Evaluation of IRS LISS III data for apple orchard inventory - A case study covering Jubbal

Evaluation of IRS LISS III data for apple orchard inventory - A case study covering Jubbal - Kotkhai block of Shimla district

N. S. Mehta¹, Nitin Bhatt¹, R. S. Thapa², Arvind Sharma³, R. K. Sood² and S. Panigrahy¹
¹Space Applications Centre, Ahmedabad - 380 015
²H P Remote Sensing Cell, Shimla – 171 009
³State Department of Horticulture, Shimla –171 002

The remote sensing technology has potential in estimating crop acreage and production at district, regional and national level due to its multi spectral, synoptic and repetitive coverage. This technology is being used operationally by many advanced nations for monitoring natural resources. Application of space - borne remote sensing is of particular significance to India, because it is the second largest producer of fruits (469 lakhs tons per year) and vegetables (580 lakhs tons per year) in the world. Shimla district has occupied place of pride in the field of horticulture not only in the state but also in the country, it is the biggest apple growing district in Himanchal Pradesh. Apples are grown on 83 percent of horticultural land in the district. This study discusses results obtained on apple orchard inventory covering Jubbal - Kotkhai block of Shimla district, using IRS LISS III data. IRS LISS III data of May 27, 1999 were used to classify apple orchards of the block. Unsupervised classification was carried out using ISO - data clustering after masking the higher / lower reflectance classes such as cloud and shadow. This classification grouped apple orchards into four groups. The out put of ISO - data clustering was used to collect ground truth (GT) for apple orchards and other land cover classes. After ground verification, apple orchards were classified into three main classes on the basis of their density i.e. dense, moderatively dense and sparse orchards. Other land cover classes identified are forests, agricultural fallow, barren hills etc. The GT collected was also used to train the classifier.

MXL classification was carried out using three (green, red and IR) and four (green, red, IR and MIR) band data, using complete enumeration method. Analysis carried out using three band data shows that apple orchards are occupying 33.18 percent of geographical area of the block against 26.08 percent provided by the State Department of Horticulture, where as using four band data area under orchards comes out to be 24.14 percent. MXL classification using four band data shows more area under null class as compared to three band data. There is also some overlap in signature between dense apple orchards and broad leave mixed forest. Overall classification accuracy using three band data was found to be 93.53 percent and including MIR band the accuracy comes out to be 91.82 percent.