RAINFALL AND AGRICULTURE IN CENTRAL WEST AFRICA: Predictability of Crop Yields in Burkina Faso* Pauline A. Dibi Kangah Rafael D. Flor* International Research Institute for Climate and Society Tropical Agriculture Program* Columbia University, New York - USA
OUTLINE Research Objective I - Rainfall in Central West Africa (CWA) Research Objective II & III - Agriculture & Rainfall vs. Yields Research Objective IV - Predictability of Crop Yields in Burkina Faso (BF) Conclusions
STUDY AREA AND MAIN GOALS Location of study area 24 Choice of Central West Africa (CWA) for study area (geographical 22 location, size, orientation) Algeria 20 Africa Mali 18 Rainfall is the major controlling Mauritania factor in tropical agric. 16 Agriculture is the prime economic parameter Latitude North 14 12 MAIN GOALSBurkina Faso Guinea Togo/Benin Assess rainfall 10 variability, agriculture, & their relationships in Central West Africa 8 Ghana Reveal the predictability Côte d'ivoire of crop yields in 6 Liberia Burkina Faso 4 Senegal Gulf of Guinea Niger -14-12 -10-8 -6-4 -2 0 2 4 Longitude West-East Countries Capital cities Other main cities
RESEARCH OBJECTIVE I To analyze rainfall characteristics and trends in Central West Africa Rainfall Variability from the Guinea Coast to the Sahel Multidecadal Isohyetal Trend Determination Average Annual Rainfall Regimes Mean Onset/Cessation Dates
Concluding Remarks for Objective I The region Number experienced of wet & dry years determines large scale dominant drought characteristic especially for each decade in (above & below mean) 1972, 1973, 1983, 1984, & 199010 12 10 1.58 1.5 6 Decadal scale rainfall: 1 Maliwet decades (1930s, 1950s, 6 1 Burkina Faso 4 1960s) 0.54 & dry decades (1940s, 0.51970s, 1980s, 1990s) 2 02 0-0.50-0.5 0 Strong -1 northward rainfall gradient -1 throughout CWA for -1.5-1.5 many study measures (annual rainfall index/totals, Entire Study Region 1.5 1.5 multidecadal isohyetal trends, monthly/daily rainfall 1930 1935 1940 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 10 Wet 1 Côte d'ivoire Cote d'ivoire Wet 1 Entire Study Region patterns, 8 number of rain days per year/season/month, & Dry Dry 0.5 0.5 onset/cessation 06 dates of the rainy 0 seasons) -0.5-0.5 4 Rainfall gradient explains the diversity of cropping and the -1-1 2-1.5-1.5 geographic locations of crop types -- Comprehensive 1930 Mali 1930-39 1935 1940 Normalized Rainfall Index Wet Time Burkina Series Faso (1930-98) 0 rainfall study is essential foundation for analyses of agriculture & rainfall/crop relationships 1930-39 1940-49 1945 1940-49 1950-59 1950 1955 1950-59 1960-69 1960 1965 1960-69 1970-79 1970 1975 1970-79 1980-89 1980 1985 1980-89 1990-98 1990 1995 1990-98 Dry 1930 1930 8 10 8 6 4 2 0 1930-39 1935 1935 1940 1940 1930-39 1940-49 1945 1945 1950 1950 1940-49 1950-59 1955 1955 1960 1960 1950-59 1960-69 1965 1965 1970 1970 1960-69 1970-79 1975 1975 1970-79 1980 1980 1980-89 1985 1985 1980-89 1990 1990 1990-98 1995 1995 1990-98 Wet Dry
RESEARCH OBJECTIVE II & III To detect & assess the extent to which crop acreage & production, and the impacts of rainfall variations are valuable indicators of agricultural change, including intercountry comparisons with common crops (cotton, maize, rice) as related to rainfall Average Spatial Distribution of Crops - Representative Regions & Crops -- Crop Parameter Relationships Correlations (1970-1998) b/w Rainfall & Cotton/Maize/ Rice Yields (1) Annual Rainfall Indices vs. Raw & Detrended Crop Time Series (concurrent & lag correlations) (2) Months/Seasonal/Annual Indices vs. Raw & Detrended Crop Time Series (concurrent correlations) (3) Crop regions & associated PCA-based regions: Months/ Seasons/Annual totals/onset-cess. vs. Raw Crop Time Series
Correlation Results between Rainfall & Cotton/Maize/Rice Yields (1) Mixed results - Evidence of positive relationships (2) Coefficients = more + when linear crop trends removed (3) More + coefficients for concurrent than lag correlations (4) Mali = strongly positive esp. Sep/May-Oct/Annual (5) BF/CI/CWA/zones = complex (weaker + & more -) (6) Mali = short & average to long rainy seasons (7) BF & CI = average to long rainy seasons
Concluding Remarks for Objective II & III Analyses identified 12 dominant regions & 8 major crops: Dominant Crop Regions = 12 (4 per country) Mali/Burkina = Koulikoro, Faso Kayes, (cereals Sikasso, Segou &- Burkina staple Faso = crops) Center, East, -Haut Côte Bassin, d Ivoire Mouhoun - Côte d Ivoire = Center, Central-west, South, West (starch/root plants & cash crops) Relationships b/w crop parameters reveal that Acreage T and production are more strongly related than either acreage and yield or production and yield G Mb Crop yields show strong variations Su Sa that are related to Ky Ko Na CNa rainfall variations -- Interpretations Ma SUMMARY OF ACREAGE (A), PRODUCTION of the time series & Ca CWa (P), AND E YIELD (Y) RELATIONSHIPS CEa So correlation Trend Relation analyses suggest Mali HBthat SWa Burkina crop Faso yields Côte d Ivoire might also Total Nb N NW be related to other factors (i.e., NEpolicies, global climate CNb and environmental predictors) W CWb Cb CEb and not solely on rainfall, SWb Sb 100 0 100 200 Kilometers although lack of water is the primary constraint to crop Total 16 16 16 48 growth, especially in drought-prone areas APY = + (27/48 = 56%) 9 (5C, 4S) 11 (6C, 5S) 7 (2C, 5S) 27 (13C, 14S) APY = - (1/48 = 2%) 0 (0C, 0S) 0 (0C, 0S) 1 (1C, 0S) 1 (1C, 0S) AP = +, Y = - (14/48 = 29%) 4 (2C, 2S) 3 (2C, 1S) 7 (4C, 3S) 14 (8C, 6S) AY = +, P = - (0/48 = 0%) 0 (0C, 0S) 0 (0C, 0S) 0 (0C, 0S) 0 (0C, 0S) A = -, PY = + (6/48 = 13%) 3 (1C, 2S) 2 (0C, 2S) 1 (1C, 0S) 6 (2C, 4S) Note: (+) = same ; (-) = different -- regression-based trends for overall (sub) period
RESEARCH OBJECTIVE IV To reveal the relationships b/w crop yields and climate/environmental predictors in Burkina Faso DATA & METHODS 10 agricultural regions & 10 rainfall stations with 3 levels of information: observed monthly data and calculated April-October and July-September seasons 6 crops: Cotton, Maize, Millet, Peanut, Rice, & Sorghum 6 global predictors: CMAP (estimated rainfall data), SSTs (Global, Nino3, Atlantic Northwest, Atlantic Equatorial South),& NDVIg (Vegetation Index -- to establish environment vs. climate relations) Agriculture in BF is mostly rain-fed, therefore there is a need for a better understanding of the relationships among crop, climate & environmental data - BF crop yield residuals are correlated with individual climate/environmental predictors (1984-2003)
0.7 350 MAJOR FINDINGS 300 Regional 29 rainfall HB regime CW SWin BF M is the same for the whole C CE CN CW 250 CN CE country, 0.6 with the bulk of rain in JAS -- CMAP rainfall 27 E HB M N 200 estimates are very similar to station observed rainfall 0.5 S SW 150 25 within each agricultural region Tempetures NDVIg (C) 0.4 23 Observed Rainfall (mm) 100 S N C E 50 NDVI, with a month of difference, follows the same monthly 0.3 21 0 distribution of Janrainfall, Feb Mar with Apr May highest Jun Julaccumulation Aug Sep Oct Novof Dec biomass 300 C CE CN Months occurring in the month of September 0.2 19 250 SSTAltNW follows the opposite pattern than to the one 200 0.1 17 SSTg observed in SSTg, Nino3, and SSTAtlEqS 15 0.0 CMAP Rainfall (mm) 150 CW E HB M N S SW Cash crops Jan Feb & water Mar stress Apr Maysensitive Jun Jul crops Aug (e.g., Sep Oct cotton, Nov Dec Jan100 Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec maize) are mostly cultivated in Month regions with higher rainfall & 50 (SW region) -- Subsistence crops (e.g., millet, sorghum) 0 are mostly cultivated Jan Feb Mar in Apr regions May Jun with Jul less Aug Sep rainfall Oct Nov (S Dec region) Months AtlNW Nino3 AtlEqS
MAJOR FINDINGS (Cont ) Monthly analysis: Results were sporadic (some regions/ some months no definite patterns) Seasonal analysis Cotton & Rice: The main region of HB = some relation with CMAP & NDVIg early in the season (planting time May/Jun) & nothing during the critical stages (e.g., flowering). Maize & Peanuts: 5 main regions (M, HB, SW, CW, E) Except for M and SW, the other regions have expected (-) correlation coefficients with SSTs & (+) correlation with CMAP & NDVIg. Millet & Sorghum: Expected signs early (AMJ) & during (JAS) the season especially for the northern drier agricultural regions with CMAP/ ORR (+) & Nino3 (-) Millet, sorghum, & peanuts = most cultivated crops in
OVERALL CONCLUSIONS This is the first attempt to compile & analyze disparate records (especially CWA crop data & CMAP/NDVIg records for BF agricultural regions) Rainfall variability/crop yield relationships are important to monitor & understand CWA socioeconomic development Time series & correlation analyses reveal that variations of crop yields mostly coincide with those of rainfall in CWA The findings of the rainfall/crop yield relationships in CWA provide important guidance for the study of the predictability of crop yields in Burkina Faso Lack of proper knowledge of the associations between climate/ environmental predictors & crop yields undermine the potential value of forecasts that can support end user decisions related to crop production & food security in the region
ACKNOWLEDGEMENTS CIMMS (Univ. of Oklahoma) for both the financial support and the data collection grant National meteorological, agricultural, and environmental institutions in Burkina Faso, Côte d Ivoire, Mali, and Niger for providing the research data Tropical Agriculture Program at the Earth Institute of Columbia University Thank you all for coming & listening to this talk