Sydney Gourlay

Eyes in the sky, boots on the ground : assessing satellite- and ground-based approaches to crop yield measurement and analysis in Uganda

Crop yields in smallholder systems are traditionally assessed using farmer-reported information in surveys, occasionally by crop cuts for a sub-section of a farmers plot, and rarely using full-plot harvests. Accuracy and cost vary dramatically across methods. In parallel, satellite data is improving in terms of spatial, temporal, and spectral resolution needed to discern performance on smallholder plots. This study uses data from a survey experiment in Uganda, and evaluates the accuracy of Sentinel-2 imagery-based, remotely-sensed plot-level maize yields with respect to ground-based measures relying on farmer self-reporting, sub-plot crop cutting (CC), and full-plot crop cutting (FP). Remotely-sensed yields include two versions calibrated to FP and CC yields (calibrated), and an alternative based on crop model simulations, using no ground data (uncalibrated). On the ground, self-reported yields explained less than 1 percent of FP (and CC) yield variability, and while the average difference between CC and FP yields was not significant, CC yields captured one-quarter of FP yield variability. With satellite data, both calibrated and uncalibrated yields captured FP yield variability on pure stand plots similarly well, and both captured half of FP yield variability on pure stand plots above 0.10 hectare. The uncalibrated yields were consistently 1 ton per hectare higher than FP or CC yields, and the satellite-based yields were less well correlated with the ground-based measures on intercropped plots compared with pure stand ones. Importantly, regressions using CC, FP and remotely-sensed yields as dependent variables all produced very similar coefficients for yield response to production factors.

Collecting the dirt on soils : advancements in plot-level soil testing and implications for agricultural statistics

Much of the current analysis on agricultural productivity is hampered by the lack of consistent, high quality data on soil health and how it is changing under past and current management. Historically, plot-level statistics derived from household surveys have relied on subjective farmer assessments of soil quality or, more recently, publicly available geospatial data. The Living Standards Measurement Study of the World Bank implemented a methodological study in Ethiopia, which resulted in an unprecedented data set encompassing a series of subjective indicators of soil quality as well as spectral soil analysis results on plot-specific soil samples for 1,677 households. The goals of the study, which was completed in partnership with the World Agroforestry Centre and the Central Statistical Agency of Ethiopia, were twofold: (1) evaluate the feasibility of integrating a soil survey into household socioeconomic data collection operations, and (2) evaluate local knowledge of farmers in assessing their soil quality. Although a costlier method than subjective assessment, the integration of spectral soil analysis in household surveys has potential for scale-up. In this study, the first large scale study of its kind, enumerators spent approximately 40 minutes per plot collecting soil samples, not a particularly prohibitive figure given the proper timeline and budget. The correlation between subjective indicators of soil quality and key soil properties, such as organic carbon, is weak at best. Evidence suggests that farmers are better able to distinguish between soil qualities in areas with greater variation in soil properties. Descriptive analysis shows that geospatial data, while positively correlated with laboratory results and offering significant improvements over subject assessment, fail to capture the level of variation observed on the ground. The results of this study give promise that soil spectroscopy could be introduced into household panel surveys in smallholder agricultural contexts, such as Ethiopia, as a rapid and cost-effective soil analysis technique with valuable outcomes. Reductions in uncertainties in assessing soil quality and, hence, improvements in smallholder agricultural statistics, enable better decision-making.

Land measurement bias and its empirical implications : evidence from a validation exercise

We investigate how land size measurements vary across three land measurement methods (farmer estimated, GPS, and compass-and-rope) and the effect of land measurement error on the inverse farm size relationship and input demand functions. Our findings indicate that self-reported measurement bias leads to overreporting for small plots and underreporting for large plots. The error is nonlinear, is not resolved by trimming of outliers, and results in biased estimates of the inverse land size relationship. Input demand functions that rely on self-reported land measures underestimate the effect of land on input utilization, including fertilizer and household labor.