Remote Sensing Analysis of Malawi's Agricultural Inputs Subsidy and Climate Variability Impacts on Productivity

Agriculture in sub-Saharan Africa is characterized by smallholder production and low yields ( 1 ton ha-1 year-1 since records began in 1961) for staple food crops such as maize (Zea mays). Many years of low-input farming have depleted much of the region's agricultural land of critical soil carbon and nitrogen, further reducing yield potentials. Malawi is a 98,000 km2 subtropical nation with a short rainy season from November to May, with most rainfall occurring between December and mid-April. This short growing season supports the cultivation of one primary crop, maize. In Malawi, many smallholder farmers face annual nutrient deficits as nutrients removed as grain harvest and residues are beyond replenishment levels. As a result, Malawi has had stagnant maize yields averaging 1.2 ton ha-1 year-1 for decades. After multiple years of drought and widespread hunger in the early 2000s, Malawi introduced an agricultural input support program (fertilizer and seed subsidy) in time for the 2006 harvest that was designed to restore soil nutrients, improve maize production, and decrease dependence on food aid. Malawi's subsidy program targets 50-67% of smallholder farmers who cultivate half a hectare or less, yet collectively supply 80% of the country's maize. The country has achieved significant increases in crop yields (now 2 tons/ha/year) and, as our analysis shows, benefited from a new resilience against drought. We utilized Landsat time series to determine cropland extent from 2000-present and identify areas of marginal and/or intermittent production. We found a strong latitudinal gradient of precipitation variability from north to south in CHIRPS data. We used the precipitation variability to normalize trends in a productivity proxy derived from MODIS EVI. After normalization of productivity to precipitation variability, we found significant productivity trends correlated to subsidy distribution. This work was conducted with Google's Earth Engine, a cloud-based platform for data storage and analysis with unprecedented speed and efficient computing by making use of Google's computing infrastructure.