Resource information
and is becoming a scarce natural resource due to the burgeoning population growth and urbanization. Essentially, detecting changes in land surface is significant for understanding and assessing human impacts on the environment. Nowadays, land use change detection using remote sensing data provides quantitative and timely information for management and evaluation of natural resources. This study investigates the land use changes in Birjand of Iran using Landsat TM5 images between 1986 and 2010. Artificial neural network was used for classification of Landsat images. Five land use classes were delineated include Pasture, Irrigated farming Land, Dry farming lands, Barren land and Urban. Post-classification technique applied to monitor land use change through cross-tabulation. Visual interpretation, expert knowledge of the study area and ground truth in formation accumulated with field works to assess the accuracy of the classification results. Overall accuracy of 2010 and 1986 image classification was 89.67 (Kappa coefficient: 0.8539) and 88.78 (Kappa coefficient: 0.8424) respectively. The results showed considerable land use changes for the given study area. The greatest increase was related to Barren land class almost 378 percent. The dry farming lands reduced by almost 48 percent during the study period. Urban class has increased drastically about 219 percent, 3percent of dry farming lands, 61 percent of pastures lands, 4percent of irrigated farming land in 1986, converted to urban and industrial land in 2010 and alone 31 percent of urban land in 1986 had conformity to urban in 2010. Irrigated farming land increased about 17.16 percent predominantly due to population growth. The result of this study revealed a successful application of the ANN approach for land use change detection. Although this model demonstrated high sensitivity to training samples data, it required trial and error for attainment more accurate. But high accuracy of classification in last two years proved that ANN was highly efficient for classification of Landsat images in the study area.
