AAGW3 - Zhe Guo - SPAM Spatial production allocation model

Transcript

1. SPAM spatial production allocation model HarvestChoice / IFPRI Carlo Azzarri Melanie Bacou Mari Comanescu Cindy Cox Zhe Guo Stephen Kibet Jawoo Koo Cecile Martignac Ivy Romero Naomi Sakana Ulrike Wood-Sichra Zhe Guo on behalf of Ulrike Wood-Sichra Africa Agriculture GIS Week 2013 (AAGW3)
2. SPAM Team (*) current Ayumi Arai Melanie Bacou * Surajit Baruah Jordan Chamberlin Cindy Cox * Susana Crespo Maria Comanescu * Günther Fischer Steffen Fritz Zhe Guo * Daniel Hawes Mario Herrero Robert Hijmans Dave Hodson Glenn Hyman Stephen Kibet * Jawoo Koo * Chris Legg Andy Jarvis Nanga Kaye Amy Kapp Zahia Khan Gucheng Li Gerald Nelson An Notenbaet Navin Ramankutty Ingrid Rhinehart Ricky Roberson Ivy Romero * Cynthia Rossi Kate Sebastian * Dongsheng Sun M.T. Tenorio Philip Thornton H.T. van Velthuizen Ivan Vidangos Williams Ojo Stanley Wood Ulrike Wood-Sichra * Wenbin Wu Hua Xie Liangxi You *
3. Knowing exactly where and how a specific crop is grown is critical information to help scope research projects, set baselines, assess technology impacts, and make investment decisions. Why?
4. Not really. * Satellite imagery doesn’t reliably identify cropland nor easily distinguish crop species. * Statistics data are often too aggregated. * Global datasets come with many caveats. Don’t we already know?
5. Not quite. * Agro-ecological zones, mega- environments, watersheds, and marketsheds do not necessarily follow administrative boundaries. * Statistical data is often still too aggregated Aren’t (sub-)national statistics enough? How much maize is produced in rural areas in tropical Africa?
6. * Plausible best-estimates * Utilizes best-available sub-national statistics circa 2000 * Global 10 km grids * 4 variables x 20 crops x 4 production systems * Peer-Review & Community-driven improvements. Spatial Production Allocation Model
7. Protected Areas
8. Data Path From Admin Units to Pixel
9. 0.01 - 0.02 0.03 - 0.04 0.05 - 0.06 0.07 - 0.08 0.09 - 0.10 0.11 - 0.12 0.13 - 0.14 0.15 - 0.16 0.17 - 0.18 0.19 - 0.20 0.21 - 0.22 0.23 - 0.24 0.25 - 0.26 0.27 - 0.28 0.29 - 0.30 0.31 - 0.32 0.33 - 0.34 0.35 - 0.36 0.37 - 0.38 0.39 - 0.40 0.41 - 0.42 0.43 - 0.44 0.45 - 0.46 0.47 - 0.48 0.49 - 0.50 0.51 - 0.52 0.53 - 0.54 0.55 - 0.56 0.57 - 0.58 0.59 - 0.60 0.61 - 0.62 0.63 - 0.64 0.65 - 0.66 0.67 - 0.68 0.69 - 0.70 0.71 - 0.72 0.73 - 0.74 0.75 - 0.76 0.77 - 0.78 0.79 - 0.80 0.81 - 0.82 0.83 - 0.84 0.85 - 0.86 0.87 - 0.88 0.89 - 0.90 0.91 - 0.92 0.93 - 0.94 0.95 - 1.00 input
10. Harvested Area (ha/cell) < 250 251 - 500 501 - 750 751 - 1,000 1,001 - 1,250 1,251 - 1,500 1,501 - 1,750 1,751 - 2,000 2,001 - 3,000 3,001 - 4,000 4,001 - 5,000 5,001 - 6,000 6,001 - 7,000 7,001 - 8,000 8,001 - 9,000 9,001 - 10,000 10,001 - 11,000 > 11,000 output
11. Starting with SPAM 2000 (20 crops) IMPLEMENT SPAM 2005 (42 crops) and Verify, Validate, Complete, Re-run with user input Current status
12. # crops: ag_land crop suitabilities: irrigation, prices, rural population  SPAM 2000 [ 2000 ] [ 2005 ] 42: 20 +pulses, +oils, +cash, +fruit, +vegetables sub-national crop statistics Steffen Fritz et. al. (Geo-Wiki) GAEZ2010 (IIASA/FAO) same sources (Frankfurt, FAO, CIESIN) new version SPAM 2005 20 Ramankutty GAEZ2000 UWS
13. SPAM2005 Crop List 1 wheat 14 dry beans 28 sugarcane 2 rice 15 chickpea 29 sugar beet 3 maize 16 cowpea 4 barley 17 pigeon pea 30 cotton 5 pearl millet 18 lentils 31 other fibres 6 finger/small millets 19 other pulses 32 arabica coffee 7 sorghum 33 robusta coffee 8 other cereals 20 soybeans 34 cocoa 21 groundnuts 35 tea 9 potato 22 coconuts 36 tobacco 10 sweet potato 23 oil palm 37 banana 11 yam 24 sunflower 38 plantains 12 cassava 25 rapeseed 39 tropical fruit 13 other roots & tubers 26 sesame seed 40 temperate fruit 27 other oil crops 41 vegetables 42 rest UWS
14. crop statistics – area & yield national, sub-national, sub-sub national per crop not same for all crops in 1 country admin boundaries map GAUL 2008 (GADM ??) cropping intensity national (sub-national) per crop production systems share national (sub-national) per crop ag_land raster, 5 x 5 deg-min (approx 10 x 10 Km) crop suitabilities – area & yield raster, 5 x 5 deg-min (approx 10 x 10 Km) per crop irrigation raster, 5 x 5 deg-min (approx 10 x 10 Km) rural population raster, 5 x 5 deg-min (approx 10 x 10 Km) prices international $ (PPP) per crop Input data and scales UWS
15. physical area raster, 5 x 5 min degree (approx 10 x 10 Km) per crop and production system harvested area same as above yield same as above production same as above harvested area, yield, production national, sub-national, sub-sub national per crop and production system Output data and scales UWS
16. Challenges in preparation of Sub-National Production Data Reconcile inconsistent formatting and units Discover implausible yields resulting from erroneous data Resolve naming discrepancies between incoming statistics and geographic admin names
17. All ~ 2000, 5min grids 175 Crops (including forage crops and “grasses other”) 26 irrigated + 26 rainfed crops (including forages, and “irrigated managed grassland/ pasture”) 20 Crops M3 MIRCA SPAM Crop System Disaggregation None Irrigated , Rainfed Irrigated, Rainfed (Commercial), Rainfed (Subsistence), None Production Indicators Harvested area, Yield Harvested area Harvested area, (Physical area,) Yield Seasonality Annual Monthly Annual Global Data Products 175 x 1 x 2 x 1 = 350 26 x 2 x 1 x 12 = 624 20 x 4 x 2 x 1 = 160 Total crop & managed grassland area Managed grassland/pasture Rainfed grassland/pasture “Other crops” “Other annual crops” “Other perennial crops”
18. Primary Data & Methodological Steps Feature M3 MIRCA SPAM Statistical Reporting/ Inventory Units (SRU) 22,106 (56-2,299-19,751)* 402 (calendar units) ~24,000 (?-2,758-21,498)* Cropland extent/ Area intensity sources SAGE/Ramankutty SAGE/Ramankutty GMIA (Siebert) SAGE/Ramankutty Irrigated extent/ Area intensity source n/a GMIA (Siebert) (and M3 crop distribution) GMIA (Siebert) Cropping calendars/ Cropping intensity n/a Assembled by MIRCA team Assembled by SPAM team (for all systems) Crop-specific area downscaling Crop-specific share of pixel cropland area = crop area share of total SRU cropland In proportion to share of irrig./rainfed crop area in calendar unit (cu) and as f (cropping intensity by cu, and M3 total crop area allocation by pixel) Crop systems-specific share of pixel cropland area = f (system share of SRU crop production, pixel system-specific area and yield potential, SRU crop value/ha | all SRU pixel values) Allocation algorithm Crop by crop Crop by crop All crops simultaneously Yield downscaling Reported SRU crop yield assigned uniformly to all downscaled crop area in SRU n/a Yields vary by system and by pixel in proportion to potential system yield, scaled and area weighted to derive reported SRU yield * SRUs: C-1st-2nd Country level data only - 1st level admin units – 2nd level admin units More data, more assumptions, more modeling
19. I thought you’d never ask How can you help?
20. Production Systems Cropping Intensity CSI Collaborators’ Support Production Statistics
21. mapspam.info http://harvestchoice.org/mappr
22. Zhe Guo on behalf of Carlo Azzarri Contact: C.Azzarri@cgiar.org
23. Background  IFPRI with CIAT were using constant ratios by administrative unit to infer poverty rates from national poverty lines to PPP lines
24. How we were doing Green maps = we had the data Gray maps = we wanted to estimate $1 per day Poverty line Headcount (%) National Poverty line Headcount (%) National poverty Line headcount % $1 per day Poverty line Headcount (%) RATIO Apply RATIO above by district
25. 10 15 17 11 19 Number of poor (HC_NAL….77 total number of poor in country, according to map and GPW3_2005) % of poor (no. of poor in district divided by total number of poor in country) .13 .14 .19 .22 .25 Number of poor at $1.25 poverty Line according to POVCALNET 99 * % of poor (no. of poor in district divided by total number of poor in country) .13 .14 .19 .22 .25 Estimated $1.25 poverty line = 77, no. of poor at national poverty line 13 19 14 22 25 ?
26. ISO3_CODE Country name Survey used # sub-national units PHC $1.25 PHC $2 GINI BFA BURKINA FASO EICM, 2003 13 63.02 82.69 45.88 BDI BURUNDI CWIQ, 2006 17 79.67 92.83 33.27 CMR CAMEROON EICM, 2007 12 6.76 27.05 38.96 CIV COTE D’IVOIRE ENV, 2002-3 11 22.81 47.8 48.28 COD D.R.CONGO 1-2-3, 2004-5 11 57.06* 78.99* 42.15* ETH ETHIOPIA HICE, 2004-5 10 37.07 77.32 29.76 GMB GAMBIA PS, 1998 7 68.57 83.84 49.18 GHA GHANA GLSS, 2005 10 29.49 54.29 42.76 KEN KENYA IHBS, 2005 69 18.43 41.56 47.62 LSO LESOTHO HBS, 2002-3 10 46.10 64.76 52.44 MDG MADAGASCAR EPM, 2005 22 68.91 89.08 35.87 MWI MALAWI IHS, 2004-5 26 73.99 89.61 39.00 MLI MALI HIS, 2006 9 51.91 75.53 39.11 MRT MAURITANIA EPCV, 2000 13 22.69 47.03 39.02 MOZ MOZAMBIQUE IAF, 2002-3 11 75.86 89.03 47.74 NER NIGER ENBC, 2007 8 38.67 72.81 34.00 NGA NIGERIA NLSS, 2003-4 37 66.23 84.95 43.67 RWA RWANDA EICV, 2005-6 12 69.49 84.39 52.04 SEN SENEGAL ESAM, 2001 10 44.01 70.68 41.31 SSD SOUTH SUDAN NBHS, 2009 10 48.05 68.92 45.54 SWZ SWAZILAND HIES, 1995-6 6 79.85 88.78 61.99 TZA TANZANIA HBS, 2007 21 66.14 86.29 37.82 UGA UGANDA UNHS, 2005 4 51.06 75.41 40.83 ZMB ZAMBIA LMCS, 2006 9 69.61 82.43 57.70
29. …by AEZ …by farming system
30. Rasterization method 1. Extracted sub-national population data, separately for rural and urban population from HHs. For each country: = ,, =1 + ,, =1 t survey year, j sub-national administrative unit, n number of sub-national units per country 2. Multiplied the 5 arc-minute grid cell resolution CIESIN/GRUMP 2000 population density data (cellrate00,r , cellrate00,u ) by the proportion of population in each grid cell c (POPratec ), and by the sub- national population data (POPj,r,t and POPj,u,t ) resulting in population values per grid cell at the survey year (POPc,t ) for both rural and urban population: ,, = ∗ 00, ∗ ,, ,, = ∗ 00, ∗ ,, 3. Re-calculated each grid cell as in step 2 by multiplying the proportion of population in each grid cell by WDI 2005 rural and urban population data (WDI2005,r , WDI2005,u ): ,2005, = ,, ∗ 2005, ,2005, = ,, ∗ 2005, Total population per grid cell in 2005 is the sum of corresponding rural and urban population: = ,2005, + ,2005,