Estimation of suspended sediment concentration (SSC) in rivers is a prerequisite to address many issues related to hydrology. Therefore, we make an attempt in this study to introduce a low-cost technique to estimate the SSC. Both surface and depth-and-width-integrated water samples were collected and measured for SSC from eight tributaries in Sri Lanka over a complete hydrological year. A site-specific calibration curve was established between SSCs measured by two methods for each tributary where R2 varied from 0.72 to 0.99. The same relationship is developed in general for all tributaries studied in the hilly terrain of Sri Lanka. This generic model exhibits a strong correlation (R2 = 0.91), which will be useful to calculate an accurate SSC from a simply measured surface SSC. To select the appropriate gauging method, be it surface or depth-and-width-integrated sampling, a new concept of surface sampling threshold factor (SSTF) is introduced. The preliminarily analysis on SSTF using available data for the studied catchments reveals that surface sampling is only adequate for estimating a representative SSC if SSTF varies from 35 to 45. When SSTF deviates from this range, the SSC measured by surface sampling needs to be adjusted by depth-and-width-integrated sampling.
INTRODUCTION
Estimating fluvial sediment transport has drawn interest from many researchers because sediment load data are extensively used in evaluating environmental issues such as transport of contaminants, water quality, reservoir sedimentation, channel and harbor siltation, soil erosion, changes in soil nutrient cycling and ecological impacts (Mimikou 1982; Ferguson 1986; Syvitski et al. 2000; Horowitz et al. 2001; Horowitz 2002; Hewawasam 2003; Nyssen et al. 2009; Walling 2009). However, there is a considerable lack of data with regard to suspended sediment concentration (SSC) in rivers around the world, particularly those in the developing world (Walling 2009). It is obvious that monitoring every tributary and river in the world for their SSC transport is not feasible. Thus, more attention is being paid to predicting it from rating curves or through hydrological modeling mainly based on pre-existing field measurement data (Mimikou 1982; Becvar 2006; Wang et al. 2008; Jain 2012). In Sri Lanka, data generated by field measurements of SSCs are very rare, even though 103 individual river basins exist in the country (Figure 1). As a consequence, scientists and decision makers recurrently face difficulties when they deal with sediment related problems. Moreover, answers to some fluvial sediment related problems of the country are yet to be found. Quantification or prediction of suspended sediment loads in Sri Lankan river basins thus becomes imperative.
To quantify suspended sediment load in a particular river, both SSC and corresponding river discharge data are required (Thomas 1988). To determine a representative annual total of suspended load, continuous monitoring of SSC over several years is preferred, at least covering one hydrological year (Ndomba et al. 2008). In a humid tropical environment, rainfall is seasonal and it changes frequently throughout the year. Small rivers in tropics are more sensitive to temporal variation, for instance, land use changes due to seasonal crops and resulting sediment delivery is comparatively higher than large rivers (Lu et al. 2006). Soil erosion is a serious environmental problem in the tropics and the sediments transported in rivers are highly affected by anthropogenic activities. Sedimentation of inland reservoirs has been identified as a significant problem (Nagle et al. 1999; Hewawasam et al. 2003; Latrubesse et al. 2005). In this context, monitoring of rivers for suspended loads over several years is particularly important for small catchments (e.g. <500 km2) in humid tropics in order to obtain a realistic value of SSC.
In the tropical rivers in Sri Lanka, discharge data are usually available but continuously monitored SSC data are very rare even though soil erosion has been recognized as a major threat (Hewawasam et al. 2003; Hewawasam 2010). The limited availability of SSC data largely reflects financial constraints as monitoring and maintenance costs involved in SSC data collection are high (Walling 2009; Akrasi 2011). Nevertheless, sediment field measurements have been carried out at 12 gauging locations island-wide, mainly in the Upper Mahaweli River, but little information is available regarding the actual measurements, apart from the field reports of Irrigation Department of Sri Lanka and NEDECO (NEDECO 1984; International – Decon Consulting Engineers Frankfurt (LIDE) 1988; Wallingford 1995). The collection and analysis of sediment data in these measurements mainly focused on the largest river basin in Sri Lanka, viz. Mahaweli River, because a series of large multi-purpose reservoirs are situated within the basin. Such field measurements in other river basins in Sri Lanka are hitherto unknown.
The amount and the nature of suspended load in rivers are affected by the availability of sediment and by turbulence in the water column (Ongley 1996). When discharge increases, coarse bed material tends to get into suspension due to turbulence, providing a high SSC. This coarser material is less homogeneously distributed in a fluvial cross-section. When the discharge is sufficiently low, suspended sediment is transported homogeneously in the section (Ying 1996). Thus, during recent decades, an important task for hydrologists has been to develop a strategy to obtain a cross-sectional representative measurement of SSC. The United States Geological Survey has proposed two methods for collecting cross-sectional representative SSC by using point and depth-and-width integrated methods (Edwards & Glysson 1970). However, collecting cross-sectional representative suspended samples is comparatively expensive and requires several field practitioners. Therefore, bucket sampling or surface sampling is still popular in many countries due to the ease of handling and the low-cost. Yet, data on the accuracy and reliability studies based on surface sampling and cross-sectional representative sampling are scarce, especially in tropical highlands. Hence, in this study, we conducted an extensive SSC monitoring program on eight tributaries of the second and the third largest rivers in Sri Lanka by using an integrated approach of daily surface monitoring and monthly depth-and-width-integrated monitoring of river loads covering a complete hydrological year.
The main objective of this research is to present a low-cost and simply applicable method to measure and estimate SSCs in small tropical catchments. Therefore, the research was focused to establish relationships between surface and depth-and-width-integrated SSCs for catchments individually and for the entire region in general via correlation and regression analyses. In order to suggest an appropriate sediment sampling method for a particular river, the research aimed at introducing ‘a surface sampling threshold factor’ (SSTF) based on catchment properties and relationships between the surface and depth-and-width-integrated SSC. The estimated SSCs in these catchments will be transformed into sediment fluxes then incorporated with bed and dissolved loads. They will finally be converted into spatially-averaged erosion rates, which is another important scope of the entire project. However, this latter part is beyond the scope of this paper.
MATERIALS AND METHODS
Study area
Kalu Ganga and Walawe Ganga (‘Ganga’ meaning river in Sri Lanka's native Sinhala), the second and the third largest river basins in Sri Lanka (Figure 1), play a significant role in the country's economy and the environment. Presently there are no reservoirs impounded in the Kalu Ganga basin. However, flood control reservoirs are proposed for this basin by the Irrigation Department of Sri Lanka, since it is situated in a heavily and frequently flooded region. A few multipurpose reservoirs have been constructed in the Walawe Ganga, located in the downstream region, far-off (ca. 4 km) from the tributaries that are selected for this study. Hence, the effect of reservoirs on sediment loads is insignificant in the selected tributaries, which may perhaps be the most influential factor in sediment loads in many rivers in the world (Walling & Fang 2003). Human activities, particularly clearing the lands for agriculture and gem mining, are common in these two river basins. Eight tributaries with catchment areas ranging from 23 to 334 km2 were selected within the two basins. The tributaries were selected based on the availability of river discharge data, ease of accessibility, and the diversity in land relief, climate and land use. These tributaries are uniformly underlain by crystalline metamorphic rocks, mainly of siliceous composition, formed under granulite and amphibolite facies conditions (Kröner & Jaeckel 1994).
The Walawe Ganga originates in the southern part of the Central Highlands of Sri Lanka, at an altitude of over 2,000 m, and drains a 2,442 km2 basin. Three morphological levels occur in the Walawe basin, namely, uplands (highland), intermediate mountainous and the lowland plain. Mean annual precipitation varies within the basin from over 3,000 mm in the mountains to around 1,000 mm along the coast (Meteorological Department, Sri Lanka 2004). Five tributaries were selected for surface water sampling within the basin: Belihul Oya, Diyavini Oya, Weli Oya, Rakwana Ganga and Hulanda Oya (Figure 1, Table 1).
River basin . | Tributary . | Gauging station . | Mean annual discharge (m3/s) . | Monitoring period of discharge . | Mean annual SSC (g/L) . | MLR (m) . | DD (km/km2) . | RG (m/km) . | Forest percentage (%) . | ME (m) . | Area (km2) . |
---|---|---|---|---|---|---|---|---|---|---|---|
Walawe | Belihul Oya | Belihul Oya | 2.7 | 1956–1960 | 0.007 | 368 | 2.6 | 81 | 54 | 1488 | 49 |
Diyavini Oya | Diyavinna | 0.5 | 1964–1967 | 0.03 | 181 | 5.5 | 52 | 29 | 230 | 23 | |
1967–1975 | |||||||||||
Weli Oya | Weli Oya | 7 | 1962–1971 | 0.03 | 366 | 2.4 | 68 | 9.2 | 697 | 261 | |
Rakwana Ganga | Timbolketiya | 5 | 1958–1967 | 0.03 | 381 | 2 | 34 | 11 | 400 | 269 | |
Hulanda Oya | Modarawana | 1 | 1956–1958 | 0.05 | 250 | 2 | 31 | 1 | 340 | 109 | |
Kalu | Kalu Ganga | Malwala | 31.3 | 1954–1979 | 0.02 | 404 | 2.5 | 97 | 32 | 400 | 334 |
2005–2007 | |||||||||||
Denawaka Ganga | Lellopitiya | 4.7 | 1958–1966 | 0.08 | 388 | 2 | 84 | 14 | 305 | 75 | |
Wey Ganga | Dela | 15.5 | 1956–2006 | 0.24 | 322 | 2.1 | 53 | 5 | 335 | 220 |
River basin . | Tributary . | Gauging station . | Mean annual discharge (m3/s) . | Monitoring period of discharge . | Mean annual SSC (g/L) . | MLR (m) . | DD (km/km2) . | RG (m/km) . | Forest percentage (%) . | ME (m) . | Area (km2) . |
---|---|---|---|---|---|---|---|---|---|---|---|
Walawe | Belihul Oya | Belihul Oya | 2.7 | 1956–1960 | 0.007 | 368 | 2.6 | 81 | 54 | 1488 | 49 |
Diyavini Oya | Diyavinna | 0.5 | 1964–1967 | 0.03 | 181 | 5.5 | 52 | 29 | 230 | 23 | |
1967–1975 | |||||||||||
Weli Oya | Weli Oya | 7 | 1962–1971 | 0.03 | 366 | 2.4 | 68 | 9.2 | 697 | 261 | |
Rakwana Ganga | Timbolketiya | 5 | 1958–1967 | 0.03 | 381 | 2 | 34 | 11 | 400 | 269 | |
Hulanda Oya | Modarawana | 1 | 1956–1958 | 0.05 | 250 | 2 | 31 | 1 | 340 | 109 | |
Kalu | Kalu Ganga | Malwala | 31.3 | 1954–1979 | 0.02 | 404 | 2.5 | 97 | 32 | 400 | 334 |
2005–2007 | |||||||||||
Denawaka Ganga | Lellopitiya | 4.7 | 1958–1966 | 0.08 | 388 | 2 | 84 | 14 | 305 | 75 | |
Wey Ganga | Dela | 15.5 | 1956–2006 | 0.24 | 322 | 2.1 | 53 | 5 | 335 | 220 |
The Kalu Ganga is the second largest river in Sri Lanka with a catchment area of 2,690 km2. The river basin lies entirely within the wet zone where the mean annual rainfall is 4,000 mm, ranging from 2,750 mm on the coast to 5,250 mm in the mountains. The Upper Kalu Ganga basin (615 km2) was selected for this study. Two tributaries (Denawaka Ganga and Wey Ganga) and the trunk of the Upper Kalu Ganga, above Rathnapura town, were selected for suspended sediment sampling (Figure 1).
River discharges are currently measured by the Irrigation Department based on river stage and discharge rating curves. Stage is measured every hour in day time by local employees in selected tributaries. The stage height is converted to river discharge by using the relationship between stage height and discharge. For the studied eight tributaries, the monthly discharge data were accessible for a period of 3–45 years. Out of them, only for three tributaries (Rakwana Ganga, Wey Ganga and Upper Kalu Ganga) were the daily discharges or stage heights obtainable for the exact dates where water samples were collected to measure suspended load concentrations. Daily rainfall in the basins is recorded by the Department of Meteorology, Sri Lanka.
Measurement of SSCs
The total sediment load in a river is the sum of the suspended sediment load and the bed load. In most global rivers, the bed load is generally lower than the suspended sediment load and represents a fraction of only ca. 10% from the total sediment load (Summerfield & Hulton 1994). The majority of the sediment gauging studies carried out world-wide rely on calculating the fraction of bed load from the estimated suspended load instead of taking direct measurements since uncertainties involved with direct measurements of bed load is large (Summerfield & Hulton 1994). For the catchments selected for this study, the bed load is about 20% of the total load according to the relationship deduced by Turowski et al. (2010) and the total sediment load can easily be detected if the suspended sediment load is known. Therefore, monitoring SSC and discharge in the tributaries is a prerequisite to obtain a representative estimate of the total sediment load.
Thus, in order to monitor SSC for each selected tributary, a 1 L water sample from the water surface was collected by a well-trained villager from a predetermined location close to the river bank every day (Figure 2). If the water samples were collected very close to the river bank, the possibilities are higher in sampling inputs of sediment fluxes derived locally from nearby bank erosion that would not represent the natural distribution of suspended load within the river cross-section. Hence, collecting a sample as close as possible to the center of the river is recommended to minimize the disturbance in natural distribution. While seeing this fact and also considering the accessibility for sample collection, the predetermined location on the water surface for sampling is traced about 3 m away from the bank along the surface towards the center. The surface water sampling campaign was continued for one hydrological year because river discharge varies markedly within a single hydrological year in parallel to changes in the monsoonal precipitation (Figure 3). Previous studies also stressed the importance of monitoring SSC over a longer period of time in order to obtain a representative value of SSC, especially during the storm periods (Lewis & Rasmussen 1999). Thus, to obtain a representative yearly figure, continuous monitoring of SSC over one hydrological year is mostly recommended (Ndomba et al. 2008). In addition, depth-and-width-integrated water samples were obtained once a month since daily surface water samples do not represent the true cross-sectional SSC of the tributary. To obtain a depth-and-width-integrated suspended sediment sample, the equal-width increment method was performed using a depth-integrated suspended sediment sampler. The river cross-sections and the divided increment in all tributaries in the Walawe and Kalu Ganga basins are shown in Figures 4 and 5, respectively. At the same time, a surface water sample was also collected at the same place of predetermined. The samples were returned to the laboratory at the Sabaragamuwa University of Sri Lanka every month. They were kept for 12 hours to allow the sediments to settle in the bottle, and the excess water was carefully decanted. After that, samples were oven dried, and the SSC in each 1 L sample was measured using an analytical balance. If the SSCs were needed to convert sediment fluxes in order to calculate spatially-averaged erosion rates in catchment scale, suspended sediment loads of the inorganic fraction are required. Hence, a few randomly selected samples from all tributaries were treated with H2O2 to decompose the organic matter and the content of organic material in the suspended load was determined. It was found that the content of organic matter in the suspended load is very low and negligible. However, the organic matter content in the suspension may vary from sample to sample, from site to site and also from season to season and thus may cause a considerable contribution in some samples. But, in comparison to other uncertainties associated in the monitoring program, correcting the suspended load for organic matter can be considered as unnecessary. Otherwise, in order to correct the suspended load for organic matter, the samples need to be treated with H2O2 one by one, which is a costly and time-consuming effort.
Using surface and depth-and-width-integrated SSCs obtained on a monthly basis, site-specific calibration curves were constructed (Figures 2 and 6). Based on these relationships, daily surface SSC was converted to the depth-and-width-integrated SSC. The differences between the surface SSC and depth-and-width-integrated SSC were calculated for each tributary. Furthermore, a generalized calibration curve was constructed using surface and depth-and-width integrated measurements of SSC in all catchments (Figure 6).
Suspended sediment rating curves and sedimentographs
Morphological factors affecting the suspended sediment transportation
The main factors that are believed to affect the suspended sediment transport within the river profile were analyzed based on available 1:50,000 digital topographic maps with 90 m resolution using Arc-GIS 9.2 software (Figure 2 and Table 1). The mean local relief (MLR) was obtained by averaging the difference between maximum and minimum elevation within 600 × 600 m grids across the catchments. The drainage density (DD) was calculated as the total stream length of the basin divided by the basin area. In addition, river gradient (RG) along the main trunk was obtained from the elevation differences between source and outlet of the basin divided by trunk channel length along the river. The land use of the selected basins was categorized and the area percentage of forest (FP) was obtained based on land use data in year 2001 (Department of Survey, Sri Lanka); mean elevation (ME) was calculated from the elevations of each grid intersection. However, the effects of MLR and ME on suspended sediment transportation in rivers is approximately similar. Therefore, the most significant topographic factor of MLR is used in the subsequent sections when the data are interpreted. However, the values of ME are also presented in the Table 1 to characterize the catchments.
RESULTS AND DISCUSSION
Suspended sediment concentrations
The surface and depth-and-width-integrated sediment concentrations are summarized in Table 2. Comparatively high standard deviations were observed in the measured surface SSC in all catchments (Table 2). This could be attributed to the seasonal fluctuations in the rainfall intensity over the hydrological year (Figures 3 and 8). The highest standard deviation was recorded in the Wey Ganga, a tributary in the Kalu Ganga basin, which has the highest level of human impact in the studied catchments. The Wey Ganga tributary is mostly affected by gem mining within the Kalu Ganga basin, as clearly observed during frequent field visits to the area. SSC was, accordingly, highly variable, controlled not only by seasonality in weather but also by human activities. Variations in mean annual depth-and-width-integrated SSC across the river profiles in the tributaries of Walawe and Kalu Ganga basins are illustrated in Figures 4 and 5, respectively.
. | . | Surface SSC (g/L) . | Depth – Integrated SSC (g/L) . | . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|---|
River Basin . | Tributary . | Minimum . | Maximum . | Mean with standard deviation . | Minimum . | Maximum . | Mean with standard deviation . | R2 of site specific relation . | r value . | P value . |
Walawe | Belihul Oya | 0.000 | 0.080 | 0.005 ± 0.01 | 0.000 | 0.071 | 0.005 ± 0.009 | 0.99 | 0.998 | 0.000 |
Diyavini Oya | 0.001 | 0.170 | 0.022 ± 0.02 | 0.005 | 0.040 | 0.021 ± 0.016 | 0.72 | 0.851 | 0.002 | |
Weli Oya | 0.001 | 0.340 | 0.024 ± 0.054 | 0.008 | 0.297 | 0.028 ± 0.046 | 0.92 | 0.957 | 0.000 | |
Rakwana Ganga | 0.001 | 0.530 | 0.034 ± 0.048 | 0.004 | 0.48 | 0.034 ± 0.044 | 0.85 | 0.919 | 0.000 | |
Hulanda Oya | 0.001 | 0.620 | 0.05 ± 0.072 | 0.006 | 0.75 | 0.064 ± 0.090 | 0.84 | 0.913 | 0.000 | |
Kalu | Kalu Ganga | 0.001 | 0.620 | 0.018 ± 0.043 | 0.002 | 0.741 | 0.022 ± 0.051 | 0.43 | 0.660 | 0.038 |
Denawaka Ganga | 0.010 | 0.400 | 0.042 ± 0.037 | 0.013 | 0.0387 | 0.043 ± 0.038 | 0.87 | 0.931 | 0.000 | |
Wey Ganga | 0.01 | 2.81 | 0.25 ± 0.353 | 0.016 | 2.76 | 0.247 ± 0.346 | 0.99 | 0.993 | 0.000 |
. | . | Surface SSC (g/L) . | Depth – Integrated SSC (g/L) . | . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|---|
River Basin . | Tributary . | Minimum . | Maximum . | Mean with standard deviation . | Minimum . | Maximum . | Mean with standard deviation . | R2 of site specific relation . | r value . | P value . |
Walawe | Belihul Oya | 0.000 | 0.080 | 0.005 ± 0.01 | 0.000 | 0.071 | 0.005 ± 0.009 | 0.99 | 0.998 | 0.000 |
Diyavini Oya | 0.001 | 0.170 | 0.022 ± 0.02 | 0.005 | 0.040 | 0.021 ± 0.016 | 0.72 | 0.851 | 0.002 | |
Weli Oya | 0.001 | 0.340 | 0.024 ± 0.054 | 0.008 | 0.297 | 0.028 ± 0.046 | 0.92 | 0.957 | 0.000 | |
Rakwana Ganga | 0.001 | 0.530 | 0.034 ± 0.048 | 0.004 | 0.48 | 0.034 ± 0.044 | 0.85 | 0.919 | 0.000 | |
Hulanda Oya | 0.001 | 0.620 | 0.05 ± 0.072 | 0.006 | 0.75 | 0.064 ± 0.090 | 0.84 | 0.913 | 0.000 | |
Kalu | Kalu Ganga | 0.001 | 0.620 | 0.018 ± 0.043 | 0.002 | 0.741 | 0.022 ± 0.051 | 0.43 | 0.660 | 0.038 |
Denawaka Ganga | 0.010 | 0.400 | 0.042 ± 0.037 | 0.013 | 0.0387 | 0.043 ± 0.038 | 0.87 | 0.931 | 0.000 | |
Wey Ganga | 0.01 | 2.81 | 0.25 ± 0.353 | 0.016 | 2.76 | 0.247 ± 0.346 | 0.99 | 0.993 | 0.000 |
The site-specific calibration for each tributary was constructed and is illustrated in Figure 6. All plots have positive linear relations and gradient of the graph or ‘m’ and intercept or ‘c’ vary from tributary to tributary. The generalized site-specific calibration for Sri Lankan tributaries is also given in Figure 6. The graph shows a strong correlation in between surface and depth-and-width-integrated SSC with an R2 of 0.91. Moreover, the relationship indicates that surface sampling was underestimated at high SSC, during the period of high rainfall or discharge (Figure 8). When the discharge is high, the sediment distribution within the river profile is inhomogeneous, hence the difference between surface and depth-and-width-integrated SSC is significant. In contrast, at medium and low SSC and discharges, the difference between the surface and depth-and-width-integrated SSC becomes minor since the sediment distribution within the river profile is reaching homogeneity. A possibility exists to apply this generalized model in order to convert easily monitored surface SSC into more a representative depth-and-width-integrated SSC for the Sri Lankan tributaries and also for other tributaries elsewhere in which similar conditions prevail.
To determine the correlation strength of site-specific calibrations, the correlation test was performed at 95% confidence level and resulting r and P values are given in Table 2. The results clearly illustrate that the studied tributaries have a stronger correlation between surface and depth-and-width-integrated SSC except for the Kalu Ganga at Malwala. The Kalu Ganga at Malwala is the largest tributary monitored in this study with a stream width of 77 m at the gauging station. The stream widths of all other tributaries are relatively narrow ranging from 13 to 47 m. In addition, this is the deepest tributary studied where sampling required a small boat. Hence, the increment interval that was used to collect depth integrated samples by the equal-width increment method is likely to be insufficient and the sampling method may also not be adequately accurate to obtain a satisfactory site-specific relationship for the Kalu Ganga at Malwala. Therefore, we exclude this tributary in subsequent interpretations.
The cross-sectional representative SSC was estimated based on the site-specific calibration relationship. The resulting depth-and-width-integrated SSCs in the tributaries of Walawe and Kalu Ganga basins are presented in Table 2.
Suspended sediment rating curve and sedimentographs
The rating curves were constructed by plotting depth-and-width-integrated and surface SSCs against either river discharge or river stage height (Figure 7). The rating curves for our study catchments show a weak relationship between SSC and river discharge/stage height. R2 values are low and vary between 0.5 and 0.23. In general, elsewhere in the world, most rivers have shown a remarkable correlation between sediment concentration/load and river discharge or river gauge heights. However, some rivers, for instance the Colorado River in the USA, have not shown a significant relationship between SSC and water discharge due to several reasons. One of the major reasons for a poor relationship is the complex sources of water and sediment in the river. Seasonally complex sources can reflect irrigation and other human interventions upstream (Syvitski et al. 2000; Walling 2005; Zhang et al. 2006, 2009; Vanacker et al. 2007; Lu et al. 2012). Mining activities can also be a significant influence on sediment loads (Latrubesse et al. 2005; Walling 2005). In a similar fashion, the rating curves for our study catchments are poorly correlated, which can be attributed to seasonally complex sediment inputs to the fluvial system.
The sedimentograph of each tributary, which demonstrates the variation of SSC with time over the period of gauging, are shown in Figure 8. As illustrated, the higher SSC for these tributaries is observed during the intense rainy and high discharge periods from October to December. This observation is comparable to what is observed in other tropical rivers elsewhere in the world where sediment deliveries and discharges are in the peak in wet seasons (Van Maren & Hoekstra 2004).
Difference between surface and depth-and-width-integrated SSC
The differences between the depth-and-width-integrated SSC and the surface SSC were calculated for each tributary and a clear discrepancy was observed for them in some tributaries. The difference as a percentage was calculated with respect to the annual mean surface SSC (Figure 9), which clearly reveals the existence of three different scenarios. The first scenario corresponds to the situation in which the overall difference between surface and depth-and-width-integrated SSCs is comparatively low at all surface SSCs. This situation was observed in Wey Ganga and Denawaka Ganga at Dela and Lellopitya gauging stations, respectively (Figure 9(a)). When the differences in these two tributaries are analyzed more precisely by measurement to measurement, surface SSC measured during the time at low surface SSC levels is slightly lower than the depth-and-width-integrated SSC. During the periods of high surface SSC levels, the measured SSC by surface sampling is observed to be slightly greater than that of the SSC measured by depth-and-width-integrated method (Figure 9(a)). The maximum difference during the time at low SSC in Wey Ganga and Denawaka Ganga were 2.4 and 4%, respectively (Figure 9(a)) whereas the maximum difference during the time at high SSC of the same is 20 and 16% (Figure 9(a)), respectively. However, as an annual average, the difference in SSCs measured by using the two methods of surface sampling and depth-and-width-integrated sampling in these two tributaries is found to be negligible over the period of sampling. The second scenario was experimental in Hulanda Oya at Modarawana and Kalu Ganga at Malwala, where all measured depth-and-width-integrated SSCs were greater than the surface SSC. In addition, when the surface SSC increases, the difference between the surface and depth-and-width-integrated SSCs also increases (Figure 9(b)). This observation highlights the importance of conducting a depth-and-width-integrated sampling together with surface sampling. The third situation was observed in Belihul Oya, Rakwana Ganga, Weli Oya and Diyavini Oya. Here, in large number of samples (ca. 90%), depth-and-width-integrated SSC is greater than the surface SSC as seen in Hulanda Oya and Kalu Ganga. In addition, at a particular value, depth-and-width-integrated and surface SSC are almost equal and afterwards surface SSC has increased in size than the depth-and-width-integrated SSC (Figure 9(c)). The maximum difference of the depth-and-width-integrated SSC minus surface SSC are 4, 13, 24, and 9% for Belihul Oya, Diyavini Oya, Weli Oya and Rakwana Ganga, respectively. A discrepancy of the surface SSC minus depth-and-width-integrated SSC is also observed in these four tributaries, but only in a limited number of samples (ca. 4–10% in ca. 350 measurements) even though it is hypothetically unlikely. This could be attributed to sampling or any other errors associated during the monitoring or any other unknown anthropogenic contribution to the sediment exports in rivers. The reasons behind these three different phenomena may depend on whether a sediment laden tributary enters the main river a short distance up stream, how turbulent the flow is at the sampling point and relative flow velocity of the river at different sampling gauges. However, it is difficult to provide the roots and mechanisms that initiate the observed scenarios at this stage without a study on fine focusing, which is beyond the scope of this paper.
It is imperative to mention that the above discussion on whether differences between surface and depth-and-width-integrated sampling are negligible and/or important is based solely here on monitoring of rivers covering only one hydrological year. It is obvious that hydrological conditions of the rivers prevailed during the time of monitoring will not remain the same over time and is subjected to change with climatic fluctuations and anthropogenic perturbations. Therefore, surface sediment sampling together with occasional depth-and-width-integrated sampling is recommended over the monitoring time in future in order to ensure the stability of the relationships and/or if needed, to readjust the relations.
Main factors affecting suspended sediment transportation
The gradient (m) of the resulting graphs presented in Figure 6 is explained as the relation between surface and depth-and-width-integrated SSC. When m = 1 and for a constant intercept, the difference between the surface and depth-and-width-integrated SSCs is zero, meaning that the distribution of suspended sediment is uniform throughout the river's depth. When m deviates from 1, the difference between surface and depth-and-width-integrated SSCs becomes significant, implying that only surface sampling is not sufficient for a reliable estimate. Therefore, at this stage, our effort is to develop a relationship between the gradient (m) of the site-specific graphs and first order morphological factors (Table 1) that are believed to influence the SSC in rivers while introducing a factor called the ‘SSTF’.
SSTF – concept and methodology
Different geomorphological and environmental factors that affect the transportation behavior of sediments within a river profile are subdivided into two groups in this study. The factors that change the stream flow to more turbulent from laminar leading to increase the difference between the depth-and-width-integrated SSC and surface SSC are listed under ‘Group 1’. The other set of factors (‘Group 2’) shifts stream flow from turbulent to laminar, subsequently decreasing the difference between the depth-and-width-integrated SSC and surface SSC.
We therefore derive the following relationship assuming that a total of ‘n’ factors in the fluvial system belong to Group 1. In Figure 10, gradient of site-specific calibration is a fixed value (G) for all graphs in Group 1 under the conditions of , and for i = 1 …… n.
In Equation (9), ‘A’ denotes the factors that change the stream flow to more turbulent from laminar, thereby increasing the difference between the depth-and-width-integrated SSC and surface SSC. The factors of B influence the stream while shifting its flow from turbulent to laminar, decreasing the difference between the depth-and-width-integrated SSC and surface SSC.
SSTF – application to the studied catchments
The ‘SSTF’ plot (Figure 11) shows a strong relation (R2 = 0.81, r = –0.89, p = 0.006) between difference of depth-and-width-integrated and surface SSCs or ‘(1–m)’ value and surface SSTF value derived for each tributary, even though the number of samples used for this plot is limited. Moreover, it has been revealed that ‘(1–m)’ = 0 (i.e. the difference between surface and depth-integrated SSCs become zero) when the SSTF is around 40. When SSTF <40, surface SSC is greater than the depth-and-width-integrated SSC implying the surface SSC tends to be overestimated (Figure 11). In contrast, when the SSTF >40, surface SSC is lower than the depth-integrated SSC implying the surface SSC is likely to be underestimated.
Now, we focus our discussion on calculating an uncertainty associated in measuring the SSC by surface sampling instead of depth-and-width-integrated sampling. This error, the deviation of surface SSC from depth-and-width-integrated SSC, is to be x% of the depth-and-width integrated SSC. The x% represents the uncertainty that caused mainly by the inhomogeneity of SSC across the river profile, which is a fraction of the total error related in the SSC. The temporal inhomogeneity of the SSC, sampling uncertainties and analytical uncertainties are the other main contributors that make an error in the SSC.
Accordingly, Wey Ganga and Denawaka Ganga have inconsequential differences between surfaces and depth-and-width-integrated SSCs and ‘m’ values are 0.96 and 0.98 respectively. Based on this interpretation, we propose that the difference between surface and depth-and-width-integrated SSCs becomes negligible when SSTF values range approximately from 35 to 45, with corresponding ‘m’ values ranging from 0.97 to 1.03 (Figure 11). For such tributaries, surface sampling would be sufficient to infer a representative SSC.
It should be mentioned here that the preceding discussion and the preliminary values presented for SSTF is based on the analysis of suspended load data from seven small catchments covering one hydrological year and considering only four morphological factors. It is indeed necessary to increase the number of gauging stations to as many as possible and increase the length of the gauging period to as long as possible to establish a better relationship between surface SSC and depth-and-width-integrated SSC aiming at an increased number of samples. Also, it is important to incorporate all the influencing factors that are responsible for suspended sediment transport behavior in the stream to obtain more reliable values for SSTF. Therefore, this preliminary study in the tropical highlands of Sri Lanka merely proposes a conceptual framework in applying a SSTF as a novel idea when a river load gauging program is planned.
CONCLUSIONS
It is obvious that the suspended load concentration of rivers is not homogeneous across the river cross-section and therefore depth-and-width-integrated sampling is a necessity to obtain a realistic estimate of suspended loads in rivers. However, monitoring of rivers for depth-and-width-integrated suspended concentrations on a regular basis is costly and not feasible for countries with financial constraints. Therefore, in this study we have successfully introduced a time-saving, cost-effective but reliable approach to estimate depth-and-width-integrated SSC in tributaries in the tropical highlands of Sri Lanka. The technique relies on the site-specific relationship between the daily monitored surface sediment concentration and the monthly monitored depth-and-width-integrated suspended load concentration developed individually for monitored tributaries. Moreover, an overall correlation was obtained for surface and depth-and-width-integrated SSC (R2 = 0.91) as a generic model that may have broader applicability to calculate a more robust SSCs for rivers in Sri Lanka or even elsewhere that possess the same geomorphologic settings.
Furthermore, we evaluated the necessity of depth-and-width-integrated suspended sampling for tributaries while introducing a threshold factor called ‘SSTF’, which was derived by considering the morphological factors of a particular catchment. In the prevalence of a large number of river basins and even with thousands of creeks and tributaries, detailed monitoring of all branches of the river system for suspended load concentrations is not realistic. Therefore, the rating curves that are constructed based on the relationship between SSC and the river gauge heights/discharge data are usually used to predict the un-gauged SSC. However, in our study, the rating curves constructed for small catchments in the Central Highlands of Sri Lanka show poor correlations, highlighting the risk of using them to calculate SSCs in humid catchments with seasonal rainfall variability. Therefore, continuous surface suspended load monitoring together with or without regular depth-and-width-integrated monitoring depending on the ‘SSTF’ is proposed as an easy and economically viable method to estimate accurate annual loads of suspended loads in small tropical catchments.
ACKNOWLEDGMENTS
The authors gratefully acknowledge the staff of the Irrigation Department, Colombo, Sri Lanka, for providing hydrological data and National Science Foundation (NSF), Sri Lanka, for awarding a research grant (RG/2005/DMM/04) to TH to conduct this research. We wish to express our sincere gratitude to Hasitha Baduraliya for his support in mathematical formulations and Nishantha Attanayake for gainful discussions. We are grateful for the in-depth reviews by three anonymous reviewers and the handling editor.