Estimation of land coverage from a land cover classification derived from remotely sensed data |
| |
| Authors: | Giles M Foody |
| |
| Institution: | (1) Department of Geography, University College of Swansea, Singleton Park, SA2 8PP Swansea, United Kingdom |
| |
| Abstract: | Remotely sensed data are an attractive source of land cover information. In many applications the required information relates to the extent or coverage of land cover class(es) in a region, which is generally derived from a count of the pixels allocated to the class(es) of interest in a classification. A highly accurate classification is not required for the derivation of accurate estimates of class coverage, provided the classification is accompanied by appropriate information on its quality. For instance, the information on classification quality contained in the classification confusion matrix can be used to significantly increase the accuracy of the estimates of land coverage. This is illustrated with reference to a case study focused on the estimation of despoiled land coverage in administratively defined local district in industrial South Wales from Landsat TM data. The accuracy of the investigation was assessed relative to a map of despoiled land cover for this region produced by conventional methods. From an image classification of moderate accuracy, the classification accuracy ranged form 57–83% between the districts investigated, a pixel count provided estimated of despoiled land coverage that were only poorly correlated to the mapped coverage;r = 0.27. Using the information on the pattern of error in the class allocation contained in the classification confusion matrix the estimation accuracy was increased significantly, with a correlation ofr = 0.81 observed between the remote sensing based estimate and the mapped land coverage. Furthermore, the r.m.s. error in despoiled land coverage estimation was reduced by approximately half, to less than 1% district area, when the classification was used in conjunction with information on the pattern of classification error. |
| |
| Keywords: | |
| 本文献已被 SpringerLink 等数据库收录！ |
|
