Diagnosis of degraded pastures using an improved NDVI-based remote sensing approach: An application to the Environmental Protection Area of Uberaba River Basin (Minas Gerais, Brazil)

Abstract

Pasture degradation represents a global environmental problem that urges mitigation. A fundamental step towards restoration of degraded pastures is the identification and accurate mapping of these areas. In Brazil, the area of degraded pastures is immense and therefore remote sensing is a cost-effective way to map it. In this study, an improved method based on NDVI values extracted from satellite images is presented, and tested in the Environmental Protection Area of Uberaba River Basin (EPAURB) located in the state of Minas Gerais, Brazil. The EPAURB covers an area of approximately 528.1 km 2 , 50.9% of which is pasture. The innovative features of this method comprise: 1) the mapping is preceded by the definition of NDVI fingerprints for healthy, smoothly degraded, moderately degraded and degraded pasture (called physiognomies), based on non linear relationships between NDVI values and time; 2) the mapping of physiognomies accounts for the influence of geology and weather seasonality on the NDVI values. In the EPAURB the physiognomic categories were set by visual inspection and evaluation of soil characteristics (e.g., organic matter, nutrients, resistance to penetration) in the so-called characterization ground truth sites also termed buffers. Resistance to penetration and several other soil parameters showed statistically different (p ≤ 0.05) values among physiognomies. The definition of fingerprints was based on a 4-year record (2013–2016) of NDVI 16-day composite (MOD13Q1) 250 m time-series data. The map of degraded pastures was delineated on the basis of comparisons between the NDVI values of 23 satellite images covering the year of 2016 (termed NDVI pixel ) and corresponding characteristic NDVI values of degraded pasture physiognomy extracted from the corresponding fingerprint (termed NDVI buffer ). Whenever NDVI buffer,min ≤ NDVI pixel ≤ NDVI buffer,max a repetition counter (n) increased one unit. For n ≥ 3 the pixel was classified as degraded pasture. The results exposed 160.1 km 2 of degraded pasture for 3 ≤ n ≤ 18, which represents 60% of all pasture land. The areas mapped as degraded pasture were subject to a field check in 38 so-called validation ground truth sites, using resistance to penetration as validation parameter, with 84.1% success. Given the serious environmental damage posed by pasture degradation, several mitigation measures were discussed including the protection of degraded soil through the “polluter pays principle”.