Using Imaging Spectrometry to Identify Crops in California’s Central Valley

Abstract:

With a growing global population, limited resources and a changing climate, understanding and monitoring the distribution of our food and water resources is essential to their sustainability. Regional food yield estimates and water resource accounting are dependent upon accurate agricultural records. Crop mapping provides farmers, managers, and policymakers the information necessary to anticipate annual food supplies and water demands by better understanding the distribution of species. While on the ground crop accounting usually happens yearly at the county level and requires significant time and labor inputs, remote sensing has the potential to map crops and monitor their health over a greater spatial area with more frequent time intervals. Specifically, imaging spectrometers have the capability to produce imagery at high spectral and spatial resolutions, which may allow for differentiation of crops at the field-level scale. In this research 14 crop species and soil were classified in Kern County, California using canonical discriminant analysis (CDA) and Multiple Endmember Spectral Mixture Analysis (MESMA) on airborne visible/infrared imaging spectrometer (AVIRIS) imagery from June 2013. Imagery was then degraded to Landsat spectral resolution and reclassified for comparison. Results with the AVIRIS imagery show an overall accuracy of 69.0% using MESMA and 89.4% using CDA with nine out of fourteen crop species showing user and producer errors under ten percent. Lower accuracy was found for OLI data. This research illustrates great potential for field-level crop mapping with imaging spectrometry.