Low flow prediction

Question 1

Predictability

Answer 1

 Droughts are less dynamic and develop slowly. So, long term forecasts are possible and of practical relevance. (Lead time: weeks to months)

 Even alone the monitoring of groundwater levels, soil moisture, status of reservoirs, discharge etc. can help to assess the future drought potential.

 Simple forecasts for low flows are often scenario based concerning the input (e.g. scenario „no precipitation“ provides lower limit of discharge)

 Better forecasts consider meteorological predictors like circulation pattern, teleconnection, anomalies of circulations (El- Niño)

Question 2

Methods

Answer 2

1. Conceptual hydrological models:
    Forecast running the model without precipitation input (“Scenarios”)
    Application of recession curves (see section 4.1.4)
2. Stochastic time series and regression models:
    Application of ARMA – models for flow forecasts
    Application of regression models with various predictors (soil
   moisture, precipitation, snow, clime variables, etc.)
3. Soft-Computing methods:
    Artificial neural networks (ANN) and fuzzy rule based methods (FRBM); (see also “Geostatistics & Softcomputing”)
    Consideration of fuzzy knowledge and prediction of categorical variables (e.g. wet, dry, mean) is possible