Using Subgrid Parameterisation and a Forest Canopy Climate Model for Improving Forecasts of Snowmelt Runoff

Water Freedom System

Survive Global Water Shortages

Get Instant Access

ULRICH STRASSER1 AND PIERRE ETCHEVERS2 1 Department of Earth and Environmental Sciences, Section Geography, University of Munich, Munich, Germany, 2 Centre National de Recherches Meteorologiques, Centre d'Etudes de la Neige, METEO-France, Saint Martin d'Heres, France

4.1 INTRODUCTION

Mountainous catchments are the origin of many large rivers and a major source of water availability. They not only are a local resource for freshwater supply and hydropower generation but also considerably influence the runoff regime of the downstream rivers. The increasing needs for a sustainable management of river water resources and the demand for effective flood protection force to compromise between water exploitation and conservation and require a comprehensive knowledge of the dynamics of mountainous river basins. The latter is particularly important for such basins, which are dominated by perennial snow cover and glacierised areas, where spring floods induced by snowmelt usually evolve very quickly and can have disastrous effects, both for the environment and population of the downstream regions. For the quantification and prognosis of such snowmelt induced floods, forecast systems can be set up, consisting of meteorological prognoses coupled with a set of hydrological models to describe the relevant processes that govern the runoff production with proper forecasting accuracy and horizon (e.g., the European Flood Forecasting System (EFFS): http://effs.wldelft.nl).

A number of approaches exist for the spatial discretisation of a catchment to quantify the snowmelt runoff component, the largest surface water input controlling runoff during the melting season: for example, the various versions of the Snowmelt Runoff Model employ a segmentation of the catchment into elevation bands in combination with the areal depletion curve concept (Rango and Martinec 1995). Bell and Moore (1999) provide a detailed discussion of an elevation-based snowmelt model and discuss the number of bands that would be most efficient. Besides the elevation class-based approaches, a variety of subgrid parameterisations have been developed recently to deal with the issue in a more continuous manner (Liston 1999, Luce et al. 1999). All those studies show the importance of the orography for snowmelt modelling, with the vertical domain being most significant for the spatial discretisation in mountainous catchments (Braun et al. 1994).

A practical way to investigate the resolution impact on snowmelt simulation quality and to detect scale effects consists in using the same model at different resolutions and comparing the results with observations and/or with the highest resolution simulation, considered as the most accurate. Habets et al. (1999) compared two simulations

Climate and Hydrology in Mountain Areas. Edited by C. de Jong, D. Collins and R. Ranzi © 2005 John Wiley & Sons, Ltd of a large basin with different resolutions (8 • 8 km2 and 128 • 128 km2). The study shows that results are similar for both resolutions in flat valley areas, but they are very different for the alpine section of the basin. In particular, the simulation of the snowpack evolution is not realistic for spatial elements with a size of 128 km. This result was confirmed by Etchevers et al. (2001a), who simulated the waterflows in the Durance catchment with three different resolutions. For the coarsest resolution (46 • 46 km2), the annual surface water fluxes are well calculated, but the monthly partitioning is not correct. Again, the main cause is the poor simulation of the melting snowpack. With an improved resolution (8 • 8 km2), the snowmelt and river discharge is better simulated; only the flood peak in spring is still overestimated. For the finest resolution (1 • 1 km2), this systematical bias is corrected: as the variable altitude of the snow line is modeled more realistic, the melting water reaches the river more gradually and the simulated flood intensity is lower. However, computational requirements are increased by a factor of 64.

Besides elevation, aspect might also be a determinant factor to explain the snowpack evolution in alpine watersheds, but only at a small scale where its effects do not statistically compensate. The same accounts for snowdrift as a source of snowcover variability. For coarse resolution, the most promising strategy to improve the prediction of the snowmelt process dynamics is the consideration of the snow-vegetation interaction.

The goal of this study is to investigate whether a suitable compromise between accuracy and computation time requirements exists for snowmelt simulations and predictions on a regional scale. We apply the ISBA-CROCUS modelling scheme to the upper Durance catchment (Rhone-Alpes/France) at two spatial resolutions (1 km and 8 km) and investigate methods of subgrid parameterisation that can be applied at the coarse 8-km resolution and that lead to improvements in simulations. The following subgrid parameterisation methods are explored: the first is a technique that utilises the high-resolution (1 km) elevation data to derive subgrid information for topography in the 8-km model cells. In principle, this approach is based on the probability distributed principle (Moore 1985): three subareas that are not necessarily coherent are derived for each grid cell by statistically adapting the specific topographical variability. The altitudes of the subareas are used for interpolation of the meteorological variables. The second subgrid parameterisation is a forest climate model to include the effect of a forest canopy on the meteorological conditions that affect the snowpack. Both these subgrid parameterisations do not affect computation time since they are applied to the meteorological variables that are provided offline by the meteorological model. Thirdly, the parameterisation of subgrid topography is combined with the forest climate model.

In all cases, we investigate the effects on daily discharge hydrographs and the mean annual water balance in the basin. The proposed methods are transferable and can be used for improving operational snowmelt flood simulations and forecasts.

4.2 THE MODELS ISBA AND CROCUS

4.2.1 The SVAT model ISBA

The ISBA soil-vegetation atmosphere transfer (SVAT) scheme was developed for the Global Circulation Model (GCM) and Numerical Weather Prediction (NWP) model of the French Weather Service Meteo-France (Noilhan and Planton 1989, Noilhan and Mahfouf 1996). ISBA calculates the energy balance of the surface (bare soil, vegetation and snow) on the basis of the force-restore method and has six prognostic variables: the soil water contents of the surface, root zone, deep soil and interception reservoirs, and the surface and deep soil temperatures. The deeper soil layers can feed the root zone by capillary rise, and only the water of the root zone is directly available for transpiration. Two parameterisations are particularly adapted for hydrological purposes: a subgrid surface runoff parameterisation permits the model to take into account the fraction of the cell where the soil surface is saturated, and a minimum base flow parameterisation for dry soil conditions is considered, which permits the model to simulate very small discharges. Both these parameterisations are treated uniformly for the whole watershed (Habets et al. 1999). The parameters in ISBA were calibrated by Etchevers etal. (2001b) and left unchanged for all model runs.

ISBA calculates surface runoff every 5 min. In this study, the simulated runoff is aggregated to daily values for direct comparison with measured discharge. This direct comparison (without any routing) is sufficient because owing to the large relief and the relatively small size of the catchment the time of concentration is much smaller than one day (Etchevers et al. 2001a).

4.2.2 The snow model CROCUS

CROCUS is a one-dimensional snow model initially developed for avalanche risk forecasting; it simulates the evolution of the snow cover characteristics as a function of the meteorological conditions (Brun et al. 1989, 1992). The model considers the internal state of up to 50 layers of the snow pack (parallel to the surface slope) by calculating their temperature, liquid water content, density and snow type using time steps of 15min. CROCUS takes the following phenomena into account: energy exchanges between the layers of the snow pack and at its interfaces with the soil and the atmosphere, absorption of solar radiation with depth, phase changes, water transmission through the snow pack, mass exchanges due to precipitation and liquid water runoff, compaction and metamorphism of the snow.

The melt rates simulated with CROCUS are routed to ISBA and treated as precipitation falling on that part of the model cell that is not covered with snow. The temperature gradient at the snow-soil interface drives the conduction flux. No direct interaction between snow and vegetation is modeled.

4.3 THE UPPER DURANCE CATCHMENT AND UTILISED DATA

The upper Durance catchment is situated in the southern French Alps (Figure 4.1). Some of its characteristics can be depicted from Table 4.1. Because of the volcanic origin of the subsurface, the water tables are shallow and storage capacity that contributes to summer river discharge is very limited. The climate of the region is mostly dominated by a Mediterranean influence: precipitation occurs mostly in autumn (generally as snow) and due to severe storm events during spring. The relatively small total amount of precipitation and the high average altitude of the catchment are the reason for the comparably small forest coverage of only one-quarter; about half of the catchment is covered by grassland and the remaining quarter is in the high mountain environment with rocks and snow. Besides the Guil, the main tributary

Table 4.1 Basin characteristics of the upper Durance catchment

Name of the basin

Durance at La Clapiere

Mountain range

Rhone-Alpes/France

Elevation range (m)

787-4102

Latitude/Longitude

45°N, 6.5°E

Area (km2)

2170

Glacierised area (km2)

2

Forested area (km2)

26

Dominant vegetation type

Alpine pasture, larch forest

Geology

Limestone and crystalline

Mean discharge at outlet (mm)

713

Mean precipitation (mm)

1064

Mean evapotranspiration (mm)

348

of the Durance, there is no other significant river in the watershed. Downstream of La Clapiere is the Serre-Poncon dam, the largest reservoir in France, which is managed with a multi-purpose objective: hydropower generation, water supply for irrigation and recreation.

Was this article helpful?

0 0

Post a comment