Drought at the global scale in the 2nd part of the 20th century (1963-2001)

The large impacts of drought on society, economy and environment urge for a thorough investigation. A good knowledge of past drought events is important for both understanding of the processes causing drought, as well as to provide reliability assessments for drought projections for the future. Preferably, the investigation of historic drought events should rely on observations. Unfortunately, for a global scale these detailed observations are often not available. Therefore, the outcome of global hydrological models (GHMs) and off-line land surface models (LSMs) is used to assess droughts. In this study we have investigated to what extent simulated gridded time series from these large-scale models capture historic hydrological drought events. Results of ten different models, both GHMs and LSMs, made available by the WATCH project, were compared. All models are run on a global 0.5 degree grid for the period 1963-2000 with the same meteorological forcing data (WATCH forcing data). To identify hydrological drought events, the monthly aggregated total runoff values were used. Different methods were developed to identify spatio-temporal drought characteristics. General drought characteristics for each grid cell, as for example the average drought duration, were compared. These characteristics show that when comparing absolute values the models give substantially different results, whereas relative values lead to more or less the same drought pattern. Next to the general drought characteristics, some documented major historical drought events (one for each continent) were selected and described in more detail. For each drought event, the simulated drought clusters (spatial events) and their characteristics are given for one month during the event. It can be concluded that most major drought events are captured by all models. However, the spatial extent of the drought events differ substantially between the models. In general the models show a fast reaction to rainfall and therefore also capture drought events caused by large rainfall anomalies. More research is still needed, since here we only looked at a few selected number of documented drought events spread over the globe. To assess more in detail if these large-scale models are able to capture drought, additional quantitative analyses are needed together with a more elaborated comparison against observed drought events