Discriminating pasture degradation in Brazil: insights from EnMAP-Based spectral resolution simulations with emphasis on Landsat Next

We evaluated the potential of Landsat Next’s planned band configuration to discriminate among five pasture degradation classes in central Brazil. Hyperspectral data from the Environmental Mapping and Analysis Program (EnMAP) were used to simulate the spectral resolution of the innovative Landsat Next mission, planned for 2031, specifically 18 of its 21 reflective bands. Other past and current multispectral missions were also considered as part of a strategy to represent a progressive increase in the number of spectral bands and coverage across the 400–2500 nm range. Classification with the Light Gradient Boosting Machine (LightGBM) and band reflectance focused on Non-degraded pastures (NDP) and four degradation levels: Low-intensity (LID), Moderate-intensity (MID), Severe agronomic (SAD), and Severe biological (SBD). The results showed the importance of the Landsat Next red-edge and shortwave infrared (SWIR) multispectral bands, particularly those near 740 nm, 1600 nm and beyond 2000 nm, in distinguishing different levels of pasture degradation. Overall classification accuracy ranged from 0.63 for sensors limited to the visible and near-infrared (VNIR) to 0.74 for instruments covering both VNIR and SWIR regions. All simulations achieved F1-scores above 0.86 for NDP. For SAD, the highest F1-score was observed with Landsat Next (0.68). For SBD, only Landsat Next achieved good classification performance (F1-score = 0.60), attributable to its three additional narrow bands centred at 2038 nm, 2108 nm, and 2211 nm. These bands captured signs of clay mineral absorption bands linked to increased soil exposure caused by vegetation depletion through biological processes.