Integration of In Situ Measurements and Multi-Sensor Remote Sensing for Above-Ground Biomass Estimation in High-Mountain Summer Pastures of Azerbaijan

Author : Karim Mirzayev

Abstract :Accurate estimation of above-ground biomass (AGB) in mountainous pastures is critical for sustainable rangeland management. This study develops and evaluates models to predict pasture biomass in the highlands of Azerbaijan using destructively sampled f ield plots (n ≈ 60) collected in September 2020 as ground truth. Remotely sensed features were extracted from temporally aligned Sentinel-1 SAR (γ⁰_VV_dB, γ⁰_VH_dB, VH–VV_dB, RVI) and Sentinel-2 optical imagery (NDVI, NDMI, MSAVI), together with topographic covariates (altitude, slope). Sentinel-1 Level-1 GRD scenes were manually processed in SNAP and Band Math for VH VV (dB) and RVI (computed in linear domain) calculated afterwords. Sentinel-2 Level-2A data were composited in Google Earth Engine using cloud masking and index derivation; all layers were coregistered at 10 m (UTM 38N). Plot-level predictors were sampled as 7 m-buffer medians to reduce geolocation and mixed-pixel effects. We compared ordinary least squares (OLS) regression and Random Forest (RF), with both raw features and principal-component (PCA) reductions of optical and SAR blocks. OLS achieved an apparent fit up to R² = 0.63, while RF attained cross-validated R² ≈ 0.30 with RMSE below the recommended 400 kg DM ha-¹ threshold. Feature importance analyses consistently highlighted altitude and optical indices (NDVI/NDMI/MSAVI) as dominant predictors, with SAR variables providing modest but complementary structural/moisture information, and PCA improving RF stability by reducing multicollinearity. The workflow demonstrates a reproducible, fine-scale (10 m) pathway for integrating in situ measurements with multi-sensor satellite data to map pasture biomass in cloud- and topography-challenged environments.

Keywords :Pasture Biomass Modeling in Azerbaijan Using Multi-Sensor Satellite Data

Conference Name :International Conference on Geosciences and Geographic Information Systems (ICGGIS-25)

Conference Place Kyoto, Japan

Conference Date 14th Nov 2025