Modelagem da qualidade de água na bacia hidrográfica do Rio Umbeluzi – Moçambique: integração dos modelos SWAT e QUAL-UFMG

Abstract

The increasing anthropogenic pressure on water resources and the scarcity of monitoring data hinder water quality management in river basins of developing countries. In this context, this dissertation aimed to assess water quality in the Umbeluzi River, Mozambique, through the integration of the SWAT (Soil and Water Assessment Tool) and QUAL-UFMG models, enabling the coupled simulation of hydrological processes and pollutant transport in a tropical, data-scarce basin. The adopted approach involved three main steps: (i) physiographic, morphometric, and hydrological characterization of the basin using geotechnologies and determination of the Water Quality Index; (ii) application of the SWAT model for the period 1994–2023; and (iii) analysis of water quality along 65 km of the Umbeluzi River main course, divided into three reaches with distinct monitoring points. The SWAT model showed good performance in streamflow simulation, with a Nash–Sutcliffe Efficiency (NSE) of 0.76, R2 of 0.78, and PBIAS of 13% during calibration, evidencing its reliability in generating monthly mean flows to support qualitative modeling. Conversely, the QUAL-UFMG model, applied to DO, BOD5, TN, TP, and thermotolerant coliforms (CTerm), presented mixed results: reasonable performance for dissolved oxygen (NSE = 0.20; R2 = 0.20; RMSE = 1.22 mg/L), but unsatisfactory results for the other variables, with negative NSE values (BOD5 = –3.71; TN = –2.34; TP = –1.71; CTerm = –12.73). These results reflect the limitations of the simulated scenario, which did not include point and diffuse pollution sources, nor the contributions and influence of tributaries, compromising the representativeness of observed peaks of organic and microbiological loads. The 65 km simulation highlighted a self-purification trend, especially for DO, but also revealed pollutant underestimation in areas with intense urban and agricultural activity. Despite these limitations, the SWAT–QUAL-UFMG integration proved to be a promising strategy for environmental assessment and water management planning in basins with limited monitoring infrastructure. It is concluded that, to improve the accuracy of future simulations, it is essential to incorporate realistic pollutant load data (point and diffuse), consider climatic seasonality, previously model tributaries, and refine kinetic parameters. The integrated approach adopted in this dissertation provides technical and scientific support to inform public policies for sanitation, environmental planning, and climate change adaptation in the Umbeluzi River basin.