IMPROVING PREDICTIVE MODELS OF IN-STREAM PHOSPHORUS CONCENTRATION BASED ON NATIONALLY-AVAILABLE SPATIAL DATA...

Area of Science:

• Watershed science
• Environmental modeling
• Water quality management

Background:

• Nationally available spatial data are increasingly used in watershed management but their local suitability is often unexamined.
• Understanding factors influencing total phosphorus (TP) concentrations is crucial for watershed health.

Purpose of the Study:

• To assess if incorporating local watershed data improves total phosphorus (TP) models developed from national datasets.
• To compare the predictive accuracy of spatial stream network (SSN) models against traditional nonspatial multiple linear regression models.

Main Methods:

• Developed and compared TP concentration models using national spatial data versus models augmented with local watershed-specific data.
• Utilized a spatial stream network (SSN) modeling approach and compared it with multiple linear regression (nonspatial) models.
• Identified significant local covariates, such as septic systems and point source TP loads.

Main Results:

• The SSN model incorporating local covariates demonstrated the highest goodness of fit and predictive accuracy.
• Nonspatial models using only national data yielded the lowest predictive performance.
• Septic systems and point source TP loads were identified as significant local predictors of TP concentrations.

Conclusions:

• Spatial stream network (SSN) modeling enhances prediction accuracy when using national spatial covariates.
• Integrating local watershed data, especially septic system locations, substantially increases the accuracy of TP predictions.
• Future national spatial databases should include detailed local variables for improved watershed management.