TY  - JOUR
AU  - Banda, Fabian
AU  - Simukoko, Leonard
AU  - Kalumba, Mulenga
AU  - Kaoma, Mwansa
PY  - 2025
DA  - 2025/04/17
TI  - Utilizing Machine Learning to Map Aquatic Weed Yields for Biochar Production: A Case Study in the Bangweulu Wetlands, Zambia
JO  - Advances in Environmental and Engineering Research
SP  - 020
VL  - 06
IS  - 02
AB  - Aquatic weeds present significant ecological and socio-economic challenges in the Bangweulu Wetlands of northern Zambia, where their proliferation disrupts aquatic ecosystems, impedes fishing activities, and affects local livelihoods. Despite these challenges, aquatic weeds also offer a unique opportunity for sustainable biochar production, a clean alternative cooking fuel that can alleviate pressure on diminishing forest resources. This study explores the application of machine learning (ML) techniques to estimate and map the spatial distribution of aquatic weed biomass, thereby enabling more efficient and strategic harvesting for biochar production. The research objectives included field-based measurement of aquatic weed biomass, analysis of environmental covariates, evaluation of four machine learning models for yield prediction, and the generation of spatial yield distribution maps. Among the tested models, Gradient Boosted Regression Trees (BRT) demonstrated superior performance, achieving an R2 of 0.63, a Mean Absolute Error (MAE) of 0.08, and a Root Mean Square Error (RMSE) of 0.29. Key predictive variables included remote sensing-derived vegetation indices (LAI, EVI, NDVI), climate parameters, and topographic derivatives from Digital Elevation Models (DEMs). Seasonal biomass yield predictions ranged from 0.70 kg to 1.18 kg per square meter, highlighting significant spatial and temporal variability. The ML-driven yield maps enable precision harvesting, which can enhance operational efficiency, reduce labor and fuel costs, and minimize environmental disturbance. Moreover, by facilitating the conversion of invasive biomass into biochar, the approach contributes to circular economy principles, reduces greenhouse gas emissions associated with traditional biomass use, and supports energy access in underserved rural areas. Overall, the integration of ML-based yield estimation into biochar production planning represents a scalable and data-driven solution that bridges environmental restoration with sustainable energy generation. The study’s methodology and findings offer valuable insights for policymakers, conservationists, and clean energy advocates aiming to harness natural resources more responsibly.
SN  - 2766-6190
UR  - https://doi.org/10.21926/aeer.2502020
DO  - 10.21926/aeer.2502020
ID  - Banda2025
ER  -