Mapping changes in agricultural cropping frequency across Zimbabwe using cross-scale time-series remote sensing data and a novel signal decomposition method

Wednesday, 17 December 2014
Ankush Khandelwal1, Jamon Van Den Hoek2, Fernando Sedano3, Vipin Kumar1 and Compton J Tucker4, (1)University of Minnesota Twin Cities, Minneapolis, MN, United States, (2)NASA Goddard Space Flight Center, Greenbelt, MD, United States, (3)University of Maryland, Washington, DC, United States, (4)NASA Goddard Space Flight Cen., Greenbelt, MD, United States
A central challenge in agricultural remote sensing is the detection of changes in intra-annual cropping frequency, often necessary in monitoring crop productivity, agricultural management practices, or policy implementation. Though remote sensing imagery offers synoptic and systematic measurements relevant to monitoring crop phenology across spatial scales, broad-scale (i.e., country-wide) changes in cropping frequency have seldom been quantified due to spatio-temporal heterogeneity in phenological and climatic cycles, signal noise, and missing data resulting from cloud cover. For example, in Zimbabwe, once the breadbasket of southern Africa, large-scale changes to agricultural production followed land distribution policies introduced in the early 2000s. The diverse impacts of land reform on the agricultural economy continue to be debated yet the underlying changes in cropping frequency and pattern have never been systematically assessed. Using Zimbabwean agriculture as a case study and MODIS 16-day composite Normalized Difference Vegetation Index (NDVI) and complementary Landsat imagery collected since 2000 across Zimbabwe, this presentation introduces a novel time-series signal decomposition and spatiotemporal clustering approach to map intra-annual cropping frequency and changes therein. MODIS-derived results indicate a massive decline in double-cropped acreage since 2000, a complete overhaul of cropping pattern with the disaggregation of large-scale commercial farms into multiple smallholder plots, and a spatial contraction of double-cropped fields to peri-urban lands, while Landsat trends capture the recent emergence of small-scale double-cropping systems unseen in MODIS data. These findings provide an independent and objective assessment of field-level changes in agricultural productivity, spatiotemporally explicit land reform effects on large-scale as well as smallholder agriculture and potential for food production, and have importance for regional water consumption, fertilizer use, and agricultural carbon emissions.