The Republican River is the primary inflow to Milford Lake and drains areas of Kansas, Nebraska, and Colorado. Milford Lake has been listed as impaired and designated hypereutrophic by the Kansas Department of Health and Environment because of excessive nutrient loading. Milford Lake had confirmed harmful algal blooms every summer from 2011 through 2017 and in 2020 and 2021.
In the lower Republican River drainage basin, the Regional Conservation Partnership Program, administered by the Natural Resources Conservation Service, provides reimbursement to agricultural producers that implement best management practices intended to decrease sediment and nutrient runoff and loading into Milford Lake. Sediment and nutrient loads could potentially be driving factors in the development of harmful algal blooms in the reservoir.
Since July 2018, the U.S. Geological Survey, in cooperation with the Kansas Water Office, has collected continuous and discrete water-quality data at the Republican River at Clay Center, Kansas, streamgage (U.S. Geological Survey station 06856600), which is about 15 river miles upstream from Milford Lake. This report documents site-specific regression models for the computation of continuous concentrations of suspended sediment, total nitrogen, total phosphorus, and total carbon developed using continuous and discrete data collected from July 24, 2018, the date of continuous water-quality monitor installation, through March 31, 2021. The objective of this study is to characterize sediment and nutrient transport in the Milford Lake drainage basin before, during, and after best management practice implementation using the models described in this report.
The explanatory variable turbidity explained a high amount (72–96 percent) of the variance in suspended-sediment, total nitrogen, total phosphorus, and total carbon concentrations. Statistical plots for the four selected models showed the desired normality and homoscedasticity in residuals, and model standard error ratios indicated that recomputing each selected model after removing a randomly selected 10 percent of the data did not substantially change model coefficients.
