Predicting nitrate discharge dynamics in mesoscale catchments using the lumped StreamGEM model and Bayesian parameter inference

摘要: Abstract The common practice of infrequent (e.g., monthly) stream water quality sampling for state the environment monitoring may, when combined with high resolution flow data, provide sufficient information to accurately characterise dominant nutrient transfer pathways and predict annual catchment yields. In proposed approach, we use spatially lumped model StreamGEM daily nitrate concentration (mg L −1 NO 3 -N) in four contrasting mesoscale headwater catchments based on years rainfall, potential evapotranspiration, measurements, monthly or concentrations. Posterior parameter distributions were estimated using Markov Chain Monte Carlo code DREAM ZS a log-likelihood function assuming heteroscedastic, t-distributed residuals. Despite uncertainty some parameters, calibration data was well reproduced across all (Nash-Sutcliffe efficiency against Log transformed NSL, range 0.62–0.83 0.17–0.88 concentration). slight increase size residuals separate validation period considered acceptable (NSL 0.60–0.89 0.10–0.74 concentration, excluding one set limited data). Proportions discharge attributed near-surface, fast seasonal groundwater slow deeper consistent expectations geology. results Weida Stream Thuringia, Germany, as opposed were, intents purposes, identical, suggesting that provides prediction dynamics. This study highlights remarkable effectiveness process based, modelling commonly available sample elucidate function, appropriate methods are used correctly handle inherent uncertainties.