Study of historical evolution (1979-2014) of key water cycle variables in the pyrenees using observations and modeled data

The natural border between Andorra, France and Spain are the Pyrenees, a South-Western European mountain range with a great environmental diversity: from Atlantic to Mediterranean climates, from high mountains to cliffs touching the sea, and from humid to semi-arid  conditions. Thus this region is particularly sensitive to climate and global change. On the other hand, this territory is the primary source of water in the region, feeding the runoffs and recharge zones of the region's main catchment basins. Rapid changes in the environment can have an influence on the availability of water resources downstream, increasing the uncertainty to an already tough water management situation.

Scientists use hydrological data to detect and quantify climate variability and change. Although data from gauging stations are basic to study the temporal evolution of water resources, more than these punctual data are needed for a regional study, as many relevant variables of the water cycle, such as evaporation, are seldom observed. Furthermore, models are necessary to study the future climate, but we need first to check if the models faithfully reproduce the intended processes. Therefore, hydrological modeling plays an important role in water resources studies, as they allow us to quantify the main components of the water balance (precipitation, evapotranspiration, drainage/recharge, runoff and streamflow) and the main stocks (soil moisture and snow) for the entire region. 

We have used observation values from non-influenced gauging stations and hydrological outputs of two different modeling tools (the fully distributed model SASER, and the semi-distributed model SWAT) to study the historical evolution (1979-2014) of the natural continental water cycle in the Pyrenees. The comparison of observational data with models, as well as models between them, will allow us to detect, evaluate and analyze the main sources of uncertainty.

We computed monthly, seasonal and annual statistics for three time periods (1979-2014, 1989-2014 and 1999-2014). Thus, we made and analyzed trends for the time series of the different variables applying a time series pre-whitening. These trends have been calculated with the Sen's slope estimator assuming that they are linear. The significance of the trends was estimated with the Mann-Kendall test on the pre-whitened time series with the statistical significance tested at the 95% level.

This work is a contribution to the EFA210/16 PIRAGUA project.