Snow is a basic component of the Earth’s surface energy balance [1
]. Climate trends, especially short-term and long-term precipitation and temperature variations, condition the snow regime and the hydrological evolution of the river flow response at the basin and sub-basin scales. The link between both climate and flow trends is crucial in mountainous areas, where small variations in temperature, among others, can produce significant impacts on precipitation (which occur as either rainfall or snowfall), snowmelt, and sublimation/evaporation. This can consequently result in very different flow signatures [2
In recent years, different authors have concluded that rainfall patterns are changing all over the globe as a result of global warming [5
]. Moreover, the analysis of precipitation events shows an increase in their spatiotemporal variability during the last decades [9
]. These variations, together with the increase in temperature, affect both the amount and distribution of snowfall, especially in semiarid regions [12
]. The impacts of this climate variability on the snow and hydrological regime are more evident over semiarid high mountain regions (i.e., areas higher than 1500 m a.s.l.) due their particularly extreme conditions [13
], in which the high variability of the annual and seasonal climate regimes is usually propagated to and amplified by the river flow. This is the case in the Mediterranean mountain areas, where alpine and semiarid conditions coexist [15
The impact of climatic variability in the medium and long term (i.e., 20- and 100-year time horizons, respectively) hydrological regimes in such areas is being assessed from different approaches on different spatiotemporal scales [14
]. The development of local climate indicators from both climate and hydrological variables are key to addressing the economic impacts of the expected changes and to assessing decision-making processes [18
This work assesses the influence of future climate scenarios on the snow regime in the Sierra Nevada Mountains, an alpine climate region in South Spain, during the 21st century. The Sierra Nevada area is a national park and a biosphere reservoir, with altitudes ranging from 2000 to 3500 m a.s.l. The park is a representative example of snow regions in semiarid environments and is characterized by a high inter- and intra-annual climate variability that influences most of the hydrological processes on different time and spatial scales [22
]. For this reason, Sierra Nevada (Spain) is part of the global climate change observatories network [24
]. The future evolution of climate and its impact on snow occurrence and persistence not only affects the hydrological regime [23
] in the area but also poses a risk for the ecosystems associated with the snow domain and the numerous endemic species identified in the park [26
For this assessment, precipitation and temperature projections in this area from two future climate scenarios derived from a Global Circulation Model (CGM) were used to estimate the evolution of the snowfall regime on both annual and decadal scales in the 2006–2100 period. Additionally, specific torrentiality indicators, which are defined in Section 2
, were analyzed together with the spatial distribution of the resulting impacts.
3.1. Long-Term Precipitation and Temperature Analysis for Future Scenarios
represents the annual evolution of the average daily mean, maximum, and minimum temperature in the Sierra Nevada area for the future projections during 2006–2100, expressed as absolute anomalies from the mean value during the reference period (Table 1
). On an annual basis, an increasing trend can be observed in all variables for both scenarios analyzed. RCP 8.5 was the most adverse scenario in terms of warming, as expected. The mean temperature anomaly follows a slightly variable regime during the future period in both scenarios, being positive in all cases. The maximum and minimum temperature anomalies exhibit a highly variable regime, with negative values in both scenarios. This can also be observed in the decadal regime. The resulting trends are significant on a high level of confidence (99%) in all cases. Autocorrelation was found only in the case of the annual mean values of the daily minimum and mean temperature under RCP 8.5 scenario. Therefore, the resulting significance in the trend of these variables can be affected by this fact.
Negative anomalies for both the annual and decadal mean values of the daily maximum and minimum temperature can be found under the RCP 4.5 scenario, and they expand over the whole future study period, with some apparent stabilization of the decadal mean daily maximum and mean temperature during the last part of the period. This is not observed under the severe scenario of RCP 8.5, with anomalies close to 6° in both variables, on both the annual and decadal scales, at the end of the study period.
The annual and decadal mean daily minimum temperature anomalies show the highest variability in all cases. Under the RCP 8.5 scenario, this variable follows a trend towards prevailing positive values during the last decades of the study period.
shows the annual and decadal evolution of the average annual precipitation in the Sierra Nevada area for future projections during the study period from 2006 to 2100 expressed as absolute anomalies from the mean value during the reference period (Table 1
). As opposed to temperature trends, a decreasing evolution is observed in both scenarios. However, on a significant level, an annual trend is found only for the RCP 8.5 scenario, the decreased trend in precipitation of which is four-fold that found in the RCP 4.5 scenario. It must be noted that the mean annual precipitation in the period differs 4% between both scenarios, being 503 and 482 mm·year−1
for RCP 4.5 and RCP 8.5, respectively. However, a large difference is found for the associated standard deviation (i.e., 44.11 and 47.37 mm·year−1
for RCPs 4.5 and 8.5, respectively).
Decadal trends are not significant, however. In fact, the first three decades of RCP 8.5 and the first four of RCP 4.5 reflect an increasing evolution, which results in a non-significant trend (i.e., no trend) in RCP 4.5. Decadal anomalies are larger and more fluctuating in the case of RCP 8.5.
These results have a clear impact on the snowfall occurrence distribution, as described in the next section. This is particularly relevant in the highest areas, as Figure 5
shows, where both increased temperature and decreased precipitation mean values for the whole study period can be observed.
3.2. Long-Term Snowfall Analysis for Future Scenarios
Based on the previous results, daily snowfall occurrence was estimated for each scenario on a 30 × 30 m grid, as described in Section 2
. In this work, the results were aggregated on the annual and decadal scales and averaged over the different spatial units in this work. Figure 6
shows the spatial distribution of the anomalies of the mean annual snowfall in the study area during the future period under both climate scenarios in this work, together with the mean annual trends during the study period. The largest impacts can be found in the highest areas (i.e., mostly negative anomalies), being both larger and more extended in RCP 8.5, as expected. Only under RCP 4.5 did some areas in the west side of the Sierra Nevada basins exhibit positive anomalies, although these were much lower in absolute values than the negative ones. Trends are negative in all cases, and they are higher in the snow domain area.
shows the mean values of the variables under analysis during the study period, averaged over each region in Sierra Nevada and the whole study area (SN) under both future climate scenarios, together with the results of the trend analysis performed for the whole study period.
The average mean annual snowfall in the study period exhibits a large variation depending on the region and scenario under analysis. These average values range between 153 mm·year−1
, in Region 5 under the RCP 4.5 scenario, and 25 mm·year−1
, in Region 2 under the RCP 8.5 scenario. Figure 7
shows the evolution of the annual snowfall anomalies averaged over each region in Sierra Nevada (Figure 1
) and the whole study area during the 2006–2100 future period under both future climate scenarios. Decreasing trend values are found in all cases, with RCP 8.5 being the most impacting scenario on both the annual and decadal scales, as expected. Trends in the different regions in Sierra Nevada show similar values in a given scenario. However, these were more pronounced in R5, the most snow-influenced region in this area. Only the snowfall evolution under RCP 8.5 showed autocorrelation on an annual basis, with a 0.05 degree of confidence. Hence, the significance of this trend can be influenced by this fact. Nevertheless, the annual scale reveals a high variability of snowfall during the future period in all of the regions under both scenarios.
The decadal evolution of the mean annual snowfall for the future period under each scenario can be observed in Figure 8
as anomalies, averaged for each region in Sierra Nevada and the whole study area. The decadal regime more efficiently shows the decrease of snowfall associated with both scenarios in all regions, with the RCP 8.5 scenario (being the most severe) causing the highest impact in snowfall occurrence in all cases. Annual variability is large in all projected decades, with the anomalies in R2 being lower than those in the rest of the regions. However, this region is also the most semiarid area in Sierra Nevada, with an average mean annual snowfall of 44 mm·year−1
during the 1961−2000 reference period, which makes the impact of future climate scenarios highly relevant for all ecotypes associated with snow occurrence in this region.
Anomalies are positive in most regions for the first half of the future period under the RCP 4.5 scenario; however, they drop to negative values in all regions after three decades under the RCP 8.5 scenario.
shows the annual evolution of the Dsnowfall indicator defined in Section 2.5
during the 2006–2100 period for each region in Sierra Nevada (Figure1) and the whole study area. This variable is also related to the annual mean intensity of snowfall (Isnowfall) that is shown in Figure 10
. Again, it is illustrated that R5, with higher Dsnowfall values (ranging between 44 and 8 days/year under the RCP 4.5 scenario and between 38 and 4 days/year under the RCP 8.5 scenario), is the most snow-influenced domain in Sierra Nevada. The semiarid character of R2 is also highlighted (ranging between 26 and 1 days/year under the RCP 4.5 scenario and 18 and 0 days/year under the RCP 8.5 scenario). Significant decreasing trends (99%) are found for both scenarios in all cases; however, relevant differences are found between scenarios.
Regarding Isnowfall (Figure10), this indicator shows similar values in most of the identified regions in Sierra Nevada, with the exception of R2 and R3, which exhibit more torrential behavior. As previously stated, these regions represent the semiarid conditions in Sierra Nevada, and their features tend to be enhanced in the future climate scenarios, especially under RCP 8.5, as expected. All resulting trends are positive (RCP 4.5: 0.006 (mm·days−1)·year−1 and RCP 8.5: 0.01 (mm·days−1)·year−1), although quite moderate under RCP 4.5. The decadal scale more clearly confirms the semiarid character of R2, with high variability, as well as the snow-dominated regime in R5.
The analysis performed in the Sierra Nevada area provides an estimate of the long-term effects of climate evolution under two different emission scenarios in a mountain area representative of Mediterranean conditions. The results highlight how climate impacts on snow in Mediterranean regions can be significantly variable in short scale spatial domains [67
] and quantify the estimated anomalies in snowfall on both an annual and decadal basis.
The two scenarios chosen for the analysis largely differ in the extent of the estimated impacts of climate on snowfall occurrence and amount on an annual and decadal basis. As expected, the more severe scenario, RCP 8.5, results in larger impacts in terms of anomalies of snowfall. This has also been reported from similar analyses in different studies across the world [14
]. Despite the fact that the decreasing evolution of precipitation on an annual basis shows no significant trends, the resulting snowfall regime for each scenario follows a significant decreasing trend associated with the long-term increasing trend for temperature.
This general significant increasing trend of temperature in future climate scenarios has been reported in climate impact analyses in other high mountain regions, like the Alps and the Himalayas [73
], as well as in other semiarid mountain areas in Chile (the Andes) and in California (the Sierra Nevada) [77
]. Similar values of the estimated trends for mean temperature have been found for the Alps (0.22 and 0.58 °C·decade−1
for RCPs 4.5 and 8.5, respectively), Sierra Nevada (0.18 and 0.46 °C·decade−1
for RCPs 4.5 and 8.5, respectively), and Chile (where an increase of 3 to 4 °C for the 21st century has been shown) [77
]. The resulting trends in minimum temperature are lower, however, than those found in these other regions (e.g., 0.3 and 1.1 °C·decade−1
for RCPs 4.5 and 8.5, respectively, in the Himalayas and 0.08 and 0.26 °C·decade−1
for RCPs 4.5 and 8.5, respectively, in Sierra Nevada). This lower value is likely due to the different ranges of altitude between both sites, but scale effects associated with the downscaling in Sierra Nevada could also influence the results. In addition, maximum temperature trends show different behaviors between both areas (e.g., 0.20 and 0.30 °C·decade−1
for RCPs 4.5 and 8.5, respectively, in the Himalayas and 0.21 and 0.66 °C·decade−1
for RCPs 4.5 and 8.5, respectively, in Sierra Nevada).
The lack of significance in the trend analysis of precipitation has also been reported in these studies [80
], with higher decreasing trend values observed during the 21st century under semiarid conditions (e.g., −9% and −19% for RCPs 4.5 and 8.5, respectively, in the Chilean Andes and −1.5% and −5.5% for RCPs 4.5 and 8.5, respectively, in Sierra Nevada). These results also show how apparent decreases of snowfall are expected in all future scenarios [74
], with a higher loss of snowfall amount in areas above 1500 m a.s.l. but a larger impact in lower areas were snowfall might not even occur in the future [74
]. The resulting impacts on snowfall differs largely among sites. For example, decreasing trends of 32 and 65 mm·decade−1
were found in the Alps for each scenario [75
], which were larger than those found in Sierra Nevada. However, in relative values, these results are closer to the decadal decrease of snowfall estimated in Canada [85
], which are 5 and 10% of the current decadal values for RCPs 4.5 and 8.5, respectively. This is in line with the 2.8 and 6.1% obtained for the same scenarios in Sierra Nevada. A key result from this analysis is the likely impact of climate scenarios on the torrentiality of snowfall. Despite the mean annual trend being a decrease (with high significance) during the study period, the extreme annual values of snowfall within each decade show an extremely high variability, with large amplitudes under both scenarios in most regions. From Table 1
and Table 2
, it can be observed how the number of snowfall events in the future period is lower than that during the reference period, while the mean duration of each event is also lower. This behavior has also been reported in the Alps [76
], with the mean value of the number of days of snowfall in future scenarios being 50% that of the reference period. That is in the range of the 33% and 76% found in Sierra Nevada under scenarios RCP 4.5 and RCP 8.5, respectively.
This impact is highlighted by the resulting Isnowfall (Figure 9
), which increases during the second half of the study period and reaches the highest values in regions R1 and R2, which represent the semiarid environments in Sierra Nevada. This can also be observed from the results in Table 3
, which shows the proportion between the mean values for each variable over the whole future time period under scenarios RCP 8.5 and RCP 4.5 for each region in Sierra Nevada and the whole study area. Whereas the ratios of the maximum and mean temperature and precipitation are very similar among regions, the ratio of the minimum temperature shows clear differences that are associated with the frequency of snowfall in each region. This greatly impacts the snowfall under each scenario, with ratios that differ among regions for both the maximum and minimum snowfall and the associated trend, which is amplified in the more severe scenario. This is especially true for the most snow-influenced regions (R5, R4). The apparent stabilization of the snowfall anomalies during the last decades of the 21st century under RCP 4.5, compared to the continuous fall estimated under RCP 8.5, stresses the importance of mitigation actions against the adaptation actions in the climate impacts simulated by the model.
sums up the regional differences found in the study area for both the mean annual values of the targeted variables and their mean annual trends. It can be easily observed how temperature trends are spatially similar despite the fact that the mean original values are not homogeneous in space, whereas the precipitation pattern is variable both for the mean values and their trends. The resulting snowfall distribution is hence variable, both as mean and trend values.
The limitations of the results discussed above are mainly associated with the skill of the downscaled data and with the simplifying hypothesis in the snowfall/rainfall partition algorithm applied in this work. Additionally, new weather stations in the higher areas are required to further validate the spatial interpolation of both temperature and precipitation. Nevertheless, the results are aligned with those projected during the 21st century in other high mountain regions and provide the first distributed assessment of future climate scenario impacts on snowfall in Sierra Nevada, which has similarities with other semiarid mountain regions that allow some transferability to other less-monitored areas.
This work assessed climate impacts on the snowfall regime in a Mediterranean mountain area in southern Spain (i.e., the Sierra Nevada range) under two different future scenarios (RCP 4.5 and RCP 8.5). The results indicate a global decrease of annual snowfall, with a significant trend that ranges from 0.21 to 0.55 (mm·year−1)·year−1 and changes depending on the scenario and region in the area. However, the major impact of climate on the snowfall regime is reflected in the torrential character of snowfall occurrence, with a decrease in the number of days with snowfall and a significant increase in the global mean snowfall intensity under both scenarios. This torrentiality is more extremely increased in the semiarid region of Andarax in Sierra Nevada, the area currently less influenced by snow. This may have a relevant impact on the fluvial regime, which is currently characterized by non-perennial flows, with the occurrence of flash-flood events.
Climate projections are not forecasts, but rather likely long-term pathways that climate variables may follow as they are driven by the long-term dynamics of the global circulation system in the Earth under different emission scenarios. The uncertainties associated with the estimation of snowfall occurrence from the assumptions made in this work add up to uncertainties in the global models, the downscaling techniques, the spatial interpolation algorithms, and the observational datasets. Nevertheless, the thus-projected trends of snowfall in the Sierra Nevada area show how climate impacts on the snow regime in Mediterranean and other semiarid regions are highly dependent on the trend towards torrentiality observed in the precipitation regime in these areas. These trends focus on the extreme value regimes of both precipitation and temperature variables, rather than on the mean value regime. Moreover, the spatial analysis’ results highlight the large heterogeneity of these impacts on the local scale and point out the need for high resolution modeling to derive further information linked to the snow regime, such as runoff and river flow, soil wetness, vegetation distribution patterns, erosion rates, water temperature, and groundwater recharge.