Regionalization analysis of environmental drivers of CONUS grazing land biomass

Authors: CHANG, J - Former ARS Employee; Gao, Feng; Anderson, Martha; Cirone, Richard; ZHAO, H - Oak Ridge Institute For Science And Education (ORISE)

Submitted to: IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/6/2025
Publication Date: 5/9/2025
Citation: Chang, J., Gao, F.N., Anderson, M.C., Cirone, R.J., Zhao, H. 2025. Regionalization analysis of environmental drivers of CONUS grazing land biomass. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. 18:12634-12644. https://doi.org/10.1109/JSTARS.2025.3568771.
DOI: https://doi.org/10.1109/JSTARS.2025.3568771

Interpretive Summary: Effective management of grazing lands requires precise biomass assessment. In recent decades, advancements in remote sensing technology have provided cost-effective and efficient alternatives to traditional field-based methods for monitoring and estimating biomass in grazing lands. However, estimating herbaceous biomass remains challenging due to the complexity of environmental factors such as precipitation, elevation, land surface temperature, vegetation cover, and soil texture. This study aims to enhance the understanding of these environmental influences on grazing lands and to identify key factors for developing reliable herbaceous biomass estimation models across the Conterminous United States (CONUS). By utilizing Earth observation data from Google Earth Engine and machine learning (ML) techniques, this paper evaluates environmental factors and compares the performance of various ML models over CONUS. The findings offer valuable insights into the development of comprehensive biomass estimation models, which can improve grazing land management under diverse environmental conditions.

Technical Abstract: Grazing lands are crucial for livestock feed and carbon storage. Effectively managing these lands requires precise biomass assessment, which is challenging due to diverse environmental conditions. This study focuses on enhancing the analysis of grazing lands in the Conterminous United States (CONUS) by combining Earth observation data with machine learning techniques. Key environmental factors, including precipitation, elevation, land surface temperature, vegetation cover, and soil texture, were used for unsupervised clustering. Data from the National Land Cover Database, MODIS, SRTM, and GPM were processed using the Google Earth Engine platform. The correlation of each cluster’s environmental factors with reference biomass from the Rangeland Analysis Platform was examined. The study identified 18 distinct environmental clusters, highlighting the complex variability within grazing lands and underscoring the importance of considering multiple environmental factors for reliable biomass estimation. The Random Forest model outperformed others in biomass estimation, with a mean absolute error (MAE) of 372.4 to 373.5 lb/acre and a coefficient of determination (R²) of 0.65 to 0.66. This research can contribute to developing adaptive biomass estimation models across diverse grazing lands by applying complex advanced models to differently categorized grazing lands, thereby improving both time efficiency and accuracy.