Remote Sensing and GIS Technologies for Denudation Estimation in a
Siwalik Watershed of Nepal
Kioshi Honda**, Lal Samarakon*, akichika Ishibashi*, yasushi Mabuchi*, Shigechika Miyajima***
*Nippon Koei Co., Ltd., Research & Development Center,
2304 Takasaki, Kukizaki, Japan
**STAR/SERD, Asian Institute of Technology, Bangkok, Thailand
***Water induced Disaster Prevention Technical Centere, Katmandu, Nepal
Abstract
Wide spread and intensive soil erosin has resulted gradual land degradationin most parts of the mountains ofNepal threatening the prosperity of Nepalese society and the natural environment. The rate of denudation is highly varying with natural factors together with the ever increase demand for forest resources and utilsable land for the growing population. Comprehensive studies extending from watershed basis to regional level have to be carried out for introducing appropriate counter measures in reducing devastating disasters.
Present study is an attempt to use remote sensing to identify the land degradation in Ratu watershed in the Central Siwalik area, and establish a method to estimate the rate of denudation in GIS perspective, Landsat data procured for a period of 20 years from 1973 to 93 were analyzed for the change of forest cover in the waterhed and topographical parameters were used in a model to estimate the probable annual soil loss. Model was further improved for flood event soil yield estimation using factors that have identified in the field as main causes for intensive erosion during a heavy rainy season.
1. Introduction
Nepal is a mountainous country with an area of 147.181 square kilometers. Two third of the total area of the country is occupied by hills and mountains. Population stands at 18.8 million (1992), with a growth rate of 2.1%. It is estimated that 75% of the total energy comes from forest, (WECS, 1992). This utilization practices lead to high degree of degradation of the land and loss of natural forest resources. Exact reasons and the actual rate of land degradation is not properly investigate for the entire country, but most of the literature published blaming the farmer for destruction of forest for energy subsidy and mountain farming. The detail examination of sedimentation in watershed basis may be a better soluition to understand the cause and effect of the flood and soil erosion process, but the limited resources may hinder this efforts.
Lack of knowledge of natural event, their occurrence, recurrence time could hinder proper planning or disaster prevention, and mitigation. It was though satellite remote sensing data could offset the information gap as they can provide timely, and repetitive data of aparticular phenomena without physically visiting the area of interest. In this study, it was decide to investigate the applicability of satellite data in land degradation assessment, and monitoring the phenomenon. Also, a soil yield estimation method was proposed, which uses generally available information based on a GIS environement.
2. Outline of the Study Area
Ratu watershed, which is situated in the Siwalik area was selected for the present study. Ratu river originates from a Churiya hills of central Nepal at an altitude of 700 meters and drains into Terai region. The watershed lies west of Kamala river and south of sun Koshi. Ratu river crosses the East-West highway aroung 86E longitude and 27N latitude. River runs dry during the dry season and transport heavy load of sediment during the rainy season with the high water flow.
The Siwalik area is composed of young tertiary strata, and contains some of the most easily erodible lithologies including unconsolidated sand and gravel. Soft and loose sandstone, mudstones, and conglomerate are the predominant rock types in the present watershed. About 70% of the watershed area is dominated by forest with varying cover densities. Shorea Robusta, locally called Sal is the major species that is found in the area. Remaining land could be grouped into crop lands and barren land according to their land use practice.
3. Data Acquisition
3.1 Acquisition of Remote Sensing and field Verification Data
It was decided to obtain satellite remote sensing data over a long time span to estimate the changes,
Specifically to observe the forest cover and its change that could be used as a key to land degradation. Also field investigation was planed to identify the physical nature of the soil erosion. It transport and sedimentation to evaluate the volume of sediment production that could use to compare satellite data evaluated soil production. A major flood event was considered to estimate the volume of sedimentation by field observations. The most recent flood event, 1993 July was selected for this as it could give the opportunity to collect the information on the extent and nature of the flood by visiting the field where no other means of information is available. Referring to available satellite data sources, data shown in the Table 1 was selected for the study. Further, during the selection of data, special attention was given to choose data that are acquired in the same season of the year to minimize irradiance variations on the surface due to positional difference of the sun. The LISS sensor data was acquired irradiance variations on the surface due to positional differences of the sun. The LISS sensor data was acquired for investigation of its capability in replacing the largely used Landsat sensor for land information extraction.
Table 1 Information on the satellite data obtained for the study
| Sensor |
Date |
Sun Angle |
Sun Azimuth |
Processing Level |
| MSS |
1973.03.14 |
Not Known |
Not Known |
System Corrected |
| MSS |
1977.03.20 |
Not Known |
Not Known |
System Corrected |
| TM |
1993.03.16 |
47° |
127° |
System Corrected |
| TM |
1995.03.22 |
46° |
127° |
System Corrected |
| LISS-II |
1995.03.15 |
47° |
127° |
System Corrected |
Referring to the availability of satellite data, and considering major flood events in the past, aerial photographs acquired in 1979 and 1992 were obtain for the study. The scale of the photographs are 1:40,000, and decide to photograph the Ratu watershed using a helicopter. A normal camera that uses visible light with 50mm focal length was used to obtain vertical photographs. Further, some of the sub-streams were surveyed for recognize the sediment formation, distribution and cludes to differentiate the age of the deposits.
3.2 Geometric Correction of Satellite Data
The spatial resolution of the acquired satellite data are 30m, 57x79m and 34m for TM, MSS and LISS, respectively. In addition, their spatial orientation also dissimilar. Therefore, the procured satellite data were brought into a common map projection (UTM) by constructing mapping function through identifying control points on 1:25,000 topographical maps. First order transformation functions were established for all the datasets, and nearest neighbor resampling was used for re-mapping the satellite data into UTM projection.
3.3 Radiometric Rectification of Satellite Data
Digital comparison of multi-sensor data requires further adjustment as the observations are made in sensor specified discrete spectral bands. The amount of energy received at the sensor from a particular earth feature is a function of received energy, reflectance, atmospheric propagation, sensor sensitivity, and the spectral bandwidth. It we consider a homogeneous land surface and similar atmospheric condition, the amount of received bandwidth. It we consider a homogeneous land surface and similar atmospheric condition, the amount of received energy is a function of bandwidth of the spectral channel. In the present study, the satellite data of three sensors adjusted for the differences in the sensor bandwidths considering linear spectral reflectance within each respective Landsat TM spectral bands. Further, histogram equalization was carried out to compensate for different atmospheric conditions that would have prevailed at the time of satellite passes having 1993 landsat TM as reference dataset.
3.4 Database Creation
The conventional maps, topographical and vegetation were digitized in establishing a GIS database in the UTM map projection for land degradation and soil erosion analysis. Further, the aerial photographs and helicopter photographs were scanned, rectified and registered into UTM projection. Thus the completed GIS database for the present study included multi-sensor temporal satellite data, aerial photographs, elevation and land cover information.
4 Method of Analysis
4.1 Land Degradation and satellite Data
As the satellite data are obtained for the dry season, classification of forest and other lands was straightforward as the crop lands are uncultivated. This can easily be accomplished by satellite data derived vegetation indexes. Further, the changes in the forest cover in its density and any deforestation can also be
Estimated by vegetation indices. Most of the vegetation indices are based on empirical evidences, and nearly all of commonly used vegetation indices are concerned only with red and near-infrared regions of the electromagnetic spectrum. NDVI based on the equation 1 was used with multi-sensor digital data that were adjusted for bandwidth differences and compensated for atmospheric variations at the time of satellite passes.
1992 aerial photographs were taken as ground truth information to classify 1992 Landsat TM data. Also, these photographic information were further compared with other dates of satellite data as the observation conditions were re-established to 1992 satellite pass. Aerial photographs covering the study area were classified, and digitally compared with NDVI images of fours dates for evaluation of forest land changes in the watershed.
4.2 Soil Erosion Model
Soil erosion in this region could be due to tectonic activities and/or high intensity continuous rainfall, (lves, 1989, Summerfield, 1991). The tectonic activities are yet to define precisely when quantifying erodibility, and these activities are rather a continental process and may not be evaluated in micro level watershed analysis. On the other hand, water erosion is more regional and can be considered in micro scale, but lack of field measurements on precipitation and runoff could hinder the application of most of the empirical soil erosion models. Current soil erosion models, Universal Soil Los Equation and, most of the ther models that are commonly used are developed by various authors on extensive field observations. Further, most of these equation requisite long term rainfall records, and measured denudation rates under different land use conditions.
Estimation of soil erosion in the Ratu watershed, utilizing a model that incorporate rainfall data is questionable due to lack of field observations. The other major factor that decide the rate of soil erosion in this watershed is the surface topography, hence a model that relate the surface topography was considered for erosion potential estimation. Honda, 1993 proposed a method for soil erosion in mountainous region by incorporating surface gradient. He has demonstrated the accuracy applying in a well monitored watershed in Japan. The annul average denudation E can be estimated as;
Where, E30 - rate of denudation at a slope of 30o; S-gradient of the pixel under consideration; S30 - tan (30o)
The denudation factor has to be identified for every spatial location of the watershed or, has to be measured for well distributed points to apply the model as it is. This is not realistic even in a well monitored suggested. As no field measurements are carried out, denudation of dominant land cover types are extracted from documented materials, and a relationship with the respective NDVI values are established. Given the NDVI value for any location, the rate of denudation is estimated based on the relationship. As the relationship between denudating and NDVI plays a critical role in this method, attempt was made to identify the potential of NDVI in representing the state of the forest cover in the watershed.
4.3 Erosion Characteristics of Storm Event
I was observed in the field that the landslides along the banks of streams, undercutting, and earth topples are some of the factors that could highly contribute to sediment transport during a storm event. This phenomenon was specially visible in the sub-streams, (JICA, 19960. Therefore, it was decided to include the distribution of sub-streams or a factor that represents the valleys in the flood event soil erosion model.
