Génération multi-sites de variables météorologiques horaires en zone alpine

Water resources management and hydrological risks are based, to a large extent, on our competence to generate relevant hydrological scenarios under present and future climatic situations. In alpine areas, characterized by highly complex topography, the availability of discharge data is generally very limited and the hydrological regime is frequently affected by hydraulic infrastructures (e.g. dams for hydro-power production). Flood scenario estimation is thus a complicated task; classical methods are badly adapted to such a context. The present work proposes a global concept based on the simulation of meteorological weather variables transformed subsequently in flood scenarios using a rainfall-runoff model. The meteorological scenarios are obtained from simulation using a stochastic weather generator conditioned by both atmospheric circulation indices and meteorological variables. Atmospheric circulation indices are derived from reanalysis data set provided by the National Centre for Environmental Prediction (NCEP) for the period 1982-2001. Appropriate circulation indices and meteorological series were selected by means of sensitivity analysis. The stochastic generator developed within the framework of this research combines two approaches widely used in recent prediction purposes. The first uses a set of generalized linear models – that extend the classical linear models – to explain regional and daily meteorological variables using the appropriate set of predictors. The second uses nearest neighbor method, based on the hypothesis that similar synoptic situations will produce similar meteorological situations, to disaggregate daily regional meteorological variables into hourly and locally distributed variables. The generation process was applied to the Upper Rhone River basin, an alpine catchment located in the Valais canton – Switzerland (upstream to Lake Leman) to generate a series of hourly meteorological scenarios (precipitation, temperature and lapse rate temperature) at multiple sites (48 raingauges and 11 thermal stations). The ability of the stochastic generator to reproduce statistics of key hydrological variables and extreme weather events was evaluated at multiple spatial (ranging from 100 to 5500 km2) and temporal (ranging from 3 h to 3 days) scales. These statistics are classical mean, standard deviation and lag-1 correlation coefficient for total and liquid precipitation, as well as temperature. Other associated variables, such as the snow/rain line altitude have been studied. In general, statistics related to meteorological scenarios show moderate to very good agreement between simulations and observations. For extreme events, distributions of annual maxima precipitations and of seasonal maximum and minimum temperatures are also well reproduced, even if results obtained at both hourly and locally scales show less good – but still satisfactory – performance than those obtained at both daily and regional scales. This is definitely not surprising since the calibration of the different model parameters has been performed at those latter scales. The meteorological scenarios are next used to simulate 20-year discharge time series at a number of discharge stations throughout the Rhône watershed using an appropriate rainfall-runoff model. The catchment was divided into elevation bands in order to be able to take into account the variation of temperature with altitude; this is of high importance, especially in areas showing highly complex topographical features, which is usually the case of the alpine watersheds. Here again, statistical properties of the hydrological scenarios show good accordance with observations, especially for flood events. Finally, a number of flood scenarios, corresponding either to floods localized in some specific sub-basins or to floods generalized over the whole basin, were extracted from the simulated discharge time series; they can be used to study the hydrological behavior of the basins submitted to various meteorological solicitations.

The effect of seasonal soil frost on the alpine groundwater recharge including climate change aspects

Daniel Bayard

In alpine areas, the snow cover plays an important role as a water reservoir. Water is stored as snow over the winter and released in spring, recharging mountain aquifers through infiltration. These aquifers are essential, especially for supplying water for human activities during dry seasons. Numerous studies have shown that locally soil frost can drastically reduce the water infiltration. However, we know much less about the hydrological impact of soil frost at a larger scale, in particular with regard to groundwater recharge. This was our motivation for initiating an extensive field experiment in the southern Swiss Alps. Several methods at different spatial scales were adapted to explore (a) the local effect of a partially frozen ground on the snowmelt discharge in an alpine area, (b) the key processes influencing groundwater recharge during snowmelt periods, such as snow cover and soil frost evolution, snowmelt water runoff types, (c) the large-scale effect of soil frost on the aquifer recharge, and (d) winter situations that are critical with respect to flooding. In order to take into account spatial, altitudinal and climatic differences, the field study was run for two winter seasons at Gd St Bernard (2500 m), a site with very high winter precipitation and strong winds, and at Hannigalp above Grächen (2100 m), a ski resort with a rather dry winter climate. The different components of the soil water balance (lateral runoff, deep percolation, liquid soil water content) were measured on delimited plots of 5 m2. Additionally, a dye tracer experiment visualized the snowmelt infiltration patterns into the frozen, respectively non-frozen ground. At Gd St Bernard, investigations of the liquid water content and the soil temperature were additionally conducted at two secondary sites, differing in their orientation, so as to gain a better insight of the spatial soil frost expansion. Finally, an analysis of the water-table variation at the village of Grächen (some 500 m below Hannigalp) during the snowmelt was carried out for a 10-year time period, including our experimental seasons. We complemented our field investigations by using a numerical soil-snow-atmosphere transfer model that simulates the seasonal development of the soil frost and the snow cover, as well as lateral and vertical water flow in the top soil. This model was used as an upper boundary condition to a 2-D groundwater model, to calculate the outlet flow at the base of a simplified alpine aquifer. The two winters investigated had contrasting meteorological conditions, which resulted in intriguing differences and similarities with respect to the soil physical conditions: the first one having a thick snowpack and hardly any soil frost, and the second one with hardly any snow until January and a deep and persistent soil frost. In spite of a very thick snowpack, the water balance measurements from winter 2000/2001 showed for both sites zero to low surface and subsurface flow due to the fact that soil frost was local and the soil permeability high. During the next winter, approximately 25% of the total meltwater run off laterally. The soil infiltration capacity was mainly reduced by the presence of a basal ice sheet, caused by the freezing of meltwater after a mid-winter snowmelt event. The field data were used to calibrate and validate a physically-based one-dimensional model. To allow an accurate parameter setting for the water repellent soil of Hannigalp, laboratory infiltration experiments on soil columns were additionally carried out. Using the resultant information, the model predicted correctly most measured soil parameters, in particular good matching was observed between measured and simulated snowmelt discharge. At Gd St Bernard, the simulated results were less satisfactory, especially during the first winter, when large discrepancies were noted between measured and simulated lateral runoff. Apart from the shortage in the soil parameter characterization, the deviation was a result of the simplified description of the experimental field. The model was, for example, not able to simulate surface runoff generated by a steep slope under unfrozen conditions, as the soil surface is considered by the model to be flat. Nevertheless, results were accurate enough to reproduce the occurrences of snowmelt events and the accumulated values of snow-melt discharge pathways under frozen conditions. Long-term weather data were used to simulate the soil frost depth at both sites. The simulation results showed that the probability of the occurrence of frozen soils was lowest on slopes with south or north exposure. On southerly exposed plots, the heat stored in the soil during the summer inhibited the development of deep and persistent soil frost, while most pore ice thawed due to underneath heating. On north plots, the building of soil frost was rare, as early snowfall insulated the ground from the atmosphere. However, during snow-poor winters, the frost penetrated deep into the ground, and the low underneath heat flux at that location hardly affected the frost depth until snowmelt. The snow depth and average soil temperature were hence the two components affecting the frost depth extent. The altitude soil frost variation was related to the spatial extent of these two components. The higher the altitude, the higher the precipitation and the lower the mean air/surface temperature. Simulating the whole Grächen recharge area, seasonal soil frost was rare between 1800 m and 2000 m, as the snowpack was thick enough to insulate the ground, and the soil was warm enough to inhibit deep soil frost. In lower areas, the shallow snowpack enabled the soil to freeze during most winters, whereas at higher areas, deep and persistent soil frost was simulated due to the low underneath heating. A 10-year water-table depth record was used to examine possible large-scale effects of seasonal soil frost. We noted that the water-table rise at snowmelt was lowest at the end of winters with extensive soil frost. However, such winters were characterized by low precipitation, and reduced snow cover. In contrast, winters with little soil frost were mostly snow rich. Consequently, only a reduced number of winters were directly comparable. Considering these winters, the reduction in the water-table rise at snowmelt due to seasonal soil frost varied only between 10 and 30%, as, during snowmelt, most meltwater was able to re-infiltrate into the very permeable ground in lower unfrozen areas. We may assume that a soil with a smaller infiltration capacity can increase the influence of the soil frost on the aquifer recharge. Finally, we investigated the effect of a changing climate on the hydraulic and thermic regimes at different altitudes in the area of Grächen. We assume an increase in the air temperature of 2°C generates deeper soil frost on lower areas, as the resultant smaller snow cover does not insulate the soil from the atmosphere any more. However, it does reduce the soil frost on higher, snow rich areas, due to the warmer air temperature. As a whole, minor changes were simulated on the mean groundwater recharge, as more/less meltwater was able to infiltrate in higher/lower areas. By increasing additionally the precipitation by 15%, the soil frost diminished slightly over the whole catchment. By combining both climate change scenarios, we may expect an increase in extreme events, as, on the one hand, the rainfall intensity increases, and, on the other hand, rain on snow events over frozen soils occurs more often in lower areas. In practice, this work shows the importance of better investigating the hydrothermal behaviour of alpine regions. Such a knowledge is necessary, as more and more efforts are put into managing efficiently, durably and globally the available water resources. This work stresses also the necessity to consider seasonally soil frost in alpine risk management. As an example, flooding is mostly a consequence of strong rain precipitation at high altitudes combined with heavy snowmelt. Obviously, the risk of extreme runoff events increases during frozen winters, when the soil infiltration capacity is further reduced by the soil ice. Also, it has been shown that unstable slopes are mostly related to specific hydrogeological situations, like strong precipitation, which induces a sharp increase in the underneath water circulation. The melt water in spring is hence a potential activating factor, and a frozen soil may reduce the risk of sliding, as less water infiltrates into the ground.

EPFL2003

Generation of meteorological scenarios from NCEP reanalyses. Application for the generation of flood scenarios for the Rhone upstream to Lake Leman

Abdelkader Mezghani

Optimal water resources management requires hydrological scenarios for the climate situation under consideration. These scenarios may be produced from meteorological scenarios thanks to an appropriate hydrological model. The Laboratory "Hydrology and Land Improvement" (HYDRAM) developed a combined downscaling model for the multisite stochastic generation of the meteorological variables required for the generation of such scenarios. The model combines a statistical downscaling model and a k-nearest neighbour resampling approach to generate hourly precipitation and temperature series from NCEP reanalyses. It was applied for the upper Rhone catchment. Observed statistics are well reproduced for both meteorological variables. Then it was used for the generation of a suite of flood scenarios at different hydrological stations of the studied catchment. The stochastic generator can also be applied to downscale climate experiments from global and/or regional climate models for future climate conditions.

2007

Improving Alpine Flood Prediction through Hydrological Process Characterization and Uncertainty Analysis

Cara Christine Tobin

Among the many challenges of Alpine flood prediction is describing complex, meteo-hydrological processes in a simplified, robust manner that can be easily integrated into operational forecasting. In this dissertation, improved methods to characterize these processes are developed and integrated into the hydrological modeling component of an operational flood forecasting system used in the Swiss Alps. Detailed studies are conducted to improve hydrological model inputs, processes and outputs. Improvements, detailed in four chapters of this thesis, address the overarching goal of this work – the reduction of flood forecasting uncertainty. The accuracy of flood predictions in Alpine areas is contingent upon adequate interpolation of meteorological forcings, which has significant impacts on discharge volumes and flood peaks. This thesis demonstrates an improvement in the interpolation of temperature and precipitation inputs using a robust variogram which considers anisotropy and using a geostatistical interpolation method to distribute inputs in space and time. Results show that using elevation as the external drift factor better describes orographically-induced precipitation and temperature gradients. Also, the consideration of anisotropy is integral in detailing spatial patterns of precipitation induced by storm advection. Hydrological flood forecasting in mountainous areas also requires accurate partitioning between rain and snowfall to properly estimate the extent of runoff contributing areas. Partitioning is improved in this work by using a new method to integrate Limited Area Model output. Unlike standard hydrological procedures in inferring snowfall limit estimates based on dry, ground temperature measurements, Limited Area Model output considers the vertical, humid, atmospheric structure in its snowfall limit calculations. In effect, this method provides good estimates of runoff contributing areas in the spring as evidenced by validation on discharge measurements and satellite images of snow coverage. Accurately describing snowmelt processes on a sub-daily scale is also of critical importance in Alpine flood forecasting. However, the complex topography of the study region has limited observations available for validation. This thesis presents the development of a new physically-based snowmelt method applicable to regions with limited data. This method uses only daily minimum and maximum temperatures to mimic the effects of radiation. A comparative analysis of snowmelt methods is validated with snow lysimeter data and with a unique, distributed meteorological dataset collected by a wireless weather station network. Results demonstrate that the new method is competitive with more complex snowmelt methods as shown by accurately reproducing diurnal snowmelt cycles. Conveying limits of certainty on flood prediction outputs to users is critical because of epistemic and aleatory errors inherent to environmental modeling. Due to the presence of these errors, the GLUE methodology and multi-criteria performance ideas have been adapted with a fit-for-purpose uncertainty estimation technique in the final part of this thesis. With this method, hydrological model parameters are constrained based on hydrograph behavior, with a particular focus on flood peak response. A key component of the technique is a visualization tool which shows acceptable ensembles of discharge with respect to individual and combined criteria. By integrating the aforementioned input and process improvements into the hydrological model, calibration achieves model outputs that capture observed river discharge. Also, the uncertainty associated with hydrological modeling output error is reduced. Findings of this thesis are applicable to operational flood forecasting in general and have proven utility in improving hydrological model predictions in mountainous regions. Due to the novelty of the developments in terms of new methods or the use of tools and data sources previously unexploited in flood forecasting, further testing of the improvements is recommended. Future research in quantifying the chain of uncertainty produced by combining probabilistic forecast inputs with the hydrological output ensembles is also critical when the improved flood forecasting model becomes fully operational.