Rainfall-runoff events in a middle mountain catchment of Nepal

Journal of Hydrology (2006) 331, 446– 458 available at www.sciencedirect.com journal homepage: www.elsevier.com/locate/jhydrol Rainfall-runoff events in a middle mountain catchment of Nepal Juerg Merz a,*, Pradeep M. Dangol b, Madhav P. Dhakal b, Bhawani S. Dongol b, Gopal Nakarmi b, Rolf Weingartner a a b PARDYP, University of Bern, Bern, Switzerland PARDYP, International Centre for Integrated Mountain Development, Kathmandu, Nepal Received 6 September 2005; received in revised form 14 May 2006; accepted 16 May 2006 KEYWORDS Summary The generation of runoff and the associated processes are important for the understanding of flood generation and sediment mobilisation. However, only few studies of this kind were conducted in the Hindu Kush-Himalayan region. This paper presents detailed rainfall event analyses in the Jhikhu Khola catchment in the middle mountains of Nepal’s Himalayas followed by analysis of runoff events in erosion plots on different land use. It shows that rainfall events in the catchment can be divided into four clusters: minor events, medium events, high intensity events and large events with each cluster having particular characteristics. Annually about nine high intensity events occur, most of them during the monsoon season. Large events, about one event per year, generally occur during the monsoon and the post-monsoon season. Both the high intensity and the large events are potentially important for the generation of floods in the catchment and beyond. Runoff events in the catchment are closely correlated to the event rainfall intensity parameters and the proposed clusters. Depending on land use another surface flow process is expected. While on degraded land infiltration excess flow is the key process in terms of runoff generation, on agricultural land saturation excess overland flow is more relevant. The bulk of the runoff is generated in a few major rainfall events. Particularly on agricultural land only few events cause the total annual runoff. c 2006 Elsevier B.V. All rights reserved. Rainfall events; Runoff events; Runoff generation; Cluster; Nepal; Middle mountains  Introduction Floods are an annually reoccurring issue particularly in the lowlands but also the inner valleys of the Hindu Kush-Hima* Corresponding author. Tel.: +977 15537279. E-mail address: jmerz@gmx.net (J. Merz). layas (Hofer, 1998; Chalise and Khanal, 2002; Agarwal and Narain, 1991). To understand the process of flood generation under different circumstances and conditions, runoff generation studies were and are being conducted extensively throughout the world (Pearce et al., 1986; Leibundgut et al., 2001; Uhlenbrook, 2005). This improved understanding aims in general to improve hydrological models for more 0022-1694/$ - see front matter c 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.jhydrol.2006.05.030  Rainfall-runoff events in a middle mountain catchment of Nepal efficient and more accurate flood estimation and forecasting. While the processes involved are widely accepted (Scherrer, 1997), the importance of the different mechanisms in flood generation in particular and their prediction and estimation are still subject to scientific discussion. Different studies have been conducted in New Zealand with a particular focus on the Maimai catchment on the South Island. Mosley (1981) described the importance of subsurface flow, previously believed to be of less importance in the generation of floods. He mainly held rapid throughflow through macropores responsible for this as was later also emphasized by Germann (1990). Pearce et al. (1986) favoured the theory of displacement of old water or piston flow effect to explain the rapid response through subsurface flow. Merz and Mosley (1998) show the impact of landsliding on the hydrological processes and runoff generation in the Tutira catchment of the North Island of New Zealand. The impact is mainly due to an increase in area of potentially saturated areas as well as loss of soil from impermeable areas and therefore loss in soil water storage capacity. In Europe a number of studies have been undertaken with a focus on runoff generation. In the Brugga catchment in southern Germany, Uhlenbrook et al. (2002) identified three main runoff components: direct runoff from saturated and impermeable areas, shallow groundwater flow with piston flow and groundwater ridging and finally deep groundwater flow with matrix flow. Results from the Leissigen catchment near Bern in Switzerland show that on the very wet areas saturation overland flow and in the other areas mainly matrix flow seems to be important (Laemmli, 2000). On the basis of rainfall simulation experiments, Scherrer (1997) showed the variability of runoff generating processes at different sites in Switzerland. However, Hortonian (or infiltration excess) overland flow at rainfall intensities of 50–100 mm/h occurred most often. Saturation overland flow only occurred in follow-up experiments. Lateral soil matrix flow and macropore flow was less important at the selected sites. Interesting results on runoff generation in forest areas are presented in Badoux et al. (2006): He tried to understand the process of flood generation in a nested approach combing different scales (from a plot to a catchment). The number of investigations of runoff generation in the context of the HKH region however is limited. Collins et al. (1998a) report that in the terraced land of the Middle Hills in Nepal both infiltration excess and saturation overland flow contribute to runoff generation. Merz (2004) showed that there is a distinct difference between the different land uses, where degraded land experiences some 20 times greater monthly runoff than agricultural land. On the degraded land the runoff corresponded roughly the seasonal rainfall pattern. It was also observed that monthly runoff coefficients greatly varied between the different land uses and that these coefficients peaked in the wet season on the degraded land, whereas on agricultural land they peaked during the pre-monsoon season. It is important to note that though there is a lot of information on runoff generation from Europe, America and the Pacific, this information is only to a limited extent applicable to the Asian conditions. In the Nepal middle mountains rainfall is governed by high seasonality with about 76% of 447 the rain occurring during the 4-month period of the monsoon season from June to September (Merz et al., 2006). Over the entire country rainfall ranges from 200 mm in the shadow of the High Himalayas to 5000 mm in the area around Pokhara (Chalise et al., 1996). In the middle mountains East of Kathmandu, where the study area of this paper is located, annual precipitation ranges from about 1000 to 3000 mm. From a runoff generation perspective, the impact of irrigation with prolonged saturation of large areas should also be mentioned. The time of this extended saturation coincides with the time of highest rainfall input, the monsoon season as well as the time with the most intense rainstorms. Saturation overland flow therefore is assumed to play a major role in the runoff generation process as also indicated by Collins et al. (1998b). The studies in the region focus primarily on surface flow processes, while sub-surface flow mechanisms are still to be investigated. In order to understand the runoff generation processes many studies have been conducted using event analysis (e.g., Mosley, 1979; Naef et al., 1986; Merz, 2004). In general these studies comprise of the rainfall event separation, identification of typical events of different magnitudes, the derivation of characteristic parameters and finally their statistical interpretation. In the case of this study the event analyses are done to answer the questions when and under what conditions runoff occurs as well as under what conditions important floods are generated. The event analyses here are done from two different perspectives in the Jhikhu Khola catchment (JKC). Firstly, precipitation event analyses investigate runoff triggering mechanisms. Secondly, erosion plot events are studied to investigate surface runoff generation. The main objective of the study is to investigate surface runoff generation processes at the plot level in relation to rainfall events in a mountainous area of the Himalayas. Study site and methodology The People and Resource Dynamics in Mountain Watersheds of the Hindu Kush-Himalayas Project (PARDYP) undertakes research for development in four catchments in China, India, Nepal and Pakistan in the field of watershed and natural resources management including water resources management and hydrology. The activities of PARDYP Nepal focus on the catchment of the JKC. The JKC is located in the middle mountains of Nepal (Fig. 1), about 45 km east of Kathmandu on the Arniko Highway. It covers 111.4 km2 in total area. Further important information on the watershed is compiled in Table 1. The measurement network consists of nine recording raingauges of tipping bucket measurement principle, some installed since 1993. In total seven sites have time series longer than 2 years. With the exception of two raingauges at sites 4 and 6 all raingauges have 0.2 mm/tip capacity and are connected to HOBO event loggers. The other two tipping buckets have a capacity of 1.0 mm/tip. Rainfall is logged at 1 min intervals which are later aggregated to lower resolution data. The orifice of all raingauges is located at 1 m above ground according to the standards used in Nepal. The data from the recording raingauges is cross-checked with the manually read standard raingauges of 1 day 448 J. Merz et al. Figure 1 Table 1 Location map with monitoring network of the JKC. Main characteristics of the JKC Parameter JKC 2 Catchment size (km ) Altitudinal range (masl) Physiography General aspect Annual rainfall (mm/a) at main meteorological stations (period) Spatial rainfall variability in the JK (mm/a) 111.4 800–2200 Flat valley bottom of alluvial origin; short and steep slopes on northern and southern margin South-east (main valley extending from north-west to south-east) 1167–1418 (1998–2000) at site 12 (865 masl) 1100–1700 (1998–2000) between 830 m and 1700 masl temporal resolution, which are installed in the compound of the same measurement site. The measurement period from 1993 to 2000 in the case of the JKC was normal in the context of the long-term records and therefore representative for the catchment. Statistically no difference between the long-term records of the Department of Hydrology and Meteorology, Nepal (DHM) at Panchkhal and Dhulikhel and the short-term records of the project at the same locations could be established (Merz, 2004). In terms of the maximum and the minimum annual rainfall the years of the study period were all within the range of the long-term records. In terms of findings of this study this suggests, that the results can be assumed to be representative for the conditions in the JKC. The season’s definitions used in this study are according to Nayava (1980). The main seasons experienced on the Indian sub-continent are southwest monsoon (June–September), post-monsoon (October–November), winter (December–February) and pre-monsoon (March–May). Carver (1997) defined for his erosion study a rainfall event on the basis of the storm separation time Smin (i.e., time between two distinct rainfall events without any rain) and a minimum precipitation amount Pmin. For the JKC and the years 1993–1995, he proposed to define an event with Pmin = 3 mm and Smin = 120 min. This Pmin value was chosen because storms below 3 mm are unimportant for sediment generation. One hundred and twenty minutes is derived from the investigation of the numbers of events with different Smin. A major change in the number of events could be observed between Smin = 60 min and Smin = 120 min. In terms of Smin the same could be shown for the data from 1993 to 2000 and therefore this value was used here as well. In terms of Pmin it was shown that the value of 3 mm is not appropriate for studying runoff generation. While on degraded plots already events with as little as 2 mm rain can produce runoff and sediment loss, on cultivated plots runoff generation only starts at more than 5 mm rainfall. For this reason Pmin was set to 2 mm and Smin to 120 min for the purpose of rainfall event separation. A rainfall event can be characterized with a variety of parameters. A selection of parameters as proposed by Mosley (1979) and Wuethrich (1999) and used in this study is given below. Magnitude and duration of the event: – Ptot (mm) rainfall amount during the event; – tP (min) rainfall event duration. Intensity of the event: – I10max (mm/h) max 10 min rainfall intensity during the event; Rainfall-runoff events in a middle mountain catchment of Nepal 449 – I30max (mm/h) max 30 min rainfall intensity during the event; – I60max (mm/h) max 60 min rainfall intensity during the event; – Iave (mm/h) average rainfall intensity during the event. runoff volume with the help of a calibrated conversion chart. The results from all drums are then summed up to determine the total event runoff from the plot. Shape of the hyetograph: Rainfall event description – P25 (%) rainfall amount after 25% of the event duration in % of total rainfall; – P50 (%) rainfall amount after 50% of the event duration in % of total rainfall; – P75 (%) rainfall amount after 75% of the event duration in % of total rainfall. The number of rainfall events varies considerably according to the number of observation years and missing data. Table 3 presents the summary of all sites and events. Most events were observed at site 6, where data is available since 1993. The minimum of 249 events was observed at site 12 during a 3-year study period from 1998 to 2000. This is followed by site 15, where data is likewise available since 1993. Site 16, which has data available since 1993, was later excluded from the following analyses as there were too many gaps in the automatic rain gauge data (519 days), for the rainy season in particular. In general it can be said that a large sample number is available which allows to make statistically valid observations. Annually there are on average about 86 rainfall events. Seventy-three percent of these events occur during the monsoon season and 20% of the total number of events in the pre-monsoon season. About 10–30 events at each site were also observed during the post-monsoon season. The rainfall events during the winter season in the study period numbered about 5–30. According to rainfall amount the most frequently occurring events are the 2–5 mm events accounting for more than 30 up to 45% of all events (Fig. 2). Due to this strongly left skewed distribution of events, the analyses below are all based on the median values of the distribution as proposed by Helsel and Hirsch (1992), as the mean does not seem to be appropriate for this purpose due to the strong maximum outliers. Results and discussion For runoff generation investigations eight erosion plots were established, but only four plots provide enough data for further statistical analysis at this point. The other four plots only provide up to 2 years data and are therefore omitted from the analyses. For the location of these plots refer to Fig. 1. Erosion monitoring plots with areas of about 100 m2 per plot (note that plot 6 is only about 65 m2; Table 2) were established on different land use. The plots consist of a (usually) 20 m long by 5 m wide area delimited by metal sheets. Runoff and eroded material is collected in a gutter system funnelling them to a first 200 l drum. To provide further storage volume other drums (often four) are placed in sequence. The second last drum has a 10-slot divider only allowing 1/10 of the runoff into the last drum. This allows to measure higher runoff rates with a lower number of drums. Runoff observations from these plots are taken by local observers after major rainfall events by measuring the water level in each drum. However, they visit the sites and make observations at 9:00 every day even if there is no rainfall in order to ensure clean and empty drums. Runoff volume is calculated by converting these water levels to a Table 2 Erosion plots in the JKC Site Site name Land use Elevation (masl) Aspect Orientation Plot size (m · m) Area (m2) Slope () 4 6 14 16 Baghkhor Bela Kubindegaun Bhetwalthok Degraded grass land (d) Rainfed terrace (a) Degraded (d) Rainfed terrace (a) 940 1240 880 1200 N N S S NE NW SW SW 19.8 · 5.03 13.61 · 4.54 20.62 · 5.02 12.64 · 7.66 99.6 61.8 103.5 96.82 11.5 24.7 15.0 6.7 Table 3 Rainfall events at selected sites (in brackets: Number of missing days; note that the missing station numbers relate to other monitoring sites than rainfall stations) Site Period Pre-monsoon Monsoon Post-monsoon Winter Total 3 4 6 12 14 15 16a 1993–1996 1997–2000 1993–2000 1998–2000 1997–2000 1993–2000 1993–2000 38 (0) 80 (2) 150 (10) 55 (26) 75 (45) 130 (4) 113 (115) 270 270 511 181 247 453 299 20 11 35 9 10 27 25 14 12 34 4 7 27 26 342 373 730 249 339 637 463 a Excluded from the analyses. (0) (63) (141) (40) (23) (94) (367) (5) (33) (21) (0) (0) (5) (36) (0) (1) (22) (0) (1) (9) (1) (5) (99) (194) (66) (69) (112) (519) 450 J. Merz et al. values showed that in more than 50% of the events more than one quarter of the event rainfall amount, in certain cases nearly half the rainfall amount (e.g., sites 4 and 6) occurs in the first quarter of the event duration (which may have an impact on the intensity during this time). On the south-facing slopes about 30–35% of the rainfall amount occurs in the first quarter. Another 30–35% of the rainfall occurs in the second quarter at these sites, with 20–25% of the rainfall in the third and about 12% of the rainfall in the last quarter of the event. At the north-facing sites the second quarter is rainfall poor and the two last quarters receive proportional amounts of rainfall with respect to time. Relative frequency of events of different rainfall Figure 2 amounts. For runoff generation and sediment mobilisation the large events are more important as they are responsible for the largest part of the soil loss (Nakarmi et al., 2000; Voegeli, 2002) as well as for the largest flood events (Merz et al., 2000). Carver (1997) showed that serious sediment output from the catchments occurred at a threshold of 30 mm Ptot and of 50 mm/h I10max. He therefore defined major events as events with rainfall amounts higher than 30 mm and maximum 10 min intensities of more or equal than 50 mm/h. He herewith excluded events with large rainfall amounts but minor to medium rainfall intensity from the class of major events. This is, as he has shown, certainly correct for sediment considerations. However, for flood generation this is believed to be different. While short and intense storms can lead to sharp flood peaks, long and persistent rainfall of low intensity can produce large flood volumes with minor flood peaks. Considering all rainfall events the median for rainfall amount ranged depending on the site from 5 to 8 mm with durations of 1–3 h (Table 4). The 75% quartile for rainfall amount reached a maximum of about 15 mm with a 25% quartile of about 4 mm. Fifty percent of the events in the JKC therefore are within the range of about 4–15 mm and therefore minor events. I10max ranged in general from a 25% quartile of about 6 mm/h to a 25% quartile of about 24 mm/h. The observed median was at all sites around 12 mm/h. In terms of hyetograph shape parameters there seems to be a difference between the sites on the northfacing slopes (site 3, 4 and 6) and the sites on the south-facing slope or the valley bottom (sites 12, 14 and 15). The P25 Table 4 Figure 3 (a) 1st (25%), 2nd (50%) and 3rd (75%) quartiles for rainfall amount (mm) and (b) maximum 10 min intensity (mm/h) distribution for large pre-monsoon and monsoon events at all sites. Typical rainfall events in the JKC (based on all observed sites) Parameter All events Large pre-monsoon events Large monsoon events 10 largest events Rainfall amount (mm) Event duration (min) Average intensity (mm/h) 10 min max intensity (mm/h) 30 min max intensity (mm/h) 60 min max intensity (mm/h) Rain in 1st quarter (%) Rain in 1st half (%) Rain in three quarters (%) 5–8 60–180 2–5 12 7 4.5 35–40 50–70 70–90 35 240 13 60 38 26 10–50 50–80 75–90 40 480 6 36 24 18 10–35 60–70 80–90 60 420–1200 5–12 24–48 16–34 13–24 20–40 50–60 80–90 Rainfall-runoff events in a middle mountain catchment of Nepal Events during the pre-monsoon and the monsoon seasons differ. While the rainfall amount tends to be lower during a large pre-monsoon event, the intensities, both maximum and average, tend to be higher (Fig. 3). This is a result of the convective activity during the pre-monsoon season, which is much stronger during this season than during the monsoon season. Pre-monsoon storms tend to be shorter than the storms in the monsoon season. Interestingly at most sites the event rainfall is concentrated to one quarter of the event duration during pre-monsoon events. At sites 3, 4 and 6 it is during the first quarter of the event duration. At sites 12 and 15 it is during the third quarter of the event. During monsoon season events the rain is more evenly spread throughout the event duration. The number of premonsoon events is limited, in general only 1–5 events were observed in the study period. As shown above in Table 3 most rainfall events occur during the pre-monsoon and monsoon seasons. This is also the case for the medium and large sized rainfall events, in particular the large events mostly occur during the monsoon season. The 10 largest events at all sites show a median rainfall amount of about 60 mm ranging from a 25% quartile of about 55 mm up to maximum 75% quartiles of 100 mm. The maximum intensities (median) are very low and show only values of 24–42 mm/h. The 75% quartile can reach to more than 60 mm/h. Relationships between the different precipitation parameters In order to review the full content of information of a rainfall event, a large number of event parameters were calculated accepting that many parameters are closely related and accepting that many parameters show similar characteristics of the events. For further analyses the parameters with the highest information content had to be established. This was done by means of correlation and factor analyses. As the event parameters are not normally distributed shown by Merz (2004) with a Kolmogorov–Smirnov test, the nonparametric correlation analysis according to Spearman was used to show the correlations between the different rainfall parameters. A rather strong correlation of Ptot with most parameters is evident at all sites except Iave and the shape parameters. The different maximum intensity parameters I10max, I30max and I60max in particular show a strong linear relation with Ptot at all sites. The correlation of the shape parameters P25, P50 and P75 is only limited, which shows that there is no direct relation between the shape of a rainfall event and the rainfall amount. In fact certain events show very strong rainfall in the early stages, others at late stages or throughout the event. In the section above, where the events were described, the hypotheses was formulated that in case over proportional amounts of rainfall occur in any particular quarter of the event duration, the intensity would be higher. This cannot be shown on the basis of the correlation between the shape parameters and the intensity parameters. The event duration tP is only strongly linearly related to the rainfall amount Ptot. The remaining correlations are weak. As expected the interrelation between the intensity parameters are strong. Different parameters show very similar aspects of the hyetographs, e.g., the four intensity parameters Iave, 451 I10max, I30max and I60max. In order to identify the key variables of precipitation on the basis of the different rainfall parameters multivariate statistics was applied. Weingartner (1999) suggested the use of factor analyses. For this purpose the parameters were firstly standardised and transformed to z-scores with mean 0 and standard deviation 1. The reason for this transformation is the different scales of the various parameters going into the analyses. The factor extraction was done on the basis of the principle components approach as discussed in Backhaus et al. (2006). The analysis was done for each measurement station separately in order to determine whether all stations show the same characteristics. The extracted factors, i.e., the factors with eigenvalues of at least 1 and herewith explaining at least their own variance, are rotated using the Varimax method. The results of these analyses presented in detail in Merz (2004) suggest the following components with the following key parameters (in bold and underlined, i.e., key parameters with the highest eigenvalues amongst the parameters belonging to the same factor) (refer also to the groups as mentioned above): – Iave, I10max, I30max, I60max – P25, P50, P75 – Ptot, tP Applying cluster analysis rainfall event types were identified with the help of these identified key parameters. In addition this analysis suggests to add Ptot to the set of key parameters which is also supported by Carver’s (1997) classification. The number of clusters has to be defined in advance in the case of the k-means cluster analysis. Four clusters were identified to be appropriate (Table 5). Three clusters would lump the lower clusters together into a main lower cluster, a medium cluster and a cluster of the very large events. Four clusters do not change anything in the large events, but divide the lower cluster into an additional cluster. Five clusters would result in the break down of the largest event cluster producing two very small clusters with only 2–3 cases in each. On the basis of predefined four clusters, cluster centres were identified for all monitoring sites and summarized in Table 5. In words these clusters can be described as minor, medium, high intensity and large events. Minor events are of low rainfall amount and of short duration. During the event only low intensities are observed and most of the rainfall occurs during the 1st half of the event. Medium events are of low to medium rainfall amount and medium duration. Medium intensities are measured and rainfall occurs throughout the event. The upper end of the events shows a high intensity cluster with medium amount and medium event duration, but high intensity rainfall up to 57.4 mm/h which mainly occurs in the 1st half, and a large event cluster with high rainfall amount between about 50 and 170 mm in a long duration event, but only of medium intensity. Merz (2004) identified similar clusters with similar boundaries for the case of the Yarskha Khola catchment, another study area of PARDYP Nepal located about 150 km East of Kathmandu. On average over all sites and all events, it is noted that most of the events belong to cluster 1, the minor events 452 Table 5 J. Merz et al. Final clusters for rainfall event classification, JKC Variable Cluster 1 Min Ptot (mm) tP (min) I30max (mm/h) P50 (%) Classification Description Cluster 2 Max 2.1 9.6 22 250 3.6 10.8 40.0 82.6 Minor Low amount – short duration – low maximum intensity rainfall event with most rainfall amount in the 1st half of the event Min Cluster 3 Max 9.4 32.5 98 728 5.4 20.8 29.7 80.6 Medium Low to medium amount – medium duration – medium intensity rainfall event with rainfall occurring throughout the event Figure 4 Rainfall events distribution according to the different seasons and clusters. (71.9%; Fig. 4a). 49.9% of these events occur during the monsoon season and 16.1% during the pre-monsoon season. Another 14.2% of the events during the monsoon season belong to cluster 2 accounting overall for 16.2% of the events. 10.9% of the events belong to cluster 3, while cluster 4, the exceptional events, only account for 1% of all events. Min Cluster 4 Max 12.8 45.4 46 421 18.8 57.4 43.3 91.3 High intensity Medium amount – medium duration – high intensity rainfall event with most rainfall occurring in the 1st half of the event Min Max 52.1 164.4 795 1931 9.4 21.4 36.6 62.2 Large High amount – long duration – medium intensity rainfall events with most rainfall in the 2nd half of the event Seasonally it is noted that the post-monsoon and winter seasons account for an over proportional share of events belonging to cluster 4 (Fig. 4b). During post-monsoon in particular a number of exceptional storms occurred with the characteristics of cluster 4. During the pre-monsoon season the share of minor events (cluster 1) as well as the high intensity events (cluster 3) is higher in comparison with the other clusters. Comparing the classifications of Carver (1997) based on rainfall amount only and of the cluster analysis, it can be observed, that the cluster analysis based classification puts more emphasis on the events with high intensities and high rainfall amount, which in theory are the most destructive events. It takes into account the difference between the high intensity-medium duration events, which are particularly important during the pre-monsoon season and the exceptional events with long duration and high rainfall amounts, but only medium intensities. The latter drop out of Carver’s classification, as they are not decisive for sediment mobilisation. For the remaining classes Carver’s classification and the cluster analysis based classification are generally very close with similar thresholds of 10 mm rainfall amount for intermediate/medium events and 30 mm for major/large events. In terms of annual and seasonal frequencies of the different clusters refer to Table 6. This table shows that on average over the entire catchment annually about 61 minor events have to be expected (i.e., 13.7 events in pre-monsoon, 42.5 events in monsoon, 2.8 events in post-monsoon and 2.3 events in winter season, respectively). Medium events are about 14 (0.9, 12.1, 0.4 and 0.4), while the high intensity events are about 9 (2.0, 7.3, 0.1 and 0.0). These high intensity events are mainly occurring during the monsoon season with about 6–10 events. During the pre-monsoon season 2–3 of these events have to be expected. Large events only occur exceptionally with about 1 each year either occurring in the monsoon or post-monsoon season. According to the classification of Carver (1997) 2.8 major storms have to be expected, 1.0 storm during the pre-monsoon season, 1.8 storms during the monsoon season. However, this also includes some of the large high intensity storms. Rainfall-runoff events in a middle mountain catchment of Nepal Table 6 Annual frequencies of different events classified according to clusters Cluster Pre-monsoon Site 3 Site 4 Site 6 Site 12 Site 14 Site 15 Catchmenta a Monsoon Post-monsoon Winter 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 7.8 15.0 14.0 15.0 16.0 14.5 13.7 1.5 1.0 1.6 0.3 0.3 0.5 0.9 0.3 2.8 2.3 3.0 2.3 1.3 2.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 43.8 43.5 41.1 42.3 43.5 40.5 42.5 20.0 10.8 11.4 11.7 11.0 8.0 12.1 2.5 10.8 10.3 6.0 6.5 7.5 7.3 1.3 0.3 0.4 0.3 0.5 0.6 0.6 4.0 2.3 3.4 2.3 2.0 2.6 2.8 0.8 0.3 0.5 0.3 0.3 0.5 0.4 0.0 0.0 0.4 0.0 0.0 0.1 0.1 0.0 0.3 0.1 0.3 0.3 0.1 0.2 3.3 2.5 2.9 1.0 1.3 2.9 2.3 0.3 0.3 0.9 0.3 0.5 0.4 0.4 0.0 0.0 0.1 0.0 0.0 0.0 0.0 0.3 0.0 0.1 0.0 0.0 0.1 0.1 Catchment: average of the results from the different stations in the catchment. the agricultural land behave differently. The same can be shown on the basis of the runoff event analyses. The median runoff of all measured events at sites 4 and 14 is with 40 m3/ha (=4 mm) about eight times higher than on the agricultural plots at sites 6 and 16, where only about 2–5 m3/ha (=0.2–0.5 mm) runoff are measured (Fig. 5). The values on degraded land range from a 25% quartile of about 1 mm up to a 75% quartile of 9.5 mm. On the agricultural land the statistical values consistently are below 1 mm. The median rainfall parameters corresponding to these runoff events are of medium size, i.e., between 10 and Runoff event description In the JKC the upper slopes are mainly used for rainfed agriculture and soils on these steep slopes are predominantly brown soils. Degradation of these soils is limited. In the lower areas below 1200 masl and on gentle slopes red soils are predominant and are in places heavily degraded. Erosion plots at sites 4 and 14 represent the conditions for degraded land on red soils and at sites 6 and 16 the conditions for sloping rainfed agricultural land on brown soils. On the degraded plots annually about 55 runoff events are observed. A runoff event is defined as the amount of runoff that occurred as a result of a single rainfall event. After each rainfall event the plot observers measure the runoff collected in the drums as shown in the methodology section. Generally runoff heights of 5 cm in the drums and below are omitted from the analyses. 4/5 of these events occur in the monsoon season (Table 7), 1/5 occurs in the pre-monsoon season. On the rainfed agricultural plots only about 20–30 events were observed annually with most events occurring in the monsoon season. During the monsoon season about twice as many events can be observed on the degraded plots than on the plots on rainfed agricultural land. The same approximate factor applies for the pre-monsoon season. According to the evidence presented above and the results on sediment dynamics presented in Merz (2004), it is clear that the plots on the degraded land and the plots on Table 7 Year Pre 1993 1994 1995 1996 1997b 1998 1999 2000 Figure 5 Event parameters (1st, 2nd and 3rd quartile) for runoff distribution of all events on the erosion plots in the JKC. Number of runoff events on the erosion plots in the JKC Site 4 (d) a Site 6 (a) a Mon Post a a 11 12 6 15 43 48 54 26 3 2 2 0 Site 14 (d) Win Pre Mon Post Win 4 1 0 0 0 5 4 1 4 8 3 9 13 14 20 19 24 22 28 25 0 0 3 1 0 1 1 0 0 2 2 0 1 1 0 0 b a 453 Pre 9 10 9 14 Mon 33 42 54 45 Site 16 (a) Post 1 1 1 0 Win Pre Mon Post 1 1 0 0 6 4 1 1 31 0 2 1 1 0 27 2 6 20 0 4 17 0 2 8 1 No rainfall data Win 0 1 0 1 2 1 0 Pre: pre-monsoon; Mon: monsoon; Post: post-monsoon; Win: winter. 1993 in the case of sites 6 and 16 and 1997 in the case of sites 4 and 14 are the initial years and therefore only of limited use. This is due to the disturbed soil conditions just after installation of the plots. b 454 J. Merz et al. Table 8 Medians of all runoff events and corresponding rainfall parameters on the erosion plots, JKC Site ROa (mm) Ptot (mm) tP (min) Iave (mm/h) I10max (mm/h) I30max (mm/h) I60max (mm/h) P25 (%) P50 (%) P75 (%) u (%) 4 (d) 6 (a) 14 (d) 16 (a) 3.7 0.4 3.5 0.2 10.4 14.8 11.0 19.4 138 166 214 270 4.6 5.8 3.5 4.1 12.6 19.2 18.0 22.2 8.4 12.6 10.0 15.4 6.3 8.4 6.7 10.5 33.3 40.0 34.5 29.6 60.0 63.6 70.6 67.4 80.0 83.3 90.9 91.7 30.9 2.7 31.2 0.9 a RO = runoff measured in the erosion plot (mm). 20 mm rainfall (Table 8). Again the plots of different land use differ slightly with the median rainfall amount of the events on agricultural land being slightly higher than on the degraded land, i.e., lower rainfall leads to more runoff and earlier to runoff than on agricultural land. This will also be shown with the establishment of lower thresholds for runoff generation on degraded plots below. Comparing the runoff coefficients a from the degraded plots and the agricultural plots, it can be shown, that degraded plots are more susceptible to runoff generation than the agricultural plots. A median 31% of the rainfall from the degraded plots runs off, while on agricultural land this value is only about 1–3%. In terms of the other parameters there is no distinct difference visible. The runoff events differ largely between the seasons. While runoff events on the not degraded plots are very seldom during post-monsoon and winter (Table 7), about 10 events on degraded plots and 5 events on agricultural plots occur during the pre-monsoon season. Most of the events occur during the monsoon season. While on the degraded plots the runoff volume largely follows the rainfall distribution (i.e., more rainfall occurs during the monsoon season, therefore more runoff occurs during the same season; Table 9), on agricultural plots runoff is higher during the pre-monsoon season or the same as during the monsoon season, though more rainfall occurred. The same can be shown with the runoff coefficient u, an overall and summarised measure of infiltration and storage processes (Scherrer, 1997). On degraded plots the runoff coefficients during the premonsoon season tend to be smaller than during the monsoon season (see also Fig. 6). On agricultural plots it tends to be just the other way round, i.e., higher coefficients during the pre-monsoon season. This is particularly interesting as the Table 9 Site Figure 6 Quartiles of the runoff coefficient distribution of the runoff events on the erosion plots of the JKC. intensity parameters only differ slightly between the seasons. However, the pre-monsoon events tend to be more intense according to the intensity parameters. The range of runoff coefficients is rather high ranging from 10% to 40% on degraded land during the pre-monsoon season and about 20% to 50% during the monsoon season. On agricultural land runoff coefficients of only up to roughly 10% are observed. Of all the rainfall events during a year only some generate runoff on the plots. Most of these runoff events however are minor events with only small amounts of runoff. Only the largest events have to be considered to be important. At sites 4 and 14, the degraded lands, about 10 events produce 50% of the total annual runoff. Twenty events produce about 75% and 30–35 events produce about 90% of the total annual runoff. On the rainfed agricultural land, i.e., sites 6 Parameter medians for all pre-monsoon (PM) and monsoon (M) runoff events P25 (%) P50 (%) P75 (%) u (%) 5.2 6.3 41.4 33.3 54.4 62.5 78.9 83.3 4.7 31.1 14.8 12.6 9.0 8.4 42.9 40.0 66.7 60.9 85.6 83.3 6.9 2.5 24.0 16.2 12.4 9.8 6.9 6.7 46.4 33.3 76.0 70.6 93.8 90.5 7.1 33.2 33.0 23.4 18.4 15.4 11.5 10.0 30.4 28.5 74.0 67.5 94.9 91.5 13.4 0.9 Season RO (mm) Ptot (mm) tP (min) Iave (mm/h) I10max (mm/h) 4 (d) PM M 1.9 4.3 7.3 11.5 120 154 4.7 4.5 18.6 18.6 8.4 8.4 6 (a) PM M 0.6 0.4 10.5 15.8 72 190 7.5 5.4 25.2 19.2 14 (d) PM M 2.3 3.7 9.7 11.1 130 249 5.6 3.3 16 (a) PM M 0.2 0.2 17.4 19.7 170 313 5.0 3.8 I30max (mm/h) I60max (mm/h) 0.24 0.36 0.27 0.17 0.21 0.35 0.32 0.21 0.22 0.35 0.43 0.27 0.16 0.23 0.16 0.20 0.20 0.32 0.43 0.15 AP4 AP3 AP2 AP1b API14 0.35 0.22 0.21 0.32 0.20 0.20 API10 0.26 0.51 0.18 0.19 0.20 0.32 0.43 API7 API1a P75 P50 APIx (mm/day) sum of rainfall x days before the event divided by x. APx (mm) rainfall x days before the event. a b 0.24 0.82 0.37 0.67 0.51 4 6 14 16 Site Site Site Site 0.31 0.82 0.56 0.74 0.70 0.38 0.23 0.42 0.39 0.62 0.36 0.63 0.49 0.72 0.38 0.71 0.51 0.81 0.39 0.74 0.53 P25 I60max I30max I10max Iave a tP Ptot Site Surface runoff at the plot scale is caused by a number of factors. Collins et al. (1998a) show that both infiltration excess and saturation excess processes contribute to runoff generation in the middle mountains of Nepal. Kandel et al. (2002) therefore use a surface runoff model, which incorporates both processes after accounting for the canopy interception losses. There is no argument that the triggering mechanism for surface runoff is rainfall. However, the questions, what Correlation coefficients for runoff – summary of the four erosion plots in the JKC (grey shaded: agricultural plots) Relationship between rainfall and runoff events Table 10 and 16, only 5 events produce 50% and 10–15 events produce 75% of the total annual runoff. At site 16 15–20 events produce 90% of the total annual runoff. At site 6 the same is achieved by 20–30 events. The importance of selected large storms is even higher in the case of soil loss (Nakarmi et al., 2000; Merz, 2004). The 10 largest runoff events on the erosion plots of the JKC show a similar, but much clearer picture as above for all events and the seasonal break down. There is a distinct difference between the events on the agricultural plots and the plots on the degraded land (Fig. 7). The events on degraded lands show median values of about 25 mm runoff during the largest events. On the agricultural land the largest events only record about 6 mm runoff. A difference is also observed in terms of the runoff coefficient a. On the degraded plots 40–50% runoff on average, while on the agricultural land only medians of 16.2% at site 16 and 32.3% at site 6 were observed. In summary it can be said that the runoff behaviour on the degraded plots differs strongly from the runoff behaviour on the agricultural plots. While a clear seasonal pattern can be observed on the agricultural plots, no such pattern is visible on the degraded plots. An average runoff event on a degraded plot produces about 3–4 mm runoff, while on a rainfed agricultural plot it only produces 0–0.5 mm. In comparison, the ten largest events on a degraded plot produce on average about 20–25 mm runoff, while on a rainfed agricultural plot they only produce 5–6 mm. Seventy-five percent of the annual runoff is produced during about 20 events on the degraded plots. On rainfed agricultural plots only 10–15 produce the same percentage of runoff. API30 Figure 7 Event parameters (1st, 2nd and 3rd quartile) for runoff distribution of the 10 largest runoff events on the erosion plots in the JKC. 0.21 0.38 0.26 0.18 455 AP5 Rainfall-runoff events in a middle mountain catchment of Nepal 456 J. Merz et al. parameter leads to lower or higher runoff, remains and will be discussed here. The Spearman correlation coefficients (the erosion plot data is not distributed normally; Merz, 2004) show that there is a distinct difference between the plots on agricultural land and the plots on degraded land (Table 10). The runoff amounts from the degraded plots show a strong correlation with the rainfall amount Ptot as well as the intensity parameters, I60Max in particular. The runoff amounts on the rainfed agricultural land however show only poor relations with the rainfall amount as well as with the intensity parameters. This suggests that other factors, such as land management and cropping, are more important for the estimation of runoff generation on these plots. On all four plots the antecedent precipitation shows mostly significant, but only weak correlation with the runoff amounts on the plots. The rainfall 24 h prior to the event shows the highest correlation coefficients ranging from 0.20 to 0.43. On plot 16 however no significant correlation between runoff and this parameter was observed. The shape of the event hyetograph shows no or only very week correlation and can therefore be assumed to have no influence on the runoff generation. The significant correlations between rainfall amount and runoff can also be shown with the seasonally disaggregated data from four erosion plots on daily basis. Runoff rates on degraded plots are generally higher with the highest events well over 10 mm. On agricultural land the highest measured rates were between 8 and 10 mm. Thresholds of rainfall to produce runoff are lower on degraded land than on rainfed agricultural land. It is estimated to be 2 mm rainfall on degraded sites and 5 mm on rainfed agricultural land. While in Figure 8 degraded sites seasonality does not seem to have any effect, on rainfed agricultural plots pre-monsoon rainfall events seem to yield higher runoff rates than events of the remainder of the year. Above rainfall events were classified into four clusters according to event rainfall amount, maximum 30 min intensity, rainfall event duration and shape parameter P50. Comparing the runoff events from the erosion plots with these rainfall event clusters, it is evident that cluster 3 event rainfall events, i.e., high intensity events, show most responsible for runoff generation on the degraded plots (Fig. 8). This is followed by cluster 2 events. The large amount-long duration events (cluster 4) are only marginally responsible for runoff generation and often in the same range as the runoff generated by cluster 1 events. On the agricultural plots the presented picture is different between the two plots at site 6 and 16. While at plot 6 there is a clear dominance and role of cluster 4 events responsible for runoff generation, at site 6 both cluster 3 and 4 events show similar impact. In both cases events of clusters 1 and 2 do not show much impact. A possible explanation for this difference between the plots is the importance of different runoff generating mechanisms on the different land use. While on degraded land it is mainly infiltration excess overland flow, which contributes to the runoff, on agricultural land saturation excess overland flow gains importance. This is shown with the delayed reaction of the runoff process on agricultural land and the lower correlation between rainfall intensity and runoff parameters shown in Table 10. In order to verify this assumption detailed soil moisture analyses on different land uses would be required. Relationship between runoff and the rainfall clusters. Rainfall-runoff events in a middle mountain catchment of Nepal Summary and conclusion The understanding of the rainfall behaviour particularly during single rainfall events is the basis towards improved understanding of the generation of destructive floods triggered by rainfall. It was shown here, that – Annually about 86 events occur, of which about 73% occur in the monsoon season and 20% in the pre-monsoon season; – The event distribution is strongly left skewed with the events between 2 and 5 mm being most frequent; – During an average rainfall event about 5–8 mm rainfall is observed during 1–3 h and with a maximum 10 min intensity of 12 mm/h; – The pre-monsoon events show higher maximum intensities, shorter duration and less rainfall than the events during the monsoon season; – The total event rainfall volume Ptot is strongly correlated with most other rainfall event parameters; – The rainfall event duration tP, the maximum 30 min intensity I30max and the rainfall that occurred in the first half of the event P50 are the key variables in a rainfall event; – Four clusters can be identified on the basis of Ptot, tP, I30max and P50: – Cluster 1 – minor events: Low amount – short duration – low maximum intensity rainfall event with most rainfall amount in the 1st half of the event. – Cluster 2 – medium events: Low to medium amount – medium duration - medium intensity rainfall event with rainfall occurring throughout the event. – Cluster 3 – high intensity events: Medium amount – medium duration – high intensity rainfall event with most rainfall occurring in the 1st half of the event. – Cluster 4 – large events: High amount – long duration – medium intensity rainfall events with most rainfall in the 2nd half of the event. – Annually about 9 high intensity events and 1 large event occurred. – Out of these 7 high intensity events occurred during the monsoon season and 2–3 during the pre-monsoon season; – Large events generally occurred during the monsoon season or the post-monsoon season. The gained information on the rainfall events needs to be put in perspective of the runoff at the plot scale. The analysis of the runoff events on the erosion plots showed that – total rainfall volume and the maximum intensities, I60max in particular show the highest correlations with runoff; – on degraded plots a rainfall event volume threshold of 2 mm was observed for runoff generation; – on rainfed agricultural land the threshold was 5 mm; – infiltration excess overland flow is the dominant surface runoff generation process on degraded land; – saturation excess overland flow is the dominant surface runoff generation process on rainfed agricultural land; – degraded lands are more prone to runoff generation than rainfed agricultural land; 457 – degraded lands yield higher runoff rates than rainfed agricultural lands; – degraded lands do not show seasonal effects, while agricultural land shows a clear seasonality; – rainfed agricultural lands are more prone to high rates of direct runoff during pre-monsoon than during the remainder of the year; – only a few large rainfall events cause a large portion of the annual runoff on the erosion plots; – event rainfall volume and maximum 60 min intensity are the key variables in terms of runoff generation; The Himalayan farmers are often held responsible for downstream flooding Hofer (1998). However, on the basis of the analysis above it is noted that the current management of agricultural land seems to be beneficial to flood generation. Agricultural land shows only marginally contributing to runoff generation on the plots. Degraded land on the other hand, where runoff seems to be generated by the infiltration excess overland flow generation mechanism, seems to contribute to a large extent (though the overall portion of degraded land is rather low). These results show the importance of rehabilitation of degraded areas and the proper management of the land resources. While these mechanisms can be statistically shown at the plot level, it remains to be seen whether these processes will also be detectable at the catchment scale. Especially as on that scale sub-surface flow mechanisms will play an additional role to the surface flow mechanisms discussed here. Acknowledgements The PARDYP project is jointly financed by the Swiss Agency for Development and Cooperation (SDC), the International Development Research Centre (IDRC) and the International Centre for Integrated Mountain Development (ICIMOD). The authors would like to acknowledge the continuous support of the donors and the continuous assistance of the colleagues in PARDYP Nepal. The comments of two anonymous reviewers were highly appreciated and helped to improve the manuscript. References Agarwal, A., Narain, S. (Eds.), 1991. Floods, flood plains and environmental myths. State of India’s Environment Report 3. Centre for Science and Environment, New Delhi. Backhaus, K., Erichson, B., Plinke, W., Weiber, R., 2006. Multivariate Analysenmethoden, Eine anwendungsorientierte Einfuehrung. Springer Verlag, Berlin. Badoux, A., Witzig, J., Germann, P.F., Kienholz, H., Luscher, P., Weingartner, R., Hegg, C., 2006. Investigations on the runoff generation at the profile and plot scales, Swiss Emmental. Hydrological Processes 20, 377–394. Carver, M., 1997. Diagnosis of headwater sediment dynamics in Nepal’s middle mountains: implications for land management. Unpublished PhD Thesis. University of British Columbia, Vancover. Chalise, S.R., Khanal, N.R., 2002. Recent extreme weather events in the Nepal Himalayas. In: Snorrason, A., Finnsdottir, H.P., Moss, M.E. (Eds.), The Extremes of the Extremes: Extraordinary 458 Floods. Proceedings of a Symposium held in Reykjavik, July 2000. Publ. No. 271. IAHS, Wallingford. pp. 141–146. Chalise, S.R., Shrestha, M.L., Thapa, K.B., Shrestha, B.R., Bajracharya, B., 1996. Climatic and Hydrological Atlas of Nepal. ICIMOD, Kathmandu. Collins, R.P., Dickinson, A., Johnson, R., 1998a. The use of GIS and remote sensing techniques to predict erosion in the Nepal Middle Hills. In: Penning de Vries, F.W.T., Argus, F., Kerr, J. (Eds.), Soil Erosion at Multiple Scales. CAB International, pp. 303–315. Collins, R.P., Jenkins, A., Sloan, W.T., 1998b. A GIS framework for modelling nitrogen leaching from agricultural areas in the Middle Hills Nepal. International Journal of Geographical Information Science 12 (5), 479–490. Germann, P.F., 1990. Macropores and hydrological hillslope processes. In: Anderson, M.G., Burt, T.P. (Eds.), Process Studies in Hillslope Hydrology. John Wiley and Sons, Chichester, pp. 327– 363. Helsel, D.R., Hirsch, R.M., 1992. Statistical methods in water resourcesStudies in Environmental Science, vol. 49. Elsevier, Amsterdam. Hofer, T., 1998. Floods in Bangladesh – A Highland-Lowland Interaction? Geographica Bernensia G48. Institute of Geography, University of Bern, Bern. Kandel, D.D., Western, A., Grayson, R., Turral, H., 2002. Surface runoff modelling. Part 1: testing at fine time scale and scalingup to daily scale. In: Proceedings of the 27th Hydrology and Water Resources Symposium 2002, 20–23 May 2002, Melbourne. Laemmli, M., 2000. Vergleichende hydrologische Untersuchungen in drei Kleinsteinzugsgebieten des Fulwasserbaches. Publikation Gewaesserkunde Nr. 250. Department of Geography, University of Bern, Bern. Leibundgut, Ch., Uhlenbrook, S., McDonnell, J. (Eds.), 2001. Runoff Generation and Implications for River Basin Modelling. Band 13 Freiburger Schriften zur Hydrologie. Institute for Hydrology, University of Freiburg, Freiburg. Merz, J., 2004. Water Balances, Floods and Sediment Transport in the Hindu Kush-Himalayan region. Geographical Bernensia G72. Department of Geography, University of Bern and International Centre for Integrated Mountain Development, Bern and Kathmandu. Merz, J., Mosley, M.P., 1998. Hydrological behaviour of pastoral hill country modified by extensive landsliding, northern Hawke’s Bay, New Zealand. Journal of Hydrology (NZ) 37 (2), 113–139. Merz, J., Dongol, B.S., Weingartner, R., 2000. Impact of land use on generation of high flows in the Yarsha Khola catchment, Nepal. In: Allen, R., Schreier, H., Brown, S., Shah, P.B. (Eds.), The People and Resource Dynamics Project. The First Three Years (1996–1999). Proceedings of a Workshop held in Baoshan, 2–5 March 1999. ICIMOD, Kathmandu. pp. 185–198. J. Merz et al. Merz, J., Dangol, P.M., Dhakal, M.P., Dongol, B.S., Weingartner, R., 2006. Rainfall amount and intensity in a rural catchment of Nepal’s middle mountains. Hydrological Sciences Journal 51 (1), 127–143. Mosley, M.P., 1979. Streamflow generation in a forested catchment, New Zealand. Water Resources Research 15 (4), 795–806. Mosley, M.P., 1981. Delimitation of New Zealand hydrologic regions. Journal of Hydrology 49, 173–192. Naef, F., Zuidema, P., Koella, E., 1986. Abschaetzung von Hochwassern in kleinen Einzugsgebieten. In: Spreafico, M. (Ed.), Abschaetzung der Abfluesse in Fliessgewaessern an Stellen ohne Direktmessung. Beitraege zur Geologie der Schweiz – Hydrologie Nr. 33. Schweizerische Geotechnische Kommission and Hydrologische Kommission der Schweizerischen Naturforschenden Gesellschaft, Bern. pp. 195–233. Nakarmi, G., Schreier, H., Merz, J., Mathema, P., 2000. Erosion dynamics in the Jhikhu and Yarsha Khola catchments in Nepal. In: Allen, R., Schreier, H., Brown, S., Shah, P.B. (Eds.), The People and Resource Dynamics Project. The First Three Years (1996–1999). Proceedings of a Workshop held in Baoshan, 2–5 March 1999. ICIMOD, Kathmandu. pp. 209–217. Nayava, J.L., 1980. Rainfall in Nepal. The Himalayan Review 12, 1– 18. Pearce, A.J., Stewart, M.K., Sklash, M.G., 1986. Storm runoff generation in humid headwater catchments 1. Where does the water come from? Water Resources Research 22 (8), 1263–1272. Scherrer, S., 1997. Abflussbildung bei Starkniederschlaegen. Identifikation von Abflussprozessen mittels kuenstlicher Niederschlaege. VAW Mitteilungen, vol. 147. Versuchsanstalt fuer Wasserbau, Hydrologie und Glaziologie, ETH, Zuerich. Uhlenbrook, S., 2005. Von der Abflussbildungsprossforschung zur prozess-orientierten Modellierung - eine Review. Hydrologie und Wasserbewirtschaftung 49, H.1:13–H.1:24. Uhlenbrook, S., Frey, M., Leibundgut, C., Maloszewski, P., 2002. Hydrograph separations in a mesoscale mountainous basin at event and seasonal timescales. Water Resources Research 38 (6), 1–14. Voegeli, Y., 2002. Relationship between rainfall, runoff and sediment dynamics. Investigations on a bari field in a subcatchment of the Jhikhu Khola catchment, Nepal. Publikation Gewaesserkunde No. 270. Department of Geography, University of Bern, Bern. Weingartner, R., 1999. Regionalhydrologische Analysen – Grundlagen und Anwendungen. Beiträge zur Hydrologie der Schweiz – Nr. 37, Bern. Wuethrich, P., 1999. Hochwasserverhaeltnisse in einem Wildbach – Zur Hydrologie des Fulwasserbachs im Testgebiet Leissigen (Berner Oberland). Publikation Gewaesserkunde No. 227. Department of Geography, University of Bern, Bern.