FcstDisagg

Forecast Disaggregation

Frequently, the seasonal forecast volume must be temporally disaggregated to a monthly or daily timestep. Three potential methods are proposed for forecast disaggregation.

Analog year

The analog year approach is frequently used by forecasters for forecast disaggregation (eg., Boroughs, 2013). In this approach, the forecaster selects a year thought to be similar to the current runoff year and scales the analog year to match the seasonal forecast volume.

Using an analog year approach is subjective and may result in cognitive biases by the forecaster due to the availability heuristic (Tversky and Kahneman, 1974). The availability heuristic is a potential bias in which people tend to overestimate the probability of occurrence of events more easily recalled.

quantile mapping

Quantile mapping uses the exceedence probability of the forecasted streamflow volumes to map to historical hydrographs, as described in (Pagano and Garen, 2003). The exceedence probability of each historical trace is calculated and the corresponding exceedence is selected from the forecast in question. The historical trace is then scaled to match the forecasted volume.

Downscaling using k-nearest neighbors

Downscaling using k-nearest neighbors has yet to be implemented. This approach has the advantage of maintaining the spatial and temporal correlation structure of historical conditions. The theory is based upon (Gangopadhyay et al., 2005).