TXT

improved landuse cover on the range of modelled sediment yield...

By Herman Rodriguez,2014-08-25 18:39
11 views 0
improved landuse cover on the range of modelled sediment yield...

     ??ÎÄÓÉsgne??Ï×

    pdfÎĵµ?ÉÄÜÔÚWAP?Ëä?ÀÀÌåÑé???Ñ????ÒéÄúÓÅÏÈÑ?ÔñTXT???òÏÂÔØÔ?ÎÄ?þµ????ú?é????

     Available online at www.sciencedirect.com

     Mathematics and Computers in Simulation 78 (2008) 367?C378

     The impact of improved landuse cover on the range of modelled sediment yield from two sub-catchments of the Mae Chaem, Thailand

     M.G. Hartcher a,? , D.A. Post b

     b a CSIRO Land and Water, Davies Laboratory, Townsville, Australia CSIRO Land and Water, Black Mountain Laboratories, Integrated Catchment Assessment and Management Centre (iCAM), Australian National University, Canberra, Australia

     Available online 18 January 2008

     Abstract A sediment source, transport, and deposition model known as SedNet has been applied to two sub-catchments of the Mae Chaem River in Thailand, the Mae Suk (95 km2 ) and the Mae Kong Kha (91 km2 ). The applied models were analysed to determine the dominant sources and sinks of suspended sediment in these catchments, and to examine the impact of an improved landuse coverage on the range of modelled sediment loads. SedNet model scenarios were run using landuse grids from 1995 to 2003. The 1995 landuse classi?cation was derived from satellite imagery and contained a relatively undifferentiated landuse classi?cation (lumping all forest types into one category); the other in 2003 was derived from a mix of satellite imagery, ground truthing, and mapping, and better differentiated between landuse types (dividing forest types into evergreen, deciduous, and pine plantations). Results indicate signi?cant differences in predicted sediment export in 2003 compared to 1995. It is dif?cult to say whether these changes were due to actual changes in landuse between 1995 and 2003, or due to the improved landuse classi?cation in 2003. The source areas of suspended sediment also changed signi?cantly between 1995 and 2003, and these changes in source area can clearly be linked to the improved mapping of landuse in 2003. The improved landuse classi?cation in 2003 also led to signi?cant reductions in the range of possible suspended sediment exports from both catchments. Despite these improvements in landuse classi?cation, signi?cant uncertainty in predicted suspended sediment yield still exists for both sub-catchments. Further improvements in identifying the total volume and source areas of suspended sediment will best be achieved through an improved landuse coverage which identi?es the type of crop being grown. ? 2008 IMACS. Published by Elsevier B.V. All rights reserved.

     Keywords: SedNet; Landuse; Erosion; Thailand

     1. Introduction The Mae Chaem catchment is approximately 3900 km2

    in area, and is located in the north-western region of Thailand forming part of the Ping drainage basin. The catchment is representative of large areas of Southeast Asia, where intense competition for land and water use requires management options which maintain socio-economic opportunities yet

     Corresponding author. Tel.: +61 7 4753 8695; fax: +61 7 4753 8650. E-mail addresses: Michael.Hartcher@csiro.au (M.G. Hartcher), David.Post@csiro.au (D.A. Post).

     0378-4754/$32.00 ? 2008 IMACS. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.matcom.2008.01.012

     368

     M.G. Hartcher, D.A. Post / Mathematics and Computers in Simulation 78 (2008) 367?C378

     minimise environmental problems such as erosion, low dry season ?ows, and water pollution [8]. The population in the catchment in 1994 was approximately 92,000 comprising 49,000 Thai locals and 43,000 hill tribe people originating from Laos and Myanmar (Burma). The Mae Chaem catchment is a relatively steep catchment ranging from 250 to 2570 m elevation, with small narrow ?oodplains. Rainfall is highly variable from year to year with 95% of yearly rainfall occurring in the wet season from May to October [11]. Population pressure on the landscape from expanding agriculture is a critical factor, with hillslope erosion due to forest clearance a dominant issue for the region. The major crops grown in the region are rice, maize, vegetables, and tree crops. Due to a combination of landscape classi?cation and forest zoning policies, there is little remaining land available for development [8]. A number of studies have also focused on catchment resources and hydrologic response to landuse change in the Mae Chaem catchment, including Perez et al. [10], Merritt [8], Merritt et al. [9], Croke et al. [1], and Hartcher et al. [4]. Two of the Mae Chaem sub-catchments, namely, the Mae Suk (95 km2 ) and Mae Kong Kha (91 km2 ), are representative of typical landuse and have been classi?ed at a ?ner resolution than for the whole of Mae Chaem. Fig. 1 shows the location of the two sub-catchments within the Mae Chaem catchment, Thailand. These ?ner resolution landuse maps were derived in different ways with the 1995 landuse coming from satellite imagery and the 2003 landuse from a mix of satellite imagery, ground truthing, participatory mapping, and better differentiation between landuse types. The 2003 landuse classi?cation also contained a more detailed representation of forest type than the 1995 coverage. This paper presents SedNet modelling results for landuse scenarios describing the Mae Suk and Mae Kong Kha sub-catchments in 1995 and 2003.

     Fig. 1. Location of Mae Suk and Mae Kong Kha sub-catchments within Mae Chaem catchment.

     M.G. Hartcher, D.A. Post / Mathematics and Computers in Simulation 78 (2008) 367?C378

     369

     2. The SedNet model The acronym SedNet stands for the Sediment River Network Model. SedNet is a software package originally developed by CSIRO for use in the Australian National Land and Water Resources Audit for use in assessing water quality in the major catchments throughout Australia [12]. It is now being applied at regional scales such as river catchments, using more detailed inputs [2,4,13]. SedNet models estimate river sediment loads by constructing material budgets that account for the main sources and stores of sediment. SedNet models use a simple mean annual conceptualisation of transport and deposition processes in streams. Information on SedNet model development is detailed in a series of CSIRO Land and Water technical reports and other related publications such as Gallant [3], Lu et al. [7], Prosser et al. [12], and Young et al. [14]. On-line documentation and SedNet software is available via the CRCCH toolkit website [15]. 3. Methods and data The base data for Mae Chaem, such as the digital elevation model (30 m DEM), landuse, stream ?ow, and rainfall grids, as well as the 2003 landuse for the Mae Suk and Mae Kong Kha sub-catchments were supplied by the World Agroforestry Centre at Chang Mai University, and the 1995 landuse was supplied by the Land Development Department in Thailand. 3.1. Stream links and watershed The basic unit of a SedNet Model is a stream link. These stream links are generated automatically from the DEM. Topology was created for each stream link to identify its upstream and downstream relationship to other stream links and its overall position within the system (stream order). For each stream link, a unique watershed is identi?ed by a polygon area. The watersheds, as well as providing measurement of upstream catchment area for hydrological

    parameterisation, de?ne the areas within which spatially distributed erosion data need to be summarised for each stream link [5]. Fig. 2 illustrates the suspended sediment budget of a river link within SedNet. 3.2. Hydrological setting To run SedNet, hydrological parameters for prediction of sediment transport and deposition within the river system need to be estimated and attached to each stream link. In general, channel width, mean annual ?ow and bank full discharge are generally only known in a few places, so regionalized values were created based on potential evapotransporation/rainfall ratios and catchment area. As with all other SedNet studies, connectivity, channel gradients, and stream order information were derived during stream link creation within the toolkit. There are no signi?cant reservoirs or lakes in the Mae Chaem catchment, while some ?oodplains occur in the lower Mae Suk sub-catchment, but none in the Mae Kong Kha. These are the areas which SedNet models as depositional, where ?ne sediment may potentially be

    deposited. Coarse sediment deposition in the model occurs in-stream, based on transport capacity estimates.

     Fig. 2. Conceptual diagram of the SedNet river sediment budget for one river link [13].

     370

     M.G. Hartcher, D.A. Post / Mathematics and Computers in Simulation 78 (2008) 367?C378

     3.3. Gully and stream bank erosion As the purpose of this study was to investigate the effect of improved landuse classi?cation on predicted hillslope delivery of sediment, gully and stream-bank erosion, and the contribution from roads and landslides, were not considered. However, anecdotal evidence suggests that rates of gully and stream-bank erosion in this catchment are low, while there is evidence of landslides, although their signi?cance is not certain. 3.4. Hillslope erosion Hillslope erosion was estimated using the revised universal soil loss equation (RUSLE) where: soil loss (t/ha/year) = R ?Á K ?Á L ?Á S ?Á C ?Á P (1)

     R is the rainfall erosivity factor, K the soil erodibility factor, L S the hill length/slope factor, C the vegetation cover factor, P the landuse practice factor (not used). All factors were represented as spatially variable grids (30 m cells), allowing for derivation of a spatially distributed hillslope erosion grid. An additional term, the hillslope delivery ratio (HSDR) is also used to account for re-deposition of hillslope sediment before it reaches a stream since not all of the sediment that is eroded from a hillslope makes its way into a stream. The total sediment delivered to a stream depends on both the hillslope erosion and on the Hillslope Delivery Ratio (HSDR), such that: Total sediment delivered to stream = soil loss ?Á HSDR. (2)

     3.4.1. Rainfall erosivity factor (R) Rainfall erosivity is a measure of the intensity of rainfall events and so is determined by climatic data. For Mae Chaem we used an annual average value based on the existing monthly rainfall grids. The grid cells used in the available rainfall data were 1 km. The average annual rainfall grid had the following equation applied to create a rainfall erosivity grid: R = 38.5 + 0.35P P represents mean annual precipitation [8]. 3.4.2. Erodibility factor (K) Erodibility is a measure of the susceptibility of the soil to erosion. It is based on the nature (structure, texture, etc.) of the topsoil. A K-factor grid was supplied by Chiang Mai University based on existing soils data. 3.4.3. Hill slope/length factor (LS) The hillslope factor accounts for the fact that soil erosion increases with increasing slope. A grid of slope in degrees was created from the existing DEM. Length of slope was not incorporated, and so slope length was left as a constant value (=1). 3.4.4. Cover factor (C) The 1995 landuse, supplied by the Land Development Department in Thailand, was classi?ed using Landsat

    TM imagery. This unsupervised classi?cation appears to have created errors such as tree shadow and hill shading effects, and has resulted in more disaggregated landuse classes. The 2003 landuse grids for the two sub-catchments, supplied by ICRAF at Chiang Mai, were created by overlaying the map outputs from participatory mapping with the forest maps (from surveying) for the year 2003, then performing ?eld surveys to verify the classi?cation. Figs. 3 and 4 compare the 1995 and 2003 landuse classi?cations for both Mae Suk and Mae Kong Kha respectively. Cover factors for each landuse were taken from an existing table of ??Crop Management Factors?? for Thailand [8]. The C-factor represents a comparison of soil loss with that expected from freshly tilled soil and has a range between 0 and 1 where higher values mean more erosion. Some updated cover factors were given to fallow and forest types, however the old cover values were still applied to the 1995 classi?cation, as they were based on the best knowledge at that time. (3)

     M.G. Hartcher, D.A. Post / Mathematics and Computers in Simulation 78 (2008) 367?C378

     371

     Fig. 3. Comparison of 1995 and 2003 landuse classi?cation for Mae Suk sub-catchment.

     3.4.5. Landuse practice factor (P) This accounts for the effects of contours, strip cropping or terracing. As data on these are not available for the Mae Chaem catchment, this factor was not used (i.e. set to 1), although it may be accounted for to some degree in the choice of C-factors. 3.4.6. Hillslope delivery ratio (HSDR) The HSDR is traditionally set as a constant value in the SedNet model. However, it is recognised that factors such as slope, vegetation cover, and distance from stream can all affect the HSDR [5]. HSDR is a number between 0 and 1 where 0 means that none of the sediment eroded from a hillslope is delivered to a stream, and 1 means that all of the sediment eroded from a hillslope is delivered to a stream. In practice, HSDR is typically between 0 and 0.2. We therefore based HSDR on the empirical observation that hillslope erosion occurring close to streams is more likely to ?nd its way into a stream than sediment eroded at a distance from streams. The exact nature of this relationship is still under investigation [6]. In this study, similarly to [5], we chose to reduce HSDR exponentially with distance from stream according to: HSDR = 0.2844 ?Á e?9.1?Á10

     4d

     (4)

     d is the distance from stream. This relationship is shown in Fig. 5. This equation was chosen because it provides an average HSDR of 0.1 for the whole of the Mae Chaem catchment. It provides a HSDR of approximately 0.12 for both the Mae Suk and Mae Kong Kha sub-catchments.

     372

     M.G. Hartcher, D.A. Post / Mathematics and Computers in Simulation 78 (2008) 367?C378

     Fig. 4. Comparison of 1995 and 2003 landuse classi?cation for Mae Kong Kha sub-catchment.

     Fig. 5. The assumed relationship between HSDR and distance from stream.

     4. Total sediment loads The ?rst component of this study was to examine how improving the landuse cover classi?cation might affect the total volume and source areas of predicted sediment sources within our study areas. As can be seen in Figs. 6 and 7 there is a signi?cant difference in the hillslope erosion grids between the 1995 and 2003 landuse classi?cations for both

     M.G. Hartcher, D.A. Post / Mathematics and Computers in Simulation 78 (2008) 367?C378

     373

     Fig. 6. Hillslope erosion grid for Mae Suk.

     sub-catchments. It appears that the 1995 classi?cation, which was based on satellite imagery alone, has a generally more even spread of erosion rates, although it also showed overprinting effects from hill shadow in the image classi?cation process. The 2003 classi?cation results in the hillslope erosion classes being more uniform in speci?c areas and clearly demarcating landuse differences not identi?ed in 1995. The sub-catchment results (Tables 1 and 2) indicate that we ?nd greater precision (although not necessarily accuracy) in identifying sediment sources where more discrete landuse classes are given. The breakdown of forest from the 1995 classi?cation, into 4 or 5 discrete classes in the 2003 classi?cation shows that ??hill evergreen forest?? comprises around 30% of each sub-catchment and approximately 50% of total forest, yet has a relatively small predicted sediment yield with a cover factor of only 0.003. In the 1995 classi?cation we were unable to differentiate between forest types and therefore had to apply a single cover factor of 0.020 for all forest types. The improved classi?cation has cover factors for forest classes ranging from 0.003 to 0.088. However, the lack of discrete classes for the ?eld crop and fallow landuses has left many questions unresolved, relating to the effects of crop choice and fallow practices on sediment yield. In 1995, fallow and ?eld crops were predicted to contribute approximately 59 kt/year (87%) of sediment for Mae Suk (Table 1). In 2003 the fallow, ?eld crop, and fruit tree landuse were predicted to only contribute 43 kt/year (77%) of total predicted sediment yield for Mae Suk. This reduction may be attributed to a lower cover value for fallow, even though it appears from Fig. 3 that some fallow has become permanent ?eld crops. In the Mae Kong Kha 1995 classi?cation, fallow and ?eld crops contributed approximately 21 kt/year (70%) of sediment. In the 2003 classi?cation, ?eld crop and

    fruit trees contributed 19 kt/year (54%) of total sediment yield, and there was no fallow due to the introduction of permanent ?elds (which are not left fallow). In reality the ?eld crops are probably quite variable, and may have a variety of different management practices resulting in different C-factors. In

     374

     M.G. Hartcher, D.A. Post / Mathematics and Computers in Simulation 78 (2008) 367?C378

     Fig. 7. Hillslope erosion grid for Mae Kong Kha.

     the case of the Mae Kong Kha, the 2003 landuse classi?cation, which involved more ground truthing surveys, has likely yielded a more accurate depiction of crops than the 1995 classi?cation, which was based on satellite classi?cation with limited ground truthing. The total predicted sediment yield for the Mae Suk has decreased from 67.79 kt/year in 1995 to 56.58 kt/y in 2003. This is probably due to a lower fallow C-factor of 0.100, as advised by Chiang Mai University, compared with the 1995 value of 0.200. Conversely, the total predicted sediment yield for the Mae Kong Kha has increased from 30.26 kt/y in 1995 to 35.94 kt/year in 2003. This appears to be the result of improved forest classi?cation where mixed and dry deciduous forests, which comprise approximately 30% of the Mae Kong Kha, have C-factors of 0.040 for 2003, as opposed to 0.020 for all forest in 1995. Also, the total forest area has decreased by about 6% while ?eld crop area, with a higher C-factor, increased by 10% (all fallow becoming ?eld crop). 5. Range of possible sediment loads Although the 2003 landuse classi?cation is more detailed than that from 1995, it still has a range of possible cover factors, especially for classes such as ?eld crops. As this study focused on examining the importance of estimation of cover on sediment delivery, the 1995 Mae Chaem landuse classi?cation was compared with the 2003 landuse for predicted maximum ranges of C-factors. Assumptions were made in assigning high and low C-factor values for the different scenarios, as some of the cover classes could be interpreted very broadly, (for example fallow). The 2003 sub-catchment landuse has however been able to reduce the range of C-factors by re-classifying forest into a number of discrete forest classes.

     M.G. Hartcher, D.A. Post / Mathematics and Computers in Simulation 78 (2008) 367?C378 Table 1 Summary of vegetation cover categories and C factors for Mae Suk C-factor 1995 Land use Forest Fallow Field Crop Paddy Urban Total 2003 Land use Hill Evergreen forest Mixed deciduous forest Dry deciduous forest Pine forest Hill evergreen/dry deciduous forest Total forest Fallow Field crop Fruit tree Paddy Urban Total 0.020 0.200 0.340 0.280 0.000 Area (%) 70.14 21.80 6.78 1.24 0.03 100.00

     375

     Sediment contribution (kt/year) 7.98 45.45 13.59 0.77 0.00 67.79

     0.003 0.040 0.020 0.088 0.020 0.100 0.340 0.150 0.280 0.000

     35.49 8.05 13.99 0.62 2.84 61.00 17.28 16.09 0.82 3.84 0.98 100.00

     3.95 2.24 1.68 0.41 0.35 8.62 10.21 32.25 1.18 4.31 0.00 56.58

     The possible range (low to high C-factors) of total predicted sediment yield for 1995 in Mae Suk was 154.45 kt/year (minimum 16.98 to maximum 171.43 kt/year). This range was reduced by 35% to 99.76 kt/year (minimum 31.86 to maximum 131.62 kt/year) under the 2003 landuse classi?cation (Table 3). This was due to the minimum predicted sediment yield increasing from 16.98 kt/year to 31.86 kt/year and the maximum predicted sediment yield decreasing

     Table 2 Summary of vegetation cover categories and C-factors for Mae Kong Kha C-factor 1995 Land use Forest Field crop Fallow Paddy Urban Total 2003 Land use Mixed deciduous forest Pine forest Hill evergreen Dry deciduous/Pine forest Dry deciduous forest Total forest Field crop Paddy Urban Fruit tree Mine Total 0.020 0.340 0.200 0.280 0.000 Area (%) 82.63 7.86 7.13 2.29 0.08 100.00 Sediment contribution (kt/year) 7.72 9.50 11.63 1.41 0.00 30.26

     0.040 0.088 0.003 0.040 0.020 0.340 0.280 0.000 0.150 0.800

     21.17 8.21 32.97 11.38 2.97 76.70 17.70 4.65 0.73 0.10 0.09 100.00

     4.56 3.62 2.19 1.81 0.33 12.51 19.28 3.69 0.20 0.07 0.19 35.94

     376

     M.G. Hartcher, D.A. Post / Mathematics and Computers in Simulation 78 (2008) 367?C378

     Table 3 Range of vegetation cover categories and possible C factors for Mae Suk Low C-factor 1995 Land use Forest Fallow Field Crop Paddy Urban Total 2003 Land use Hill evergreen Dry deciduous forest Mix deciduous forest Hill evergreen/dry deciduous Pine forest Total forest Fallow Field crop Paddy Urban Fruit tree Total 0.010 0.020 0.250 0.100 0.000 Sediment contribution (kt/year) 4.07 2.75 10.31 0.28 0 16.98 High C-factor 0.080 0.800 0.790 0.280 0.300 Sediment contribution (kt/year) 33.06 110.39 32.59 0.79 0.03 171.43

     0.001 0.001 0.001 0.001 0.088 0.020 0.250 0.100 0.000 0.150

     2.69 0.20 0.38 0.02 0.41 3.70 2.83 27.10 1.99 0.15 1.16 31.86

     0.003 0.020 0.040 0.020 0.088 0.340 0.790 0.280 0.300 0.600

     10.18 2.32 2.89 0.45 0.78 16.62 38.32 85.71 6.30 1.89 2.31 131.62

     from 171.43 kt/year to 131.62 kt/year. The high cover factor for fallow was reduced for 2003 with improved knowledge of fallow practices in the sub-catchment; however ?eld crops still have a high range of uncertainty associated with their sediment delivery contributing somewhere between 27.1 and 85.71 kt/year from low to high cover scenarios. Forest cover has decreased by around 10% which has been converted into ?eld crop, fallow, paddy and urban.

     Table 4 Range of vegetation cover categories and possible C factors for Mae Kong Kha Low C-factor 1995 Land use Forest Field Crop Fallow

    Paddy Urban Total 2003 Land use Mix deciduous forest Pine forest Hill evergreen Dry deciduous forest/Pine forest Dry deciduous Total forest Field crop Paddy Urban Fruit tree Mine Total 0.010 0.250 0.020 0.100 0.000 Sediment contribution (kt/year) 3.86 7.06 0.69 0.51 0.00 12.11 High C-factor 0.080 0.790 0.800 0.280 0.300 Sediment contribution (kt/year) 30.77 21.91 27.11 1.40 0.47 81.67

     0.001 0.088 0.001 0.001 0.001

     0.99 3.59 1.17 0.39 0.07 6.21 13.87 1.49 0.11 0.05 0.19 21.93

     0.040 0.088 0.003 0.088 0.020

     6.13 3.60 3.58 3.91 0.43 17.65 43.96 4.43 1.02 0.20 0.21 67.46

     0.250 0.100 0.000 0.150 0.800

     0.790 0.280 0.300 0.600 0.800

     M.G. Hartcher, D.A. Post / Mathematics and Computers in Simulation 78 (2008) 367?C378

     377

     Fig. 8. Impact of improved landuse classi?cation on the possible range of hillslope erosion in the Mae Suk and Mae Kong Kha sub-catchments.

     The possible range of total predicted sediment yield for the 1995 landuse in Mae Kong Kha was 69.56 kt/year (minimum 12.11 to maximum 81.67 kt/year). This range was reduced by 35% to 45.53 kt/year (minimum 21.93 to maximum 67.46 kt/year) using the 2003 landuse classi?cation (Table 4). This was due to the minimum predicted sediment yield increasing from 12.11 kt/year to 21.93 kt/year and the maximum predicted sediment yield decreasing from 81.6 kt/year to 67.46 kt/year. Forest cover has decreased by about 7% which has been replaced by ?eld crop, paddy, and urban. Fallow has disappeared completely, having been replaced by crops with permanent ?elds. Fig. 8 illustrates the range (from low to high cover factors) of predictions for hillslope erosion in both sub-catchments for 1995 and 2003 landuse data. In both sub-catchments, the 2003 forest landuse classi?cation has improved, having discrete categories of forest types. This has allowed for more appropriate cover factors to be applied and has therefore reduced the range of possible sediment yields from both subcatchments. However, the ?eld crop classi?cation has not been re?ned so we are still left with no knowledge of the actual crops being grown and therefore a wide range of possible cover factors, and hence possible sediment yields. In addition, the spatial aggregation of landuse types differed signi?cantly between 1995 and 2003. As discussed previously, the 1995 landuse was derived purely from Landsat classi?cation, which may have introduced various errors such as tree shadow and hill shading effects. This has created a ??speckled?? classi?cation causing the landuse types to appear to be disaggregated. 6. Conclusions The results of this study have indicated that the model shows signi?cant changes in the total

    amount and source areas of suspended sediment in the Mae Suk and Mae Kong Kha catchments in 2003 compared to 1995. However, it is not possible to say whether these changes are due to actual changes in landuse between 1995 and 2003 or whether it is simply due to the improved landuse classi?cation in 2003. In reality, both of these factors have probably impacted the results. However, the improved landuse classi?cation in 2003 has improved the determination of source areas of suspended sediment within the catchment by better representing the spatial location of different landuse types.

     378

     M.G. Hartcher, D.A. Post / Mathematics and Computers in Simulation 78 (2008) 367?C378

     In addition, the improved landuse classi?cation in 2003 reduced the range of possible sediment yields in both subcatchments by around 35%. This improvement was achieved primarily through the improved classi?cation of forest types. However, while forest has been reclassi?ed into different types with speci?c cover factors, most of the sediment is predicted to come from ?eld crops, which are still not represented as unique crop types in the 2003 coverage. Also, we have not focused on the effects of disaggregated landuse and how that may relate to topography and the spatial distribution of hillslope erosion. The 1995 classi?cation clearly shows landuse as being disaggregated throughout the landscape. This is most likely the result of an unsupervised image classi?cation, and may have affected the modelling results. As there is no way to analyse this, it produces uncertainty in terms of the location and quantities of suspended sediment. In addition, the other input factors used in the RUSLE equation have not been tested in this study. The development of more spatially accurate data for rainfall erosivity, soil erodibility, and digital elevation models would support a more thorough assessment of the range of possible sediment yields due to hillslope erosion. Concentrating efforts on improving the classi?cation of crop types will provide the greatest reductions in the possible range of C-factors. This will then allow us to focus on the key source areas of hillslope erosion and sediment yield. However, despite this, the improved landuse classi?cation in 2003 has clearly improved the determination of erosion source areas and has therefore provided managers with the ability to focus erosion control measures in these areas. Acknowledgments Thanks go to Pornwilai Saipothong (ICRAF, Chiang Mai University) and Barry Croke (Australian National University) for supplying data and advice on the study areas. This study was funded through ACIAR Project FST/1999/035. References

     [1] B.F.W. Croke, W.S. Merritt, A.J. Jakeman, A dynamic model for predicting hydrologic response to land cover changes in gauged and ungauged catchments, J. Hydrol. 291 (2004) 115?C131. [2] R. DeRose,

Report this document

For any questions or suggestions please email
cust-service@docsford.com