Abstract

High-resolution satellite imagery to improve crop statistics in data-poor environments
Track: Agriculture
Authors: An Notenbaert, Lieven Claessens, Percy Zorogastúa, Roberto Quiroz

Spatial crop statistics are an important source of information for implementation of research and development projects. However, available data for crops in subsistence agricultural settings rarely correspond with actual areas. We developed a methodology combining radiospectrometry and high resolution satellite images in a GIS. An application for sweetpotatoes in Uganda uses multispectral panchromatic images and handheld spectroradiometers for ground-truthing. Digital data processing was done with ENVI and ESRI ArcGIS. An unsupervised classification clustering the image in general landuse classes was followed by a supervised classification using spectral signatures from the spectroradiometer to obtain the spatial distribution of sweetpotato. The spectral signature of sweetpotato can be differentiated from other crops by low reflectance in the visual and high reflectance in the NIR spectrum. The analysis shows that official statistics are underestimating the area under sweetpotato by 50 to 60% and shows a promising methodology to improve crop statistics in data-poor environments.