Riverine Nutrients Export From The Jiulong River:Hydrolgoical Controls And Application Of LOADEST Model

Affected by high-intensity anthropogenic activities and climate change,the nutrient?nitrogen and phosphorus?concentrations and loads export from subtropical rivers to the estuaries have become highly dynamic.It is hard to get accurate riverine nitrogen and phosphorus loads derived from the traditional monitoring data of water environment because of their low monitoring frequency and insufficient parameters.In the present study,we had a daily observation of nutrients at the outlet of Jiulong River in Fujian Province in the period of 2014-2016.According to the features of precipitation and discharge,the year of 2014 was a normal hydrological year,2015 was a mildly wet year,and 2016 was an extremely wet year influenced by ENSO.We first examined the nutrient transportation within river system and its hydrological mechanism by analyzing the variations of the observed nutrient concentration and load.Then,we developed modified LOADEST models for the Jiulong River to compare its model performances for modeling ammonium?NH₄-N?,nitrate?NO₃-N?and soluble reactive phosphorus?SRP?loads.The main findings and conclusions are summarized as follows.Affected by the subtropical monsoon climate,hydrology controls the timing and magnitude of the exported nutrient concentration and load from the river.The average annual concentration of nutrient presented a basic pattern "the extremely wet year>the mildly wet year>the normal hydrologic year".In the normal hydrological year?2014?,nutrients were diluted in the wet season and enriched in the dry season.Nutrient concentrations were high in the pre-wet season and peaked at the March in 2014.NH₄-N transported by surface runoff reached its concentration peak 1 week earlier than NO₃-N,which was originated from subsurface runoff.Affected by ENSO,the extremely wet year?2016?showed stronger dilution effect on nutrient dynamics because of the increasing intensity and frequencies of rainstorms.Storm-induced nutrient variation in the river channel was controlled by the superposed influence of source supply?enrichment effect?and rainfall runoff?dilution effect?.In 2014,the wet season?May to September?contributed 59%-73%of nutrient loads?NH₄-N,NO₃-N,and SRP?.In 2015,the wet season?May to October?contributed 73%-83%of nutrient loads.From 2014 to 2016,a total of 7 to 10 floods lasted 24%to 53%of each year but contributed 40%to 68%of the total dissolved nitrogen?DTN?loads and 48%to 73%of the total dissolved phosphorus?DTP?loads.Performance of LOADET model was influenced by the time scale of the calibration dataset,the ratio of maximum to minimum nutrient concentrations?MMR?,and the sampling frequency.What's more,different nutrient forms showed different model performances.Compared with the multi-year model?2014?2016?and the year-round model,the LOADEST piecewise model performed better in estimating nutrient loads.The "Piecewise model" was developed respectively for three periods?pre-wet season,wet season and post-wet season?.The deviation of estimated monthly nutrient loads in the extremely wet year?2016?became larger compared with that in the normal hydrological year?2014?.The maximal relative error?RE?of nutrient loads at different time scales estimated by the piecewise model presented as "nutrient loads in flood events>monthly loads>annual loads".The model behavior was closely related to the dynamic of nutrient concentration during flood events.There was a significant positive correlation between the MMR and the range of relative error?R²>0.75,p<0.01?.The deviation of annual and monthly loads of NO₃-N estimated by the piecewise model was smaller than that of NH₄-N and SRP.In the normal hydrological year,estimated NH₄-N loads?RE<10%?had smaller deviations from observed.values than NO₃-N?RE<15%?and SRP?RE<30%?during most flood events.However,the piecewise model had the best performance in estimating NO₃-N loads of flood events in the extremely wet year?RE<10%?.Weekly and biweekly sampling frequency would produce a sound estimation of monthly?-0.318%?and annual?-1 ±3%?NO₃-N loads but monthly sampling frequency produced a large deviation.NH₄-N and SRP were more dynamic?higher MMR?than nitrate and required more frequent sampling strategies to obtain robust load estimations.The model performance would be effectively improved if more frequent sampling of flood samples were observed during flood events observation.Using the online high-frequency monitoring technology of water quality and discharge,the "Direct Estimation Model" and "LOADEST Model" were integrated into the intelligent monitoring system of the Jiulong River to make real-time monitoring of nutrient loads available.To reach the goal of "annual load error<5%,monthly load error<10%",the maintain time of the online monitoring instrument should be set no more than 7 days for NO₃-N and NH₄-N measurements and no more than 2 days for SRP measurement.In conclusion,we combined observations and model studies at different time scales?monthly,seasonal,annual,and flood events-based?and summarized the mechanisms of hydrological controls on the export of different nutrient forms from Jiulong River.Moreover,an optimal model developed in this study was applied to make use of the high-frequency monitoring and simulation systems to produce a more accurate estimation of nutrient loads from the river.This study could provide theoretical guidance for the controls of total nutrient pollution of China's estuaries,monitoring and early warning of algal blooms,and the integrated management of sea and land in the context of global change.