Sustainable water stress management in arid and semi-arid regions requires precise understanding of soil-plant interactions
when implementing biochar-based strategies. This study developed an Artificial Neural Networks (ANN) model to
predict mango productivity under drought conditions using potassium-enriched biochar (KEB), addressing a significant
knowledge gap in biochar application modeling. KEB was made by pyrolyzing a maize straw–banana peel mix (1:3)
at 500 °C for 3 h. A two-year field experiment evaluated four potassium sources, i.e., C (control), KS (K₂SO₄), KEB
(potassium-enriched biochar), and KF (potassium feldspar), under two irrigation regimes representing 80% (normal) and
50% (drought) of available soil moisture. Potassium release patterns between KEB and KS (R2 = 0.83–0.97), both superior
to other treatments. Drought stress significantly impaired soil quality, reducing mango fruit yield by 20% and decreasing
soil microbial biomass carbon (MBC) and dehydrogenase enzyme activity by 22% and 14%, respectively. However, KEB
application enhanced soil quality under water stress by improving MBC, dehydrogenase enzyme activity, and soil organic
carbon (SOC), resulting in an 82% yield increase. KEB treatment also elevated chlorophyll content, proline levels, and
soluble carbohydrates, enhancing drought tolerance through improved osmotic adjustment in mango leaves. ANN modeling
identified optimal conditions for maximizing fruit yield, with 125% KEB application providing the best results. The
model established critical threshold values: SOC (3.1 g kg− 1), MBC (365.66 mg kg− 1), chlorophyll (3.7 mg kg− 1), soluble
carbohydrates (38.25 mg kg− 1), and phosphorus (2.78 mg kg− 1). These findings highlight the dual role of KEB as a sustainable
soil amendment and the utility of ANN as a decision-support tool for precision agriculture. Integrating KEB with
intelligent modeling approaches offers a promising strategy for improving resilience and productivity in mango orchards
under water-limited conditions.