Land use can offset climate change induced increases in erosion in Mediterranean watersheds
Xavier Rodriguez-Lloverasa *; Wouter Buytaertb and Gerardo Benitoa.
aGeology Department, National Museum of Natural Sciences, MNCN-CSIC, C/Serrano 115 dupl., 28006, Madrid, Spain. *
bDepartment of Civil and Environmental Engineering, Faculty of Engineering, Imperial College, Skempton Building, South Kensington Campus, Exhibition Road, SW7 2AZ, London, UK.
ABSTRACT:
The aim of this paper is to assess the impacts of projected climate change on a Mediterranean catchment, and to analyze the effects of a suite of representative land use practices as an adaptation tool to reduce climate change-driven erosion and hydrologic extremes. Relevant climatic variables from the ERA-Interim global atmospheric reanalysis of the European Centre for Medium-Range Weather Forecasts (ECMWF) were downscaled for the study area, and perturbed with the anomalies of 23 global circulation models for three emission scenarios (B1, A1B and A2). Both a projected daily rainfall time series for the period 2010 - 2100, and a single precipitation event with a one-hundred year return period were used to assess the impact of climate change. The downscaled data were fed into a distributed hydro-sedimentary model (TETIS) with five land use configurations representative of future demographic tendencies, geographical characteristics and land management policies (e.g. European Union CAP). The projected changes showed a general decrease in runoff and sediment production by the end of the century regardless of land use configuration. Sediment production showed a positive relationship with an increase in agricultural land and a decrease in natural land under present day agricultural management. According to our simulations, some conservation practices in agriculture can effectively reduce net erosion while maintaining agricultural production. As such, they can play a critical role as an adaptation tool to reduce climate change impacts in the 21st century.
KEY WORDS: Downscaling; Climate Change Adaptation; Land Use; Erosion; TETIS.
1. INTRODUCTION
The Mediterranean is one of the most sensitive regions to climate change in Europe (IPCC, 2014). Climate change projections for the southern Iberian Peninsula suggest a decreasing trend in annual average precipitation and an increase in heavy rainfalls by the end of the century (Barranco et al., 2014; Garcia et al., 2007; Rodrigo, 2010).Changes in total and extreme precipitation are projected to alter runoff production (Barranco et al., 2014), but the impact of changes on erosion and sediment yields at catchment scale has received less scientific attention (Bussi et al., 2014b).
Hydrological and environmental planning of Mediterranean watersheds requires an understanding of the future runoff and sediment yield response to climate and land use changes to enable adaptation to the potential impacts on freshwater resources and economic activities (e.g. agriculture). Sediment yields in Mediterranean catchments are mainly produced during high intensity precipitation events which may generate up to 40% of total erosion (Baartman et al., 2013; Rodriguez-Lloveras et al., 2015). At the same time, soil loss may be increased by inadequate land use and agricultural production techniques, deforestation, overgrazing, forest fires and construction activities (Boellstorff and Benito, 2005; Garcia-Ruiz, 2010; Puigdefábregas and Mendizabal, 1998). In these fragile Mediterranean environmental conditions, any soil loss higher than the estimated mean of 1.3 t ha-1 yr-1(Cerdan et al., 2010) can lead to a stage of irreversibility within a time span of 50-100 years (EEA, 1999; Gobin et al., 2004). Climate change may contribute to increased soil erosion as a consequence of the higher frequency of heavy rainfalls projected by climatic models (Döll et al., 2015; Kundzewicz et al., 2014). However, the net sediment production will result from the combined impact of climate (dryness, heavy rainfall, total rainfall) and land use conditions (agriculture, forest, shrubs), and their spatial and temporal variability. Future water and sediment yield projections need to consider different socioeconomic pathway scenarios for both climate change and land management at watershed scale.
Simulating climate change impacts on hydrology and sediment production requires the generation of time series of projected climatic variables. General Circulation Models (GCMs) produce projected climatic variables under different socioeconomic and technological development scenarios. Several studies have used climate change projections from GCMs and regional climate models to model long-term changes in sediment transport in Mediterranean watersheds (Bussi et al., 2014a; Nerantzaki et al., 2015; Nunes et al., 2013). Most of these models provide projected climate variables at a coarse spatial resolution, which reduces precipitation intensities and disregards local patterns of variability. This is one of the most critical characteristics of precipitation in the Mediterranean region (Gonzalez-Hidalgo et al., 2009; Xoplaki et al., 2004), and it is therefore important to simulate it in detail. For this reason, downscaling methods have been used to obtain representative climatic variables at smaller scales, that take local relief and elevation characteristics into account(Christensen et al., 2007). Among these, statistical downscaling uses statistical modelling techniques to extrapolate and interpolate results generated by dynamic models (Benestad et al., 2008). These statistical downscaling techniques have been widely used for hydrological projections (Barranco et al., 2014; Hertig and Jacobeit, 2008; Segui et al., 2010), but their application to the analysis of watershed erosion and sediment yields is still limited (Michael et al., 2005).Downscaling methods have the capacity to preserve the observed statistical structure (mean and dispersion) of the projected climate parameters at local scale, and to include heavy rainfalls not well reproduced by GCM projections.
Climate change impacts on soil erosion and degradation can be reduced by adequate land use practices (Boellstorff and Benito, 2005), thus effectively offsetting climate-driven increases in erosion in Mediterranean areas. When exploring the best land use practices for climate change adaptation, it is necessary to project representative land use configurations, considering non-natural variables such as demographic and socioeconomic factors, which cannot be easily predicted (Aguirre Segura et al., 1997; Arnold et al., 1998; Barriendos, 1997).
The purpose of this paper is to analyse the effects of different climate change projection scenarios on runoff and sediment production in a Mediterranean catchment, and to investigate how these effects can be mitigated by different land use configurations and agricultural techniques. The dataset of daily climate projections was obtained from the ECMWF ERA-Interim project combined with the 4th IPCC General Circulation Models (GCMs). Land use scenarios were established considering geography, demographic trends, traditional agricultural use and techniques and plant phenology. Changes in hydrology and sediment yields were calculated by routing the projected meteorological time series (2010-2100) through the TETIS distributed hydro-sedimentary model simulated under different future land use and land cover scenarios.
2. MATERIAL AND METHODS
2.1. STUDY AREA
The study area comprised a section of the upper catchment of the Guadalentin river (SE Spain, Andalusia) covering a surface area of 429 Km2 drained by the Rambla Mayor and Caramel Rivers (Fig. 1).The elevation ranges between 2045 and 687 m above the main sea level (m.a.m.s.l.; Fig. 1). The climate of the study area is Mediterranean (average annual temperature ~13 ºC) with semiarid characteristics in the lower part of the catchment, and mountainous features in the high reliefs. The average annual precipitation, calculated from 41 years of hourly precipitation records at 13 meteorological stations (Fig. 1B), ranges from 460 mm at the highest meteorological station in the catchment (1186 m.a.m.s.l.) to 320 mm at Valdeinfierno dam (697 m.a.m.s.l.). Rainfall occurs mainly in spring and autumn, whereas summers are characterized by dry conditions. As reported in Rodriguez-Lloveras et al. (2015),soils within the study area are poorly developed, in agreement with its semiarid Mediterranean characteristics. In the northern part of the catchment, soils are highly degraded with a dominant occurrence of Calcaric Regosols, Cambisols and Calcisols. In the southern part, Leptosols are dominant in the uplands and Regosols in the lowlands; the latter are especially predominant in the eastern part, where the majority of agricultural land is located. The soil organic matter content is usually moderately high, in general between 2 and 10%, with maximum values of 17%. The soils are deeper in the lowlands (depth: 50 – 100 cm) and shallower in the uplands (20 – 30 cm). The soil texture is mostly clay loam, loam and silt loam, with some sandier patches located in the central part.
Present day land use combines natural vegetated areas (49%) with areas of human uses (51%), which include agriculture (50 %, mainly cereal farming) and urban occupation (1 %). Areas of natural vegetation (trees and shrubs) are located in the higher mountainous areas and slopes, whereas human activities are concentrated in flat areas and valley bottoms (Fig. 1C).
2.2. CLIMATE CHANGE DATA AND SCENARIOS
In this study, the meteorological statistical downscaling model GLIMCLIM (Chandler and Wheater, 2002; Yan et al., 2002; Yan et al., 2006; Yang et al., 2005)was applied to daily data from 13 meteorological stations within the study area region(Fig. 1B).The GLIMCLIM downscaling model is a two-step model: the first step involves fitting a gamma distribution to the meteorological time series, while in the second step, generalised linear models are used to find a relation between the statistical moments of the distribution of the daily meteorological variables and synoptic climate variables in and nearby the study region. This predictive statistical model can be used to generate synthetic time series with statistical properties that are similar to the original observed time series. As GCMs are more skilful at projecting synoptic variables, projected changes in these variables can be used to generate stochastic time series for future periods. For more information about the GLIMCLIM procedure, see Yan et al., 2006 and references therein.
In this study, synoptic data for the period 1979 – 2012 were acquired at monthly scale from the ERA-Interim reanalysis model(Dee et al., 2011) developed by the European Centre for Medium-Range Weather Forecasts (ECMWF). A total of 124 synoptic variables for a period of 33 years (1979-2012) from 4 ERA-Interim cells that cover southern Spain were initially considered. A total of 25 synoptic variables were retained for the generalised linear model.
Climate projections of three socioeconomic scenarios considered representative of lower, medium and high greenhouse gas emissions (B1, A1B and A2 respectively;(Nakicenovic et al., 2000)were selected. The precipitation and temperature anomalies of 23 GCMs (Table 1) for these three scenarios (B1, A1B and A2) were included in the IPCC Fourth Assessment Report (AR4; (IPCC, 2007). The reference period used to calculate the climatic anomalies corresponded to the post-industrial reference period (1961-1990), considered in the 20th Century Climate in Coupled Models (20C3M). The anomalies were calculated for three 30-year periods (2010-2040, 2040-2070, 270-2100;(IPCC, 2007). The effects of climate change on hydrology and sediment production were calculated by combining the 30-year anomalies with the downscaled meteorology and applying this projected meteorological data to a distributed hydrological model.
2.3. HYDROLOGY AND SEDIMENT PRODUCTION
The TETIS distributed hydrological and sediment model was used to reproduce the hydro-sedimentary response to climatic and land use scenario data. TETIS is a conceptually-based spatially distributed model which has the capacity to simulate catchment hydrology and sediment transport budgets at event scale over long-term simulations as well as for future climate scenarios (Bussi et al., 2014a; Bussi et al., 2014b; Bussi et al., 2013; Frances et al., 2007), and it simulates all the main components of the land phase of the hydrological cycle(Frances et al., 2007). It is composed of a hydrological and a sediment transport sub-model. The hydrological sub-model is based on a cell-tank structure, where terrain is divided into cells (or pixels), each of which is conceptualised as a system of tanks which represent a hydrological process (snowmelt, canopy interception, soil static storage, soil gravitational storage and aquifer storage). The first two tanks (canopy interception and soil static storage) are filled by precipitation and are only emptied by evapotranspiration. The remaining tank depends on water flow, which is divided into overland flow and infiltration, depending on the soil infiltration capacity. The water infiltrated into the soil is separated between interflow and aquifer flow depending on the soil and aquifer properties. The total flow outlet which enters the drainage network is calculated from the sum of overland flow, interflow and base flow (Frances et al., 2007; Rodriguez-Lloveras et al., 2015; Velez et al., 2009). The total flow is routed downstream using geomorphological kinematic wave methodology, based on the drainage network hydraulic geometry (Rodriguez-Lloveras et al., 2015).
The sediment sub-model depends on the balance between flow sediment transport capacity and sediment availability (Bussi et al., 2014b). The sediment transport capacity of the overland flow is calculated by means of the modified Kilinc-Richardson equation (Julien, 2010), with the overland flow as input, and its transport capacity is determined using the Engelund and Hansen (1967) equation. The sediment available is classified into three categories according to its textural size (sand, silt and clay). Together, this parameter and flow transport capacity determine the amount of sediment transported downstream, as well as particle settling velocity, which is used to separate the transported material into suspended and deposited sediment(Bussi et al., 2014a; Bussi et al., 2013). Implementation, calibration and validation of the TETIS model for the study catchment are explained in more detail in Rodriguez-Lloveras et al. (2015).
All the 23 GCM projected rainfall and temperature datasets under B1, A1B and A2 scenarios were entered into the rainfall-runoff analysis. Results of rainfall-runoff simulations using this climatic time series provided runoff and sediment production outputs at daily resolution throughout the 21st century. The resulting GCMs with minimum, medium and maximum runoffs and sediment yields were selected to characterise hydrology and sediment transport in the study area.
A major drawback of the hydro-sedimentary model outputs based on climate change scenarios implemented in the catchment is the misrepresentation of large peak flows. The absence of extreme flow discharges is not consistent with the present hydrological behaviour of the catchment (Rodriguez-Lloveras et al., 2015), or the historical records (last 300 years) and palaeoflood records (last 1000 years) of extreme floods(Benito et al., 2010). Given the lack of rainfall extremes as a GCM limitation, the occurrence of such heavy rainfall and its effects on future hydro-sedimentary projections was tested. Thus, a statistical analysis of annual maximum daily rainfall was performed considering the annual maximum daily rainfall over the period 1971 to 2012 (Fig. 2). A square-root exponential type distribution function SQRT-ET Max (Etoh et al., 1987), frequently used in Spanish Mediterranean areas to determine maximum precipitation events (CEDEX, 1999), was fitted using the maximum likelihood method.