A Coefficient Of Agreement As A Measure Of Thematic Classification Accuracy

ABSTRACT: Changes in land use and land use in a fast-growing urbanized area have been studied by many researchers and much work is underway on this subject. Often, the region`s urban sprawl depends on many factors such as economic prosperity and population growth. Iraq is one of the countries where the settlement area has grown rapidly. Remote sensing and the Geographic Information System (GIS) are analytical software technologies to assess this well-known global phenomenon. This study illustrates the evolution of settlements in Sulaimaniyah governorate from 2001 to 2017 with Landsat satellite images from different eras. All images were classified with remote sensing software to perform a high-performance mapping of land use classification. The maximum probability method is used in solution information extracted accurately from spatial images. The landsat images of the study area were divided into four different classes. Here is: the forest, the vegetation, the soil and the subdivision. The results of the change recognition analysis show that, in the face of an explosive demographic change in the settlement area, where the record of 8.99 per cent, or 51.80 km2 over a 16-year period, and the housing area rises from 3.87 per cent in 2001 to 12.86 per cent in 2017. The accuracy assessment model was used for the analysis of classified images (LULC). The results of the accuracy show an overall accuracy of 78.83% to 90.09% between 2001 and 2017, with convincing results of the Kappa coefficient between essential and almost perfect chords.

This study will help urban development decision-makers. An error matrix is often used to organize and display the information used to assess the thematic accuracy of a soil coverage map, and many precision measurements have been proposed to summarize the information contained in this error matrix. No single measure is generally the best for all accuracy assessment objectives, and different precision measurements can lead to conflicting conclusions, since the measurements are not accurate in the same box. Selecting appropriate precision measurements is essential to achieving the objectives of the mapping project. The characteristics of some frequently used accuracy measures are described and the relationships between these measures are made available to help the user choose an appropriate measure. Precision measures that can be interpreted directly as probabilities for certain types of classification errors or correct classifications should be selected in. The accuracy of the user and manufacturer, as well as the total share of the correctly classified surface, are examples of precision measurements that present the desired probabilistic interpretation. The Kappa coefficient of the agreement does not have such a probabilistic interpretation because of the adaptation to the hypothetical random agreement that was included in this measure, and the strong reliance on kappa relative to the marginal proportions of the error matrix makes Kappa useful for suspicious comparisons. The standardization of an error matrix leads to estimates that are not consistent for those responsible for the precision economy of the map to evaluate, so this method is generally not warranted for most applications. This research was supported by a CR821782 cooperation agreement between the Environmental Protection Agency and SUNY-FSE. This manuscript has not been peer-reviewed and policy-reviewed by CEPOL and does not necessarily reflect the Views of the Agency.

David Verbyla and two critics made some helpful suggestions to improve the manuscript.


Please signup or login to answer this question.