Response-calibrated SWAT process information for daily total nitrogen prediction in a gate-controlled shallow lake

Daily total nitrogen (TN) prediction in gate-controlled shallow lakes is hindered by the mismatch between catchment nutrient export and section observations. Here, we introduce response calibration, a new section-scale concept that transforms non-point-source process information from the Soil and Water Assessment Tool (SWAT) into effective predictors by adapting it to transport lag and gate-regulated hydrodynamics. In Honghu Lake, the response-calibrated SWAT-ridge model achieved an RMSE of 0.14 mg/L and an NSE of 0.79 on the 50-day late-autumn/winter independent test set, reducing RMSE by 34.07% relative to conventional monitoring inputs. Raw SWAT outputs added directly as predictors provided no gain, with an RMSE of 0.21 mg/L, showing that their value arose from response calibration rather than from feature addition. The gains were largest during rainfall-affected periods and under low-exchange conditions, when runoff activation and lake retention amplify the source-to-section mismatch. In the 2024 forward evaluation, skill was lower in the highest-concentration quartile, where RMSE reached 0.30 mg/L, and in 2022, a year outside the calibrated regime. This framework provides a practical way to use insufficiently calibrated process-model outputs as section-oriented priors for short-term water quality prediction in hydraulically regulated lakes where inflow water quality is unmonitored.