Analysis of Water Quality Time Series Obtained for Mass Discharge Estimation

Abstract

Methods are proposed for quantifying long term mean annual riverine load reductions of the nutrient phosphorus [P] and other agricultural pollutants anticipated in southwestern Ontario Great Lakes tributaries due to farm scale nonpoint source [NPS] remediation measures implemented in the headwater catchments. Riverine delivery of NPS pollutants is a stochastic process driven by episodic hydrometeorologic events; thus, progress towards tributary load reduction targets must be interpreted as the expected mean annual reduction achieved over a suitably long, representative hydrologic sequence. Trend assessment studies reveal that runoff event biased water quality monitoring records are conceptualized adequately by the additive model EquationSource% MathType!MTEF!2!1!+- % feaagCart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn % hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr % 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9 % vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x % fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamiwamaaBa % aaleaacaWGPbaabeaakiabg2da9iqadIfagaqeamaaBaaaleaacaWG % PbaabeaakiabgUcaRiaadoeadaWgaaWcbaGaamyAaaqabaGccqGHRa % WkcaWGtbWaaSbaaSqaaiaadMgaaeqaaOGaey4kaSIaamivamaaBaaa % leaacaWGPbaabeaakiabgUcaRiabew7aLnaaBaaaleaacaWGPbaabe % aaaaa!4742!]]</EquationSource><EquationSource Format="TEX"><![CDATA[$${X_i} = {\bar X_i} + {C_i} + {S_i} + {T_i} + {\varepsilon _i}$$ where X i is sample concentration, EquationSource% MathType!MTEF!2!1!+- % feaagCart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn % hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr % 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9 % vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x % fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGabmiwayaara % WaaSbaaSqaaiaadMgaaeqaaaaa!3803!]]</EquationSource><EquationSource Format="TEX"><![CDATA[$${\bar X_i}$$ is ‘global’ central tendency, C i is discharge effect, S i is seasonality, T i is trend (local central tendency) and ε i is residual noise. As the watersheds systematic hydrochemical response embodied in components C i and S i has remained stable in the presence of gradual concentration trends, the expected mean annual load reductions may be inferred by the difference between the mean annual loads estimated by adjusting the water quality series to (1) pre-remediation and (2) current mean concentration levels where concentrations on unsampled days are simulated by Monte Carlo methods. Fitting components by robust nonparametric smoothing filters in the context of generalized additive models, and jointly fitting interactive discharge and seasonal effects as a two dimensional field C⊗S t are considered.KeywordsWater QualityGreat LakeGeneralize Additive ModelAnnual LoadHeadwater CatchmentThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.