𝑡𝑡𝑡𝑡𝑡 𝑏𝑏𝑏𝑏 𝑤𝑤𝑤𝑤𝑤 1950 − 2008
Data: IUCN, UNEP

36. SUPPLY AND DEMAND INDICATORS
𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈 𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛 𝑢𝑢𝑢 (2010)
𝐑𝐑𝐑𝐑𝐑𝐑 𝐟𝐟𝐟𝐟 𝐫𝐫𝐫𝐫𝐫 =
𝑀𝑀𝑀𝑀 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑏𝑏𝑏𝑏 𝑤𝑤𝑤𝑤𝑤 (1950 − 2008)

37. UNDERSTANDING RISK : AQUEDUCT FRAMEWORK
Overall Water Risk
Physical Risk: Physical Risk: Regulatory &
QUANTITY QUALITY Reputational Risk
 Baseline water stress
 Inter-annual variability
 Seasonal variability  Media coverage
 Return flow ratio
 Flood occurrence  Access to water
 Upstream protected land
 Drought severity  Threatened amphibians
 Upstream storage
 Groundwater stress

38. OTHER INDICATORS
Indicator Data Source Scale
Brakenridge, Dartmouth
Flood occurrence Polygons
Flood Observatory
1 degree raster
Drought severity Sheffield and Wood
Groundwater stress Gleeson et al. Major aquifers
Media coverage Google Country
Access to water WHO, UNICEF Country
Threatened amphibians IUCN Red List Polygons

39. Francis Gassert | Aqueduct Project | WRI
AQUEDUCT’S WATER RISK FRAMEWORK

40. UNDERSTANDING RISK : AQUEDUCT FRAMEWORK
Overall Water Risk
Physical Risk: Physical Risk: Regulatory &
QUANTITY QUALITY Reputational Risk
 Baseline water stress
 Inter-annual variability
 Seasonal variability  Media coverage
 Return flow ratio
 Flood occurrence  Access to water
 Upstream protected land
 Drought severity  Threatened amphibians
 Upstream storage
 Groundwater stress

41. ENABLING COMPARABILITY : THRESHOLDS
Thresholds:
1. Create categories for communication
2. Enable scoring and aggregation
Thresholds determined using:
 existing literature
 governmental or intergovernmental guidelines
 range and distribution of indicator values
 expert judgment

42. NORMALIZING AND SCORING
 Values mapped over thresholds using continuous functions
Raw value (r) : 0.1 0.2 0.4 0.8
Score : 0 1 2 3 4 5
ln 𝑟 − ln 𝑡1
𝑓BWS 𝑟 = min(5, max(0, + 1 ))
ln 𝑏𝑏𝑏𝑏

43. PUTTING IT TOGETHER : AGGREGATION

44. PUTTING IT TOGETHER : AGGREGATION
Weighted average
 For each region (j):
∑ 𝑥 𝑖𝑖 𝑤 𝑖
𝑎𝑗 = 𝑓𝑓𝑓 𝑖 𝑖𝑖 {𝑎𝑎𝑎 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝑤𝑤𝑤𝑤𝑤 𝑥 𝑖𝑖 ≠ 𝑛𝑛𝑛𝑛}
∑ 𝑤𝑖
Re-normalize to display full range of relative values
 Final displayed “overall water risk” (sj):
𝑎 𝑗 − min(𝑎)
𝑠𝑗 = 5
max 𝑎 − min(𝑎)

45. UNIQUE USERS : TAILORED WEIGHTING
WRI Default
Agriculture
Food & Beverage
Chemicals
Electric power
Semi-conductor
Oil & Gas
Mining
Construction Materials
Textile
0% 25% 50% 75% 100%
 Quantity  Quality  Regulatory and reputational

46. BALANCING INDICATORS : SETTING WEIGHTS
Exposure to risk varies– users can set own weights
Default weights set by:
 WRI water expert panel
 Corporate disclosure documentation
 Industry leader review

47. PUBLICATIONS
 Aqueduct Global Maps 2.0:
http://www.wri.org/publication/aqueduct-metadata-global
 Aqueduct Water Risk Framework:
http://www.wri.org/publication/aqueduct-water-risk-framework
 Aqueduct Data and Methodology: in prep

48. Thank you for joining!
CONTACT US
Paul Reig | preig@wri.org
Francis Gassert | fgassert@wri.org
WRI.org/Aqueduct