Institute of Remote Sensing,
Anna University, Chennai, India
Email: [email protected]
Shamla Rasheed, K Venugopal
Muhammad I Tubbsum, Tayyab I Shah
The extraction of the information from the remote sensing images for monitoring and managing the resources has always been the purpose and challenge. The updating of land and water resources is the important task in any land resource management. Specially in such dynamic phenomenon like agriculture, the changes have to be ascertained for the management and policy decisions. The use of multi temporal data for the land use change detection in agricultural areas is well known. The use of land and water resources for human kind for various purposes and the increasing needs by the humankind is quite understandable. Especially the importance of effective use of the limited water resource needs no emphasis. The water needs for the agricultural purposes in the highest of all other consumptive uses of this precious resource. Hence it is imperative to account for the irrigational needs and uses of water for making more effective and efficient water use decisions. Fro this purpose one needs to know accurate water needs for irrigation before anything. A dependable and accurate information extent of crops and their requirements in to be known for assessing this water needs. Remote sensing comes handy as dependable source. And what about accurate crop map?
The standard methodology is going in for classification of remote sensing data and comparing the areal extent under different crops. More often the standard procedure of training the classifier using the ground samples and classifying the image at one go will result in more error because of the heterogeneity of various factors. The crop growth stage, cropping pattern, crop type, etc. play a major role in the performance of the classifier and therefore on the accuracy of the crop map. It is always better to study the agricultural pattern on the study area and invoke these special conditions in the classifier. A rule based procedure that takes into account the specific facts regarding the area of interest also works in this way. But the difference in this attempt is that the standard classification procedure is embedded with the facts regarding the local conditions.
This methodology of embedding the standard maximum likelihood classifier with the local knowledge in respect of the factors influencing the classification accuracy is demonstrated through this study. Here a small irrigation command area viz. Thirumangalam main canal command area is taken for this study. A frame work of classification model based on the ground knowledge is formulated and the results of maximum likelihood classification are improved by application of this knowledge. The knowledge framework purely driven by the unique and age-old local agricultural practices is combined with standard classifier and the crop maps for three seasons pertaining to the years 1989-90,1994-95 and 2001-2002 are derived and their accuracy is assessed. This particular paper covers the methodology of combining the local knowledge with standard classification procedure for deriving agricultural land use information from remote sensing images, though the entire study envisages at assessing the agricultural performance of the command area.