Corp Discriminational in Salt Affected Soils
by Satellite Remote Sensing
R.K. Sharma, K S Sundara Sarma and D K Das
Division of Agricultural Physics,
Indian Agricultural Research Intitute,
New Delhi-110 012, India
Abstract
Crop discrimination is a basic requirement for acreage estimation and yield prediction by satellite remote sensing. In salt differences in reflectance caused by salt encrustation as well as retarded crop growth due to salinity. Studies were carried out in sultanpur (Gurgaon, Haryana, India) to find our the utility of IRS I LISS II satellite data in delineation of salt affected soils, differentiation of crops grown in these soils and the effect of various soul and management factors on crop growth.
Delineation of the saline and sodic soils could be achieved by Maximum likelihood classification (MXL) of IRS LISS II data obtained for saline souls were differentiated from normal soils are devoid of vegetation. The sodic and saline soils were, differentiated from normal soils with an overall accuracy of 78.2%. Under cropped conditions, during January, the MXL classifier could differentiate wheat, mustard, tree plantation and waste lands with classification accuracies of 74, 68.7, 87.1 and 97.0, respectively. Inclusion of additional classes viz., early and late sown wheat irrigate and unirrigated mustard improved the classification accuracies of wheat and mustard to 77.2 and 81.2%. The attributed to the poor growth of wheat and mustard due to salinity, irrigation and varying dates of sowing. These factors could explain 81% of the variability in Normalized vegetation index values (NDVI) under wheat cropped conditions. With regard to mustard salinity is the major contributor to the differences in NDVI.
Introduction
A major application of satellite remote sensing in agriculture is the identification of crops for acreage estimation and yield prediction. While discrimination and area estimation of crops growing in normal soils have been operationalized (Navalgund 1991), in salt affected souls the differences in reflectance due to salt encrustation as well as the retarded crop growth due o salinity make crop discrimination a difficult task.
The differences in the geometry and growth habits of plant types are manifested in their spectral responses which are used to discriminate various vegetation types. Single date satellite data (Dadhwal and Parihar 1985) as well as spectral temporal vegetation profiles (Badhwar et al. 1982) are used to discriminate various crops. However, the latter method is cost prohibitive and availability of cloud free satellite data on consecutive satellite passes is a limiting factor. Soil background (Major et al. 1990) is one of the most important factor interfering multispectral classification of crops through remote sensing, this is especially important in salt affected soils where crop growth is adversely affected due to salinity.
Landsat data have been used (Singha and Dwivedi 1989 and Saha et al. 1990) for delineation of salt affected soils. Recently, Gore and Bhagat (1991) delineated salt affected soils using IRS data. The present investigations were carried out to study the utility of IRS digital data for delineation of salt affected soils, differentiation of crops grown under salt affected soil conditions and effect of soil and management factors on spectral vegetation index of crops.
Materials and Methods
The study area belong to Sultanpur village of Grugaon District, Haryana State, India for which IRS-IA LISS-II scenes for the months of June 1990 and January 1991 were procured. A base map of 1:7250 scale was prepared from the cadastral map for ground truth data representation. Intensive soil sampling (120m x 125m grid) was done for the characterization of area for salinity/sodicity. By making use of spatial statistics the point data for electrical conductivity and pH were krigged for the preparation of thematic maps to be used in the GIS. Various crops, tree plantation, wastelands and area used for other purposes were carefully recorded in the base map. Information with regard to soil, water and crop management aspects like soil water transmission parameter, a, crop variety, date of sowing kind and time of fertilizer application number of irrigations applied, date of harvesting and yields etc. was also obtained. This information was digitized and converted to IDRISTI is a raster based GIS software with image analysis capability.
The IRS LISS II satellite data were analyzed on IBM PC/AT using IDRSI software. After registering satellite images to the ground truth maps, training sites were selected. These sites were used to generate signatures of various soil types and land use/land cover classes which were used in the supervised maximum likelihood classification algorithm for classifying the study area. For soil type differentiation the June 90 imageries were used, when the vegetation was sparse and for crop differentiation studies, January 91 period images, when the crop was at its peak vegetative growth, were used.
