AB24 3UF Scotland UK

16.1 INTRODUCTION

Floods are generated in mountainous basins by the complex interplay of precipitation input (i.e. rain as well as snow and ice melt) and temporally variable or constant basin characteristics (e.g. topography, soils, drainage network etc.). Therefore, firstly, for flood modelling, the understanding of flood runoff generation processes is crucial. In the last decade, numerous field investigations were performed, which were often coupled with model developments and applications. These studies focussed on identifying processes at the plot scale (e.g. Mosley 1982; Hornberger etal. 1991; Faeh et al. 1997), the headwater scale (e.g. McDonnell 1990; Hinton et al. 1994; Grayson et al. 1997; Anderson et al. 1997; Uchida et al. 2002), and the catchment scale (e.g. Merot etal. 1995; Uhlenbrook etal. 2002). A collection of state-of-the-art approaches is given in McDonnell and Tanaka (2001) and Uhlenbrook etal. (2003a). Secondly, the runoff processes need to be conceptualised adequately in the applied hydrological model. Therefore, more physically based (e.g. MIKE-SHE, Refsgaard and Storm 1996; WASIM-ETH, Schulla 1997) or more conceptual (e.g. TOPMODEL, Beven etal. 1995; HBV, Bergstrom 1992) model approaches were developed, which work in a distributed or semi-distributed manner. Each modelling approach has its strengths and shortcomings in consideration of the costs, applicability, and uncertainty, as discussed in several papers in Beven (2002). An overview of current process-oriented model developments is also given by Uhlenbrook et al. (2003a). Thirdly, the required input data for the model needs to be in an appropriate spatial and temporal resolution. Regarding rainfall input, this is difficult, in particular in mountainous basins, as precipitation is often very localised (e.g. Woods etal. 2000) and diverse along altitudinal gradients (e.g. Gurtz et al. 1997).

For flood modelling in heterogeneous catchments with an inadequate consideration of rainfall, variability can cause significant errors (e.g. Milly and Eagleson 1988; O'Loughlin etal. 1996; Koren etal. 1999). In many studies, an improvement of rainfall-runoff modelling owing to the use of distributed rainfall input data is shown (e.g. Schilling and Fuchs 1984; Quirmbach et al. 1999; Sun etal. 2000; Boyle etal. 2001). The increase of simulated runoff volumes with stronger consideration of rainfall heterogeneity was reported, for example, by Michaud and Sorooshian (1994) for a semi-arid 150-km2 watershed and Winchell etal. (1998) for a medium-sized catchment with intense convective summer storm events. In contrast, Faures et al. (1995) found for a small catchment of some hectares a decrease in simulated

Climate and Hydrology in Mountain Areas. Edited by C. de Jong, D. Collins and R. Ranzi © 2005 John Wiley & Sons, Ltd runoff volume with increased rainfall disaggregation. This indicates the dependence on the number and location of the rain gauges and the implications on the calculated basin precipitation for an investigated event. However, with increasing catchment size, an increased heterogeneity of precipitation can be observed, and the goodness of simulated runoff depends largely on the representation of the spatial and temporal variability of basin precipitation (Arnaud etal. 2002). But this variability can be crucial also for small catchments of a few hectares, in particular, in semi-arid regions (Faures etal. 1995; Goodrich etal. 1995). Obled etal. (1994) found that runoff simulations were more sensitive to the spatial variability compared to the temporal variability of precipitation, whereas Krajewski et al. (1991) showed a higher influence of the temporal aspect compared to the spatial aspect. In general, the capturing of spatially and temporally limited rain cells is difficult even with a detailed network of ground stations (Terblanche et al. 2001). The partly contrary results of the various studies depend on the one hand on the investigated event and catchment characteristics, and on the other hand on the applied model and scale.

For linking rainfall and runoff responses, Woods and Sivapalan (1999) developed a method for analysing the spatial and temporal patterns for both processes. This approach allows assessing the importance of dominant runoff generation processes and their spatial distribution in a catchment and, thus, the necessity of considering the spatial and temporal variability in hydrological investigations, that is, modelling.

The application of rainfall radar data seems to be a suitable method for capturing detailed rainfall information for catchment modelling. In recent years, radar data were more and more used for hydrological applications, for example, Obled etal. (1991) for a mountainous watershed, Winchell et al. (1998) for a 100-km2 catchment with special focus on saturation excess and infiltration excess overland flow, Quirmbach et al. (1999) for a 4-km2 urban catchment, Lange et al. (1999) and Lange (1999) for a 1400-km2 arid catchment, Ogden etal. (2000) for a 25-km2 basin with high variability of land surface parameters and for convective storm investigations, and Carpenter et al. (2001) for a macro-scale watershed (4100 km2).

Radar systems do not measure rainfall intensity itself but the reflectivity of radar beams by raindrops. The reflectivity is strongly dependent on drop size and precipitation type. Thus, the parameters aZ/R and fZ/R from the Zradar/RZ/R relation, which relate the radar reflectivity and the precipitation intensity, can vary considerably (e.g. Smith and Krajewski 1993; Pessoa et al. 1993; Ciach and Krajewski 1999; Quirmbach et al. 1999; Lange etal. 1999; Haase and Crewell 2000). The applicability of rainfall radar is strongly dependent on topography, distance from radar device, characteristics of the radar and precipitation characteristics. For instance, the radar may detect echoes from non-precipitation targets' so-called ground clutters, or the occurrence of hail may make the use of radar data impossible. In addition, the intensity of echoes decreases with increasing distance from radar because of the expansion of the radar beam and attenuation effects, and mountains can block the radar beam (Andrieu et al. 1997; Creutin et al. 1997).

For a correct transformation of radar reflectivities into rain intensities, the adjustment of the signals using ground station data is necessary (Adamowski and Muir 1989; Winchell etal. 1998; Seo etal. 1999; Quirmbach et al. 1999). But the differences in data collection can be problematic. The radar measures stepwise, for example, every five minutes. In contrast, the rain gauges record normally cumulative values over longer time steps. Additionally, ground data represent rainfall at a certain point, while the radar data average the reflectivity of raindrops of an air volume over the ground.

There are two main foci during radar data adjustment: (i) For the volume adjustment, the totally recorded rainfall amounts of ground data and radar data are compared, and (ii) for the distribution adjustment, the temporal distribution of the intensities (rain vs radar data) is considered (Maul-Kotter etal. 2001). After the decision, if either the volume or the intensity distribution is used as target value, the adjustment of measured reflectivities can occur (i) by application of a standard Zradar/RZ/R relation with spatially and temporally uniform parameters (e.g. Sun et al. 2000; Ogden et al. 2000) or (ii) by using an event- and space-dependent Zradar/RZ/R relation (e.g. Smith and Krajewski 1993; Hirayama etal. 1997; Winchell etal. 1998).

Numerous studies examined essential improvements in radar data quality by attenuation correction, considering of drop size distribution or vertical reflectivity profile correction (e.g. Sempere Torres etal. 1994; Uijlenhoet and Stricker 1999; Grecu and Krajewski 2000). This information was not available for the radar product used in this study. Hence, the operational available radar data in this study could only be corrected by using the available rainfall information from the ground stations.

Applications of radar data in mountainous catchments with their special problems due to the influence of topography are published in various studies (Delrieu etal. 1999; Jasper etal. 2002). A main error source is the affection by beam blockage as shown for instance by Andrieu etal. (1997) and Creutin etal. (1997). They corrected beam blockage by using digital terrain models and vertical profiles of radar reflectivity variability. They recommended the use of S-band radar with a wavelength of 10 cm for minimising attenuation effects in Mediterranean regions with intense rain events. The nowcasting of precipitation on the basis of radar data in an Alpine region is examined by Mecklenburg et al. (2000). They found improvements by removing small-scale features by using larger tracking areas.

The central objective of this study is to investigate the significance of the spatial and temporal variability of convective precipitation for flood modelling in a mountainous basin. Therefore, the use of rainfall radar data in addition to a classical precipitation network for deriving basin precipitation is examined. The following questions are addressed in further details:

1. What might be the contribution of highly resoluted radar data to capture spatial and temporal variability of convective precipitation events in mountainous areas?

2. What is the significance of the distribution of basin precipitation during convective cells for flood modelling in mountainous catchments?

3. How can operational available radar data be used within a detailed hydrological model in an appropriate way?

16.2 MATERIAL AND METHODS 16.2.1 Study site

The study was performed at the meso-scale Brugga catchment (40 km2), located in the Southern Black Forest Mountains, southwest Germany (Figure 16.1, Table 16.1). It is a pre-alpine mountainous catchment with a mean elevation that amounts to 986m a.s.l. The bedrock consists of gneiss and is covered by a drift of glacial and periglacial origin with varying depths (0-10 m). Brown soils have mainly developed in this drift cover material. The morphology is characterised by

Figure 16.1 The Brugga basin with instrumentation network

Table 16.1 Basin characteristics of the Brugga basin and the sub-basin St. Wilhelmer Talbach

Basin properties

Name

Mountain range Elevation range Area Geology

Dominant vegetation type

% forested

Mean precipitation

Mean runoff

Mean evapotranspiration

Brugga

Black Forest Mountains 438-1493m 40 km2

Gneiss covered by drift Forest and pastureland 71

1750 mm 1195 mm 555 mm

St. Wilhelmer Talbach Black Forest Mountains 633-1493 15.2 km2

Gneiss covered by drift Forest and pastureland 73.4

1853 mm 1301 mm 552 mm moderate to steep slopes (75% of the area), hilltops and hilly uplands (about 20%), and narrow valley floors (less than 5%). The overall average slope is 19°, calculated with a 50 x 50m2 digital elevation model. Owing to the strong variability of elevation, slope, and exposition caused by the deeply incised valleys, the catchment is characterised by a large heterogeneity of all climate elements, in particular, precipitation. This causes spatially and temporally very different elevation-precipitation gradients within the basin and articulated luv-lee effects. Further details about the basin can be found in Uhlenbrook (1999).

0 0

Post a comment