Eyes in the Sky, Boots on the Ground : Assessing Satellite- and Ground-Based Approaches to Crop Yield Measurement and Analysis in Uganda

Lobell, David B. ; Azzari, George ; Marshall, Burke ; Gourlay, Sydney ; Jin, Zhenong ; Kilic, Talip ; Murray, Siobhan

World Bank, Washington, DC

关键词: MAIZE;    CROP YIELD;    CROP CUTTING;    REMOTE SENSING;    SMALLHOLDER FARMER;   

DOI  :  10.1596/1813-9450-8374
RP-ID  :  WPS8374

学科分类：社会科学、人文和艺术（综合）

来源: World Bank Open Knowledge Repository

【 摘 要 】

Crop yields in smallholder systems aretraditionally assessed using farmer-reported information insurveys, occasionally by crop cuts for a sub-section of afarmer's plot, and rarely using full-plot harvests.Accuracy and cost vary dramatically across methods. Inparallel, satellite data is improving in terms of spatial,temporal, and spectral resolution needed to discernperformance on smallholder plots. This study uses data froma survey experiment in Uganda, and evaluates the accuracy ofSentinel-2 imagery-based, remotely-sensed plot-level maizeyields with respect to ground-based measures relying onfarmer self-reporting, sub-plot crop cutting (CC), andfull-plot crop cutting (FP). Remotely-sensed yields includetwo versions calibrated to FP and CC yields (calibrated),and an alternative based on crop model simulations, using noground data (uncalibrated). On the ground, self-reportedyields explained less than 1 percent of FP (and CC) yieldvariability, and while the average difference between CC andFP yields was not significant, CC yields capturedone-quarter of FP yield variability. With satellite data,both calibrated and uncalibrated yields captured FP yieldvariability on pure stand plots similarly well, and bothcaptured half of FP yield variability on pure stand plotsabove 0.10 hectare. The uncalibrated yields wereconsistently 1 ton per hectare higher than FP or CC yields,and the satellite-based yields were less well correlatedwith the ground-based measures on intercropped plotscompared with pure stand ones. Importantly, regressionsusing CC, FP and remotely-sensed yields as dependentvariables all produced very similar coefficients for yieldresponse to production factors.