Case Study: Modeling Historic Flows in Ungauged and Data-Sparse Basins
Upstream Tech has developed a process using machine learning hydrology models to predict stream and river flows at locations where the flow is modified by humans and where gauge data is not available, conditions we call Ungauged Actual Flows. A key motivating factor for our work predicting Ungauged Actual Flows is to gap-fill holes in the gauged data record at locations with short or entirely unavailable flow gauge data records.
A case study for predicting at ungauged locations presents our modeling approach, our gap-filling workflow, and finally shares results for prediction and validation results at several example sites. This report focuses on the extension and gap-filling of historic time series and validation. We follow a similar set of methods for generating forecasts in this context.