Elsevier

Journal of Hydrology

Volume 383, Issues 1–2, 15 March 2010, Pages 111-124
Journal of Hydrology

Comparison of three downscaling methods in simulating the impact of climate change on the hydrology of Mediterranean basins

https://doi.org/10.1016/j.jhydrol.2009.09.050Get rights and content

Summary

Studies of the impact of climate change on water resources usually follow a top to bottom approach: a scenario of emissions is used to run a GCM simulation, which is downscaled (RCM and/or stastistical methods) and bias-corrected. Then, this data is used to force a hydrological model. Seldom, impact studies take into account all relevant uncertainties. In fact, many published studies only use one climate model and one downscaling technique. In this study, the outputs of an atmosphere–ocean regional climate model are downscaled and bias-corrected using three different techniques: a statistical method based on weather regimes, a quantile-mapping method and the method of the anomaly. The resulting data are used to force a distributed hydrological model to simulate the French Mediterranean basins. These are characterized by water scarcity and an increasing human pressure, which cause a demand in assessments on the impact of climate change hydrological systems. The purpose of the study is mainly methodological: the evaluation of the uncertainty related to the downscaling and bias-correction step. The periods chosen to compare the changes are the end of the 20th century (1970–2000) and the middle of the 21st century (2035–2065). The study shows that the three methods produce similar anomalies of the mean annual precipitation, but there are important differences, mainly in terms of spatial patterns. The study also shows that there are important differences in the anomalies of temperature. These uncertainties are amplified by the hydrological model. In some basins, the simulations do not agree in the sign of the anomalies and, in many others, the differences in amplitude of the anomaly are very important. Therefore, the uncertainty related to the downscaling and bias-correction of the climate simulation must be taken into account in order to better estimate the impact of climate change, with its uncertainty, on a specific basin. The study also shows that according to the RCM simulation used and to the periods studied, there might be significant increases of winter precipitation on the Cévennes region of the Massif Central, which is already affected by flash floods, and significant decreases of summer precipitation in most of the region. This will cause a decrease in the average discharge in the middle of the 21st in most of the gauging stations studied, specially in summer. Winter and, maybe spring, in some areas, are the exception, as discharge may increase in some basins.

Introduction

The Mediterranean basin is a quasi-closed sea with a marked orography on its periphery and a high urbanization of its coastline. Its climate is characterized by mild winters and hot and dry summers. The marked orography often triggers intense events that may cause flash floods and the hot and dry weather in summer causes low flows to be long and severe. In this context, for planning purposes, it is important to evaluate the possible impacts of climate change on water resources in such a region.

Global climate models (GCM) are the main tool used to study the future climate. According to Giorgi and Lionello (2008), the study of several GCM simulations shows “a robust and consistent picture of climate change over the Mediterranean emerges, consisting of a pronounced decrease in precipitation, especially in the warm season, except for the northern Mediterranean areas (e.g. the Alps) in winter”. It is also expected that the variability increases. In fact, according to Giorgi (2006) the Mediterranean basin is one of the planet’s hot-spots of climate change.

However, GCMs do not have enough resolution to study the regional and local scales. Their current resolution of 300 km (Solomon et al., 2007) misses most of the important relief surrounding the Mediterranean basin. Furthermore, at this scale, they are often biased. This obliges us to downscale the outputs of these models.

The usual strategy in impact studies has a top to bottom structure. Global socio-economic assumptions are made (Nakicenovic et al., 2000), which are then used to force GCMs, which are then downscaled and unbiased. This downscaling can be dynamical (computationally expensive) or statistical (less expensive) (Mearns et al., 1999). If the chosen method is dynamical, a limited area atmospheric model, which can simulate in more detail the climate on a smaller area, is forced at the edges of the domain by the outputs of a GCM (Hewitson and Crane, 1996). These models are known as regional climate models (RCM) and have a typical resolution of 50 km or 25 km. Often, dynamical and statistical downscaling methods are presented as mutually exclusive, but, in fact, as it will be seen in further sections, they can be used together.

The resolution of a RCM is not enough for most hydrological models, thus they need to be further downscaled and bias-corrected (Christensen et al., 2008) to produce atmospheric forcings at the adequate resolution (10 km) (Wood et al., 2004). Thus it is necessary to further downscale the output of these models and to develop methods to reconstruct the regional climate in relation to climate on a larger scale.

In these studies, the emission scenario and the GCM are the main sources of uncertainty (Boé, 2007, Maurer and Hidalgo, 2008). But, unfortunately, each step of the downscaling procedure also has associated uncertainty. All these uncertainties add up and constitute a cascade of uncertainty that must be taken into account. Thus, a complete impact study must look at all kinds of uncertainty. Many studies, have focused on the uncertainty related to the GCM (Christensen and Lettenmaier, 2007, Hamlet and Lettenmaier, 1999, Maurer and Duffy, 2005, Minville et al., 2008, Wilby et al., 2006) but fewer studies have focused on uncertainties related to downscaling to the resolution of the impact model (Boé et al., 2007, Dibike et al., 2005, Khan et al., 2006), which might also be important and is often neglected.

Within this study we look at the impacts of climate change on the French Mediterranean basins. Our goal is to force the hydrological model SIM with three atmospheric forcings representing the climate of the future. These forcings are build from the same RCM simulation using three different methods of downscaling and bias-correction. This should enable us to estimate the hydrological response to climate change, and to estimate the uncertainties related to the last step of downscaling and bias-correction of the climate simulation.

Section snippets

The french Mediterranean context

This article is focused on the French Mediterranean region. Fig. 1 shows the French Mediterranean basin, plus some rivers that do not reach the Mediterranean sea but are Mediterranean in climatological terms. These are situated on the Massif Central.

The largest French Mediterranean basin is the Rhône. Two of the main tributaries of the Rhône are alpine and have a very important nival component. These tributaries are also heavily influenced by hydropower production. But, in our context, we are

Methodology

In this study, three different methods are used to downscale and bias-correct the outputs of one single RCM simulation, using a gridded database of observations. In the next sections, the gridded database, the RCM and the downscaling methods are described.

Description of the hydrological model

In this study, a recent version (Quintana Seguí et al., 2009) of the SAFRAN-ISBA-MODCOU (SIM) model (Habets et al., 2008) is used. This model is the result of combining the SAFRAN meteorological analysis, the ISBA surface scheme and the MODCOU hydrogeological model. Only the main features of the model are described in this paper.

ISBA (Boone et al., 1999, Noilhan and Planton, 1989) is a soil–vegetation-atmosphere transfer (SVAT) scheme. It is used to simulate the exchanges in heat, mass and

Results

Two periods of 30 years were selected to compare present and future climate. For present climate, it was chosen to study the period August 1970–July 2000. The period selected for the future is: August 2035–July 2065.

The significance of the anomalies is evaluated using an adaptation of the Student test that does not require the assumption of the equality of the variances of the compared samples. This adaptation is often referred to as the Welch’s test (Welch, 1947).

Discussion and conclusions

There are many sources of uncertainty in impact studies. The main source is related to the GCM simulation (Boé, 2007), which is often taken into account, but many studies do not take into account the uncertainties related to the final step of downscaling and to the bias-correction of GCM or RCM simulations. In this study, the uncertainties related to this last step were assessed.

Relating precipitation, it was shown that the methods produce similar long term annual averages, but there are

Acknowledgements

This work was partly supported by the program ACI-FNS “Aléas et Changements Globaux” of the French research ministry under the project CYPRIM and the French Agence nationale de la recherche under the project MEDUP. The authors are grateful to Joël Noilhan for his continuous support during this research and his previous work, which made this study possible. The authors are also very grateful to the anonymous reviewers whose comments and suggestions helped to improve the clarity of the text.

References (45)

  • J. Boé et al.

    Statistical and dynamical downscaling of the seine basin climate for hydro-meteorological studies

    International Journal of Climatology

    (2007)
  • J. Boé et al.

    Projected changes in components of the hydrological cycle in French river basins during the 21st century

    Water Resources Research

    (2009)
  • Boone, A., 2000. Modélisation des processus hydrologiques dans le schéma de surface isba: Inclusion d’un réservoir...
  • A. Boone et al.

    An intercomparison of three snow schemes of varying complexity coupled to the same land surface model: local-scale evaluation at an alpine site

    Journal of Hydrometeorology

    (2001)
  • A. Boone et al.

    Inclusion of a third soil layer in a land surface scheme using the force–restore method

    Journal of Applied Meteorology

    (1999)
  • Y. Caballero et al.

    Hydrological sensitivity of the Adour-Garonne river basin to climate change

    Water Resources Research

    (2007)
  • N.S. Christensen et al.

    A multimodel ensemble approach to assessment of climate change impacts on the hydrology and water resources of the colorado river basin

    Hydrology and Earth System Sciences

    (2007)
  • J.H. Christensen et al.

    On the need for bias correction of regional climate change projections of temperature and precipitation

    Geophysical Research Letters

    (2008)
  • J.W. Deardorff

    A parameterization of ground-surface moisture content for use in atmospheric prediction models

    Journal of Applied Meteorology

    (1977)
  • B. Decharme et al.

    Impact of an exponential profile of saturated hydraulic conductivity within the ISBA LSM: simulations over the Rhône basin

    Journal of Hydrometeorology

    (2006)
  • M. Déqué et al.

    An intercomparison of regional climate simulations for europe: assessing uncertainties in model projections

    Climatic Change

    (2007)
  • Y. Durand et al.

    A meteorological estimation of relevant parameters for snow models

    Annals of Glaciology

    (1993)
  • Cited by (145)

    View all citing articles on Scopus
    View full text