Developing models and forecasting turbidity in Esopus Creek, NY
In this repo, we share several codes for transforming flow forecasts (from NOAA River Forecast Centers and the U.S. National Water Model) into forecasts of turbidity. This intended as a proof-of-concept for building water quality forecasts from high-resolution monitoring data and a streamflow forecast source of your choice. The testbed for this approach is Esopus Creek, NY, a highly instrumented watershed in the Catskill Mountains of southeastern New York, USA.
-
00_functions: this is where all the functions live
-
01_data_discovery: this finds data from gages upstream of a single gage
-
02_observational_data_download_and_clean: this downloads discharge and turbidity data from the Esopus Creek gages
-
03_observational_data_prep: This cleans & formats observational data
-
04_event_delineation: this delineates storm events from observed discharge
-
05_model_development: this performs variable selection and trains quantile regression and LightGBM models
-
06_forecast_data_download: this downloads archived National Water Model and Northeast River Forecast Center streamflow forecasts
-
07_forecast_data_prep: this formats forecast data and calculates various antecedent parameters needed to make turbidity forecasts
-
08_make_forecasts: this generates turbidity forecasts
-
09_evaluate_performance: this evaluates the performance of benchmark turbidity predictions and turbidity forecasts