Moroccan Climate in the Present and Future: Combined View from Observational Data and Regional Climate Scenarios [chapter]

K. Born, M. Christoph, A. H. Fink, P. Knippertz, H. Paeth, P. Speth
Environmental Science and Engineering  
Introduction Morocco is located between the arid regions of the western Sahara and the moderate Mediterranean and Atlantic regions. Landscape types reach from flat areas in the north-western part to high-mountain areas in the Atlas and Rif. Therefore, we can find a large variety of climates ranging from moderate humid and subhumid climates at the northern slope of the High Atlas over mountain climates to semi-arid and arid climates south of the Atlas. In this tension field, agricultural
more » ... gricultural production and local economy depend very much on water availability, and thus, mainly on rainfall variability. In the past, periods of successive dry years have repeatedly shown the vulnerability to water scarcity, which results in threatened livelihoods of farmers and nomad families living from pasturing. This leads to a stream of migrants heading towards the large cities at the Moroccan coast and even to Europe. In order to counteract the peril of water scarcity, regional planning greatly benefits from information on future climate variability. Of course, the assessment of future climate scenarios in this very heterogeneous region is still a challenge for the climate research community. In this study, the actual state of climate research in IMPETUS West-Africa with respect to rainfall and temperature variability and of research on future climate scenarios is illustrated. We start with an overview of available information from observational data and a few remarks on their quality. In the second part, results of regional climate modelling (RCM) are discussed with respect to future climate variability especially in regions south of the Atlas Mountains, where the water stress is clearly the limiting factor of agricultural production. An analysis of climate and climate variability always starts with the acquisition of observational information in the region of interest. In Morocco, the density and quality of observations is better than in most other parts of Northern Africa, but still relatively sparse when we think of the spatial heterogeneity of climates. The first and most important source of information stems from SYNOP weather stations, which contribute to the WMO network and deliver data of a relatively high quality standard for several decades now. Unfortunately, most stations in Morocco and the western parts of Algeria are located north of the Atlas Mountains, which leaves a lack of information exactly in the southern region, where the impact of a climate shift is expected to be very strong (Fig. 1) . The SYNOP data can be extended by information from individual climate stations operated by regional water management facilities and also by IMPETUS West-Africa, but the quality standard and temporal completeness of these data is not as high as for SYNOPs. Therefore, these data are mainly used to control the quality of derived climate products and do not contribute to the description of regional climates in Morocco. In order to generate spatially distributed information on near-surface climates, the observations have to be interpolated with adequate statistical techniques. Two projects have produced world-wide data sets of climate-related meteorological parameters from a large number of available observations, which have a spatial resolution of at least 0.5° longitude and latitude. These are climate data from the Climate Research Unit of the University of East Anglia, hereafter called CRU, in the version TS 2.1 for the period 1901-2002 (Mitchell and Jones, 2005 , and the VASCLIMO data (Beck et. al., 2005) , which contain only precipitation data for the period 1951-2000. Both data sets contain monthly averages. The VASCLIMO data were subject to a more rigorous quality control and can therefore be used to improve the CRU data, where only information from the period 1951-2000 is requested. At the moment, these two data sets represent the most reliable long-term rainfall information despite the problems with data coverage described above. Fig.1: SYNOP Stations in Morocco and the western part of Algeria. Circles represent the locations of the stations, numbers the observed 1961-1990 mean annual rainfall sum in mm. Fig. 2: Annual mean rainfall in mm (isolines with grey shading) for the period 1961-1990, as presented by the VASCLIMO data (left panel) and the CRU TS 2.1 data (right panel). The numbers in white rectangles show mean annual rainfall sums in the same period, but calculated directly from observations at SYNOP stations. We begin with looking at the mean annual rainfall for the so-called climate normal period 1961 -1990 . In order to focus on the effect of orography and missing data, the annual mean rainfall is shown for a region centred on the High Atlas, which contains the area with more observations in the north and the area with hardly any observations in the south-east. Observations at stations were plotted as coloured circles in the map and allow a comparison with the interpolated products. In general, the CRU and VASCLIMO coincide very well with the observations, but in regions with less dense stations we can find larger differences. Interestingly, none of the data really represent the high precipitation in the Mountain region. The CRU data seems to accentuate the orographic effect on annual precipitation a little stronger, but overestimates the rainfall south of the Atlas. In the following, we have to keep this uncertainty in mind when looking deeper into the characteristics of climate-related parameters. Climate variability affects vegetation in both natural and agricultural environments. Therefore, our second view focuses on the well known Köppen climate classification, which is based on thresholds relevant for special vegetation types. We want to demonstrate the climate shift in the late 20 th century by comparison of climate classes obtained from a reduced version of Köppens classification scheme, which has already proven its practicability in other applications (Guetter and Kutzbach, 1990 , Fraedrich et al. 2001 ). Fraedrich et al. (2001 have shown that the -in a statistical sense -optimal length of time spans for detecting changes in the Köppen classification is 15 years. Therefore, classification is applied to the 15-year periods 1951-65 and 1986-2000. Fig 3 shows the classification results for the entire Mediterranean basin. In northern Africa we can clearly see a shift towards dryer and warmer climates; at the borders to Steppe and Desert a number of pixels shift from moderate and summer dry (Cs) to Steppe climates. The Köppen classification is quite rough and only represents a very limited number of climate classes. Especially in the Cs region, a finer distinction is desirable. Fig. 3: Climate classification after a reduced Köppen scheme applied to the CRU TS2.1 data. The maps compare the climate classes for 1951-1965 to 1986-2000 and reveal a trend towards dryer and warmer climates in the second half of the 20 th century. Table 1: Definition of classes of the reduced Köppen climate classification. T is the mean monthly temperature in 2m hight above ground, Prec is the annual precipitation sum. Max / Min T indicate the warmest and coldest month in the mean annual cycle. Name Climate Criterion 1 Criterion 2 E Ice Max T < 10°C D Snow Max T > 10°C and Min T < -3°C Cs Moderate -3°C < Min T < 18°C summer dry Cf Wet Cw winter dry Af Tropical Min T > 18°C Wet Am Monsoon climate (dry period compensated by seasonal rain) Aw/s winter/summer dry BSk Steppe {Mean T} < {Prec} < 2 {Mean T} cold (Mean T < 18°C) BSh warm (Mean T > 18°C) BWk Desert cold with Mean T < 18°C BWh warm with Mean T > 18°C The classification can be improved by merging orographic information and climate parameters statistically. As an example, we computed an aridity index by integrating the area between the mean annual march of temperature and rainfall in the Walter-Lieth diagram. The orographic information was included by a multiple linear regression of surface height and exposition (slope and direction) with rainfall and temperature. Then we divided the aridity index into eight classes and obtained the classification shown in Fig. 4 . The Walter-Lieth climate diagrams according to the climate zones are also depicted. Here, the mountain climates and the northwest-southeast aridity gradient emerges clearly -also in regions where the Köppen classification showed only the Cs climate. It has to be pointed out, however, that this purely statistical proceeding does not contain any information of regional and local scale weather systems of the mountain areas, which we expect to influence the near-surface climates. Ignoring these effects probably explains the overestimation of zone 4 south of the Atlas mountain ridge. The third view focuses on the temporal climate variability. For this purpose, we look at the original observational data of SYNOP stations and combine them into regional indices. Since station-to-station differences in rainfall can be very large, we have first divided the monthly rainfall sums into 5 classes by separating the probability distribution -estimated simply by histograms -into quintiles, as recommended by the WMO No. 100 (1983). This way, each rainfall sum can be assigned to a number between 0 (no rainfall) and 5. These numbers are then averaged over regions with similar temporal behaviour of the rainfall (Fig. 5a) . The regions were chosen with respect to typical rainfall bearing systems (Knippertz et al., 2003) : The Atlantic region (ATL) is mainly affected by synoptic systems of the midlatitudes, has a seasonal rainfall maximum in the winter months (DJF) and is connected largely with the North Atlantic Oscillation (NAO). The Mediterranean region (MED) is also affected by Mediterranean pressure systems. Both regions are subhumid. The region south of the Atlas (SOA) is less affected by North Atlantic weather variability and gains rainfall relatively often from tropical-extratropical interactions. We have to notice, that the ATL region is represented by a larger number of stations -typical 25 -whereas the MED and SOA regions only contain about 5 stations. In Fig. 5b , time series of rainfall indices for the winter months (Nov-Apr) are compared to CRU time series. The series are plotted as standardized anomalies to the 1961-1990 period. We can see, that for the ATL region winter rainfall variations (blue bars in the plot of the CRU data) are quite good in phase with the observations, but for the MED and SOA regions we can see larger differences. In this case, we tend to value the direct observations higher than the interpolated CRU product, because the latter contains also information from far away stations due to the employed statistical interpolation technique. The time series show periods of wet and dry years, which do not seem to be entirely random, but clustered in periods of 2-4 years. Fig. 4: Climate zones based on an aridity index computed as the aggregated difference of twice the temperature in °C and the rainfall for the averaged seasonal cycle. The Walter-Lieth diagrams (bottom) are representative for the zones. They show the annual cycle of temperature (black lines, right axis) and monthly rainfall sums (grey line, right axis). The numbers printed each diagram contains the average surface height, annual rainfall and annual mean temperature of each zone. The numbers in the grey scale bar are aridity indices for the zones. The data are collected from the 1961-1990 climate normal period. (a) (b) Fig. 5: (a) The Atlantic (ATL), Mediterranean (MED) and South-of-the-Atlas (SOA) regions, for which stations data were aggregated. (b) Rainfall indices based on quintiles for station data (left panels) and CRU TS 2.1, for the ATL, MED and SOA region, respectively. The left series are aggregated for the rainfall season (Nov-Apr) and correspond to the grey bars in the right panel. In addition, the graphics show filtered values (lines) and numbers of available stations. Concluding Remarks Utilizing observational data we took three different views on observed Moroccan rainfall climate. We have to admit that the quality and density of available observations causes a large range of uncertainty especially in the region south of the Atlas Mountains that is subject to large risks due to climate shifts. We can analyze the temporal variability of rainfall on long time scales and are able to identify periods of drought and of wet years. The Köppen climate classification clearly reveals the climate shift towards warmer and dryer climates in the 20 th century. Now, we have to walk a step ahead and ask for possible future developments.
doi:10.1007/978-3-540-85047-2_4 fatcat:7r34jrdkc5ctngqo44uo3ps52e