The role of bedrock geology, historical rainfall pattern and topography on rainfall-runoff function in mountainous catchment
Abstract:
INTRODUCTIONStreamflow generation is a one of essential issues in hydrology and has been debated more than 100 years [e.g., Beven, 2001]. Especially, clarifying first-order (main) control of rainfall-runoff response is one of key efforts in this field. So, many theoretical, experimental and modelling studies have been conducted. Through these efforts, it has been shown that a variety of environmental variables controlled rainfall-runoff responses and the system coverts from hyetograph to hydrograph should be very complicated. However, hydrologist still argued prediction methods about runoff responses in ungauged basin,
Intersite comparisons might be a possible approach to clarify first-order control of rainfall-runoff response, since it has been difficult to derive general hydrologic principles from single research studies at intensively studied small basins. Although intersite comparisons have been conducted in several decades [Bosch and Hewlett, 1982], intersite comparisons still gave us new information about runoff generations in recent years [e.g., Uchida et al., 2005; Ali et al., 2013].
However, to clarify first-order control of rainfall-runoff response, three issues have to be considered. First issue is a removal of rainfall effects. Many observations showed that runoff characteristics, such as peak flow, runoff ratio etc., were strongly controlled by rainfall magnitude and patterns [e.g., Tromp Van Meerveld and McDonnell, 2006]. So, we have to remove effects of rainfall magnitude and patterns on streamflow to characterize rainfall-runoff response function of a given catchment. Second issue is related to the first. To remove rainfall characteristics effect, we have to collect streamflow data under a variety of rainfall conditions for each site.
Third issue is a diversity of site characteristics. Many environmental parameters has been proposed as a potential factor controlling rainfall-runoff responses. For comprehensive examination of roles of environmental factors, we have to collect rainfall-runoff data from a variety of catchments.
However, most of intersite comparisons did not considered all of these three issues. Here we completed intersite comparisons satisfing these three issues to discuss about first-order control of rainfall-response function in headwaters..
Fig. 1 Structure of tank model
MATERIALS AND METHODS
Dataset
We used Suimon Suisitu database managed by Ministry of Land, Infrastructure, Transport and Tourism of Japan. We focused headwater catchments and collected gauging station data where drainage area was smaller than 100 km2. Also, if the observation periods were less than two years, we did not include those gauging station in our analysis. Further, if it can be thought that human impacts, such as urbanization, water use for agriculture, flood control by reservoir etc., on stream flow was large, we removed these gauging stations from our analysis. Finally, we collected 164 gauging station data in Japan. Also, we used rainfall data of Automated Meteorological Data Acquisition System (AMeDAS) operated by the Japan Meteorological Agency (JMA). We collected 1017 rain gauge data. We used the 50-m digital elevation model supplied by the Geographical Survey Institute of Japan for topographic analysis. We also used Seamless Digital Geological Map of Japan (1:200,000) published by Geological Survey of Japan to characterize geological settings.
Characterization of rainfall-runoff response
A variety of indexes and methods, i.e., direct runoff ratio, flow rating curve etc., were proposed to characterize runoff response. In this study, to remove effects of rainfall patterns, we used four tanks model as shown in Fig. 1. We selected 12 storms for calibration and other 4 storms for validation for each site. First, we classified storms into four groups in terms of total rainfall amounts (i.e., 50-100 mm, 100-200 mm, 200-300 mm and more than 300 mm). Then, we randomly selected 3 storm, two for calibration and one for validation. However, since some stations cannot get data for large storms, we selected smaller storm data for these sites. We used SCE-UA method to calibrate parameters. To define initial condition, we assumed that water storages in the upper three tanks were zero. We used rainfall data in a given basin and surrounding stations to calculate spatially averaged intensity in the catchment.
Table 1 Number of catchment Geological classification
Fig. 2 Examples of preliminary analysis for clarifying degree of contribution of each parameters on result (a) Z13 (b) B2
Table 2 Tested environmental parameters
Data preparationWe calculated flowpath length from each grid cell to the gauging station. We used D8 method to define flow direction. Then, we classified grid cells into two, hillslope and channel. For classification, we used the relationship between local slope angle and drainage area as similar to Montgomery [2001]. Then, we calculated mean hillslope length and mean flowpath, including both hillslope and channel, length and coefficient of variation of hillslope length and flowpath length.
Also, catchments were classified in terms of rock types (sedimentary, volcanic and metamorphic/plutonic rocks) and geological age (Paleozoic/Mesozoic, Tertiary and Quaternary). We defined a predominant geology for each catchment using Seamless Digital Geological Map of Japan, Results of our classification are shown in Table 1.
Multiple linear regression analysis
In this study, we used multiple linear regression analysis for clarifying roles of topography, geology and historical rainfall pattern. We conducted preliminary analysis for clarifying degree of contribution of each parameter on result (Fig. 2). If a given parameter was sensitive to results, likes Z13 (Fig. 2a), we considered that the parameter was important for describing rainfall-runoff response. Finally, we selected important seven parameters, A12, A13, A2, B1, Z12, Z13 and Z2. So, we used these seven parameters as objective variables of the analysis. Models were determined by Stepwise method and using Akaike's Information Criterion. We tested roles of 14 environmental variables (Table 2). Dummy variable was used for test of geological effects.
Fig. 3 Results of validation ((a) Runoff ratio, (b) Peak discharge)
RESLUTS AND DISCUSSIONS
Most of models calibrated using 8 storm data generally predicted the runoff ratio and peak discharge in four other storms well (Fig, 3). So, we used all of calibrated parameters for multi-linear regression analysis.
Fig. 5 shows evaluated standardized partial regression coefficient (hereafter refer to SPRC) for geological variables. The SPRC of A13 in sedimentary rock catchment was negative, while Z13 was positive. Also, SPRC of A13 and A12 in volcanic rock catchment were negative, while Z13 and Z12 were positive. These indicate that rock types gave a impact on runoff responses. Moreover, it can be considered that the runoff response become gentle with the increase of Z12 and Z13 and decrease of A12 and A13. So, the runoff response in sedimentary rock and volcanic rock catchment should be gentle, compared to the catchment underlain by metamorphic/plutonic rocks.
While, geological age information was not selected for describing most of important parameters in the tank model. This suggested that geological age might have small impacts on runoff generation in headwater catchment.
Fig. 5 Standardized partial regression coefficient of geological information
Fig. 6 Standardized partial regression coefficient of geological information
The SPRC of A13 and A12 were negative for mean total flowpath length, while Z13 and Z12 were positive (Fig. 5). This indicates that the runoff response becomes gentle as the mean total flowpath length becomes large. This result agrees well with many previous studies. For example, Dunne [1978] indicated that the peakdischarge decreased with the increase of drainage area, if the dominant runoff generation process was similar. Except for mean total flowpath length, the topographic parameters were not selected to describe tank model parameters. This result does not concur with previous studies. A number of previous studies focused on the effect of flowpath gradient and CV of flowpath length [e.g., Gregory and Walling, 1973]. Our results suggest that roles of flowpath gradient and CV of flowpath length to define of rainfall-runoff response function was relatively small, compared to other environmental parameters, likes rock type and drainage area.The SPRC of A13 and A12 were positive for mean annual maximum daily rainfall, while Z12 was negative (Fig. 6). This indicated that the runoff response in streamflow becomes sharp, as mean annual maximum daily rainfall increases. Perhaps this relation can be explained by development of effective drainage pathway under wet conditions. So, Uchida et al. [2001] reported that water flow in soil pipes contributed to rapid drainage and soil pipes was easily developed by water flow in soil pipes. This suggested that the soil pipe developed easily at the hillslope where pipeflow occurred frequently. Other climatic parameters gave a small impact on the rainfall runoff response.
A WAY FORWARDED
We found several roles of geology, historical rainfall and topography in rainfall-runoff response functions in mountainous catchment:
Rock types controlled runoff-response function, but roles of geological age were unclear.
Except for mean flowpath length, roles of topography were unclear.
Runoff response become sharp as the increase of annual maximum daily rainfall, suggesting high drainage capacity in wet hillslopes.
However, multiple regression coefficients for seven important tank model parameters ranged from 0.3 to 0.5. This indicated that tank model with our fitted parameter could not fully explain rainfall-runoff response functions. It can be thought that these are several potential issues. For example, liner regression might be too simple to describe complex system Perhaps we might have missed several environmental parameters.
References
