Marinos Eliades (Cyprus) 1; Adriana Bruggeman (Cyprus) 1; Hakan Djuma (Cyprus) 1; Corrado Camera (Italy) 2
1 - The Cyprus institute; 2 - Universita Degli Studi Di Milano
The temporal variability and unpredictability of the magnitude of rainfall due to climate change is expected to cause changes both in runoff and evapotranspiration processes. These changes are difficult to model in semi-arid watersheds. The objectives of this study are (i) to observe the water balance components of a Pinus brutia forest during three hydrologically contrasting years (2015–2017) and (ii) to improve the representation of the observed evapotranspiration patterns and streamflow characteristics of semi-arid watersheds in a conceptual, four-parameter rainfall-runoff model (GR4J).
The water balance components were measured with throughfall gauges and soil moisture and sap flow sensors at the Agia Marina Xyliatou forest site, situated at the northern foothills of the Troodos mountains in Cyprus. GR4J was applied in the midstream and upstream areas of the adjacent Peristerona watershed. To reduce model complexity, the streamflow-groundwater exchange parameter was set to zero, while maintaining a good fit. Exponents of the routing store outflow and evapotranspiration equations were made tuneable. The models were evaluated against four different Nash-Shutcliffe efficiency criteria (standard, root squared, logarithmic, and inverse) and Bias for daily streamflow. We evaluated the model with the same criteria for monthly evapotranspiration.
The results from three years (2015–2017) of monitoring show a seasonal pattern of transpiration. Even though rainfall was significantly higher in 2016 (359 mm) than in 2017 (220 mm), transpiration was lower in 2016 (107 mm) than in 2017 (166 mm). This was due to the temporal distribution of rainfall during the year and the rain in fall 2016, which recharged the fractured bedrock. The trees were found to take up water from the fractured bed rock. On average, rainfall during the driest months (July-September) was 2% of the annual rain while tree transpiration was about 7% of rain. The new GR5J-dry showed better results than the original GR4J model during the calibration period (01/01/2015–30/09/2016), as it captured the monthly distribution of the observed evapotranspiration with an NSE of 41%. Also, GR5J-dry simulated zero streamflow during the dry period and had a higher standard NSE for streamflow (88%).
During the validation period (1/10/2016–31/12/2017), both models had a lower performance, indicating that a longer observation record is needed to capture the effects of the large rainfall variabilities between years. The results show the importance of long-term observations and of hydrologic model improvements for understanding the effects of climate change in semi-arid environments.