Regional climate change projections and hydrological impact analysis for a Mediterranean basin in Southern Italy

Summary

A number of future water availability scenarios in the Crati River Basin (Southern Italy) are made by applying the outputs of three Regional Climate Models (RCMs) RegCM, HIRHAM and COSMO-CLM to the newly developed Intermediate Space Time Resolution Hydrological Model (In-STRHyM). In-STRHyM is a fully distributed hydrological model detailed enough to describe the hydrological processes of several small-medium sized Mediterranean basins. It has a relatively simple structure and is suitable for long period simulations to be undertaken within acceptable time frames. The analysis was performed using two time slices (1961–1990 and 2070–2099) with the SRES A2 (HAD3AM) and A1B (ECHAM5/MPI-OM) scenarios. Observed biases in simulated precipitation and temperature fields during the control period (1961–1990) were corrected before using meteorological outputs from each RCM as input for In-STRHyM. A sensitivity analysis is used to help verify this procedure.

While hydrological model simulations are based on different greenhouse gas emission scenarios, GCMs, RCMs and bias correction parameters and hence show noticeable differences, they all agree on a general reduction in future water resource availability. An increase in the average annual temperature between +3.5 °C and +3.9 °C and a decrease between −9% and −21% in the cumulative annual precipitation are projected, leading to a drastic reduction in snow accumulation of between −82% and −92%. Evapotranspiration is expected to increase in the wintertime and decrease in summertime, with the water stress period increasing on average by 15 days. Mean annual reductions are predicted for root zone soil moisture (between −12.8 ± 1.9% and −20.7 ± 1.9%, with reductions reaching −37.7 ± 2.4% during summer), groundwater storage (−6.5 ± 1.4% and −11.6 ± 1.6%), surface runoff (−25.4 ± 6.0% and −41.2 ± 5.0%) and a significant increase in runoff variability.

Introduction

The Mediterranean is a ‘hot spot’ for climate change (Giorgi, 2006, IPCC, 2007). Several investigations (e.g. Ulbrich et al., 2006, Giorgi and Lionello, 2008, Sheffield and Wood, 2008) project significant impacts on both mean precipitation and variability, with relevant consequences on land surface water availability, decreasing by 20% according to Mariotti et al., 2008. The confirmation of these results and the general improvement in identifying the variability and changes in the main hydrological processes, is a crucial task for better management of water resources in Mediterranean areas. These areas are characterized during the last decades by large increases in population, rising living standards, the development of irrigated agriculture and new activities (particularly tourism related), where many aquifers are already overexploited and surface waters are endangered (Cudennec et al., 2007).

Among other issues in its recent White Paper on the adaptation to climate change (COM, 2009), the European Commission highlights the importance of developing methods, models and prediction tools. The methodology usually followed to assess the hydrological consequences of climate change basically consists of a three-step process (Xu et al., 2005): (1) the development and use of general circulation models (GCMs) to provide future global climate scenarios under the effect of increasing greenhouse gases, (2) the development and use of downscaling techniques (both statistical methods and nested regional climate models, RCMs, which are being continuously improved) for “downscaling” the GCM output to the scales compatible with hydrological models, and (3) the development and use of hydrological models to simulate the effects of climate change on hydrological regimes at various scales. However, uncertainties within this framework have to be taken into account such as the internal variability of the climate system, model structure and parameterizations at different spatial and temporal scales, the downscaling techniques and bias correction methods and the choice of future climate scenarios.

Concerning the reliability of models, the performance of each GCM (i.e. a model by which the future climate can be assessed) must be carefully analyzed. Currently, many GCM intercomparison projects have been or are being performed (e.g. PILPS, Henderson-Sellers et al., 1993; or C⁴MIP, Friedlingstein et al., 2006), leading recently to the ambitious Program for Climate Model Diagnosis and Intercomparison (PCMDI). Several studies have been conducted for the purposes of analyzing GCM downscaling techniques (e.g. Murphy, 1999, Murphy, 2000, Hewitson and Crane, 2006, Schmidli et al., 2007) and comparing different RCMs (Feng and Fu, 2006, Inoue et al., 2006, Jacob et al., 2007, Smiatek et al., 2009, van Roosmalen et al., 2010). However, fewer studies have focused on uncertainties related to downscaling to the resolution of the hydrological impact model (Kunstmann et al., 2004, Kunstmann and Stadler, 2005, Dibike and Coulibaly, 2005, Prudhomme and Davies, 2009, Quintana Seguì et al., 2010). The typical resolution of a RCM (10–50 km) is not enough for most hydrological models and thus they need to be further downscaled and bias-corrected (Christensen et al., 2008). Further downscaling of results from GCMs/RCMs to individual sites or localities for hydrological impact studies is particularly needed in several regions of the Mediterranean, where the basin size is often relatively small (about or less than 10³ km², e.g. in Italy or Greece). For this reason, the bias correction technique is important where the “delta-change” method (Hay et al., 2000) is widely used. Other procedures have been suggested (among others, e.g. Kunstmann et al., 2004) that also account for the uncertainty of the input data.

In this study, the hydrological impact of climate change in a 10³ km² Southern Italian catchment is assessed using several models and by adopting two different SRES scenarios (Nakícenovíc et al., 2000). Specifically, the results of two RCMs (HIRHAM and RegCM) are applied using the HAD3AM GCM with the A2 scenario, and the RCM COSMO-CLM with the A1B scenario of the ECHAM5/MPI-OM GCM. In this particular field, this work represents the only available high resolution climate change data from regional downscaling experiments with grid resolutions finer than 20 km. Common scenarios for different GCMs were not available. Recently, regional data is also available from the ENSEMBLES project (Linden et al., 2009), where several RCMs are run with different GCM forcings, but at a spatial resolution of only 25 km and only one (A1B) climate scenario. In our work, the precipitation and 2 m temperature fields derived from each RCM are compared with observed data and bias-corrected through the method proposed by Kunstmann et al. (2004). These corrected fields are then used in a new distributed hydrological model, the Intermediate Space Time Resolution Hydrological Model (In-STRHyM), developed for the purpose of evaluating the flow regime over long periods in the main watersheds of Southern Italy. The stability of this procedure is then analyzed by varying the correction parameters.

The innovation and specific contribution of this study to the scientific debate about future climate assessment can be summarized in the following items:

• A spatially explicit hydrological model specifically suited for climate impact analyses in relatively small study areas was developed and applied. The use of this model allowed a detailed resolution of the impact study (daily time scale and 1 km space scale) not previously available. Previous studies in the target region were limited to the application of coarse resolution GCMs without any detailed water balance models.

• Most of the climate change results available so far are general and related to the whole Mediterranean area. The downscaling approach used from GCMs to RCMs down to the new distributed hydrological model allowed the generation of images providing a spatial distribution of the changes in the main hydrological variables for a relatively small basin.

• Results were derived from three different RCMs with the most recent climate change data at a spatial resolution below 20 km for the analyzed area. These RCMs were in turn derived from two GCMs by applying different SRES scenarios, thereby reducing uncertainty from the specific choice of GCM, RCM and emission scenario. Hence, this study addresses a specific gap in understanding by providing results at the resolution of the hydrological impact model.

The main objectives of the paper are: (1) the preliminary validation of the distributed hydrological model, (2) a water resource availability assessment based on the chosen climate change scenarios, and (3) the comparison of synthetic and distributed results achieved for the main components of the hydrological cycle. These are obtained by using different models and scenarios in order to evaluate their agreement with each other, and the reliability of the trends.

In the next sections, the study area and the climatological and hydrological models are described. Specifically, the hydrological model is calibrated and validated. Following this, results of regional climate modelling are shown and then the bias correction is explained. Finally, synthetic and distributed results from several components of the hydrological cycle (snow, evapotranspiration, root zone soil moisture, groundwater storage and surface runoff) are shown and compared.