Evolutionary algorithms for predicting aboveground carbon stocks in mopane woodlands in Mozambique
Tropical forests are crucial for global climate regulation and carbon cycling. Mopane woodlands, a tropical dry forest covering southern Africa, feature high ecological-socioeconomic importance. In Mozambique, charcoal production is a major driver of Mopane degradation and aboveground carbon (AGC) l...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Taylor & Francis Group
2025-12-01
|
| Series: | Carbon Management |
| Subjects: | |
| Online Access: | https://www.tandfonline.com/doi/10.1080/17583004.2025.2504937 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Tropical forests are crucial for global climate regulation and carbon cycling. Mopane woodlands, a tropical dry forest covering southern Africa, feature high ecological-socioeconomic importance. In Mozambique, charcoal production is a major driver of Mopane degradation and aboveground carbon (AGC) loss. Accurate AGC estimation is essential for climate mitigation strategies. We applied machine learning techniques to predict stand-level AGC in Mopane woodlands across Mabalane and Chicualacuala districts, Gaza Province. Two evolutionary algorithms were tested: (1) a hybrid Genetic Algorithm and Random Forest (GARF), and (2) Genetic Programming (GP) using symbolic regression. In total, 139 predictor variables were derived from remote sensing, biophysical, and bioclimatic datasets. Field data included 114 cluster plots. Both algorithms reduced the dataset by 95.6%. Observed AGC ranged from 1.313 to 28.476 MgC ha−1. GARF predictions ranged from 2.910 to 19.459 MgC ha−1 (nRMSE = 0.427; MBE = 0.08), while GP showed a wider predictive range (1.721–23.503 MgC ha−1; nRMSE = 0.428; MBE = 2.731×10−17). GARF relied on optical and bioclimatic variables, whereas GP operated independently of variable type. Both approaches were effective for feature selection and AGC prediction. However, GP produced a more interpretable model, offering advantages for replication and use in operational carbon inventories. |
|---|---|
| ISSN: | 1758-3004 1758-3012 |