Located in the Brazilian semiarid region, Castanh & atilde;o is the largest non-hydropower dam in Latin America, which is used for water supply, irrigation and fish farming. After 2010, this reservoir became the largest producer of tilapia in cage aquaculture in Brazil (up to about 3,000 ton/month). However, a severe drought in the region from 2012-2018 resulted in extremely low water storage (2%) and very high total phosphorus (TP) concentration in the reservoir (> 0.20 mg/L), causing massive mortalities of tilapia and significant increases in water treatment costs. This research combines field surveys, lab studies and mathematical modeling to develop a practical tool to model phosphorus dynamics in Castanh & atilde;o reservoir and estimate its carrying capacity for its multiple uses and highly variable hydrological conditions. The model considered complete-mixing under transient conditions, including three sources of TP: the watershed, sediments and fish cages, and two sinks: settling and withdrawal. The model was calibrated with primary data of hydrology and water quality, achieving good performance: NSE = 0.62 and PBIAS =-0.03%. The external load was dominant during the wet period, while the internal loads became more relevant during the dry period. Scenario analysis also suggested that for the current tilapia production (about 500 ton/month) and water storage for the study period (about 15%), only a 80% reduction in the total loads could bring TP to the mandatory level (0.03 mg/L). Under these conditions, the carrying capacity of the reservoir was 100 ton of TP per year. Another simulated scenarios, potential remediation measures and their implications on water resources management are also discussed.