Show Full Disclaimer
Combined Uncertainties in the CLIRUN-II Results
As stated previously, the GRDC-UNH runoff data is the predicted average monthly runoff over the years 1986 to 1995.  In general, a longer (in time) calibration runoff time series is preferred over a shorter calibration runoff time series for the best CLIRUN-II runoff estimation results.  But, since the GRDC-UNH data is the best global data available, the model was calibrated using the average monthly values (12 values for each grid cell).  There are issues that arise when using average monthly runoff data for calibration, instead of time series runoff data.  The biggest, and most obvious, issue involves the loss of the peak values, either low or high runoff estimations, in the calibration process.  During the calibration process, many sets of the calibrated basin parameters (soil information, land cover information, etc.) are estimated.  Once a set is estimated, the model estimates a time series or runoff values.
These runoff values are then compared to the observed runoff time series used for calibration.  If a given criterion is met, the estimated basin parameters are considered to be a good match, and are used.  If the criterion is not met, a new set of parameters are estimated and the process starts over.  In the case of a time series, the criteria is basically the sum of squares of residuals is used over the entire time series.  Alternatively, when the average monthly values are used, the sum of squares of residuals is used to compare the GRDC-UNH mean monthly values with the modeled mean monthly values.
 In the latter method, the extreme values are not taken into consideration because the calibration data (GRDC-UNH) does not have extreme values.  But if the observed weather data contains extreme weather events like very dry or rainy months, then CLIRUN-II will produce extreme runoff events as well, since it is a physically based model, sensitive to weather input.  So the extreme events will be produced, but there is no way to know if the extreme events that are modeled are the extreme events that actually (or would) occur in a given basin.  As stated previously, the CRU TS datasets are ‘space optimized’ rather than ‘time optimized.’  When the data was interpolated (both in time and space) to fit the half-degree grid and the years 1901 – 2002, the interpolation was ‘relaxed to the 1961 –
1990 mean.’  So the extreme weather values could be missing from the CRU database where station coverage is poor.  If these extreme weather values are missing from a particular time and location in the CRU TS dataset, there would be no information to allow CLIRUN-II to produce extreme runoff events (for the given place and time).  So, given that the UNH-GRDC dataset and the CRU TS dataset both have the tendency to under-represent extreme events (runoff and weather, respectively), there is a good chance that extreme runoff events are under-represented in the CLIRUN-II results.
Water and Climate Change: Understanding the Risks

The data presented here was developed by the World Bank Water Anchor with the aim to gain insights into climate change impacts on potential future hydrology and to establish a common platform of information on the behavior of key hydrologic drivers across World Bank regions at a scale appropriate for policy and investment decisions.

Here you can explore basin and country level hydrological statistics for 3 emissions scenarios, 23 Global Circulation models and two future time periods (2030-2039 and 2050-2059).

Please be warned that this data is not intended for use in any design study.

Click here to get access to data and indicators