Ground and Satellite Sensor Integration for Smallholder Agriculture Monitoring

Keywords: remote sensing, smallholder, maize, Zambia, data fusion
Abstract Type: Paper Abstract

Abstract

With food demand expected to significantly increase in the coming decades, increasing agricultural production through intensification (as opposed to extensification) is critical if sub-Saharan Africa is to become food self-sufficient. Satellite-based agricultural monitoring has improved dramatically in recent years, due to a combination of higher-resolution satellite sensors, new methods for detecting cropland and estimating crop productivity, and increased integration of data from sensors, crop surveys, field trials, and crop modeling. However, two obstacles limit agricultural monitoring. First, cloud cover can obscure satellite images for much of the growing season. Second, a lack of data on smallholder management practices limits the ability to evaluate soil, weather, and management interactions.

To address these obstacles, this study uses ground-mounted sensors, which are the only way to accurately and consistently monitor crop growth (using a multispectral sensor). This project uses networks of Arable Marks, a ground-based multispectral and meteorological sensor, installed in smallholder maize fields of Zambia and Kenya. The sensors provide consistent, continuous measurements of surface reflectance and derived vegetation indices (VI) like NDVI, albeit at a limited spatial footprint. To expand this spatial footprint, we have developed a model using smoothed time-series of multispectral and radar satellite imagery as inputs, and the ground-sensor based VI as an output. This presentation explains the advantages and challenges of this data integration technique, including curve-smoothing, model validation, and how to derive important phenological events, like start of season and peak greenness, from the model output.