Evaluation of Landsat-Based Area Measurement Accuracies for Surface Water Area in New Mexico

An effort to refine the techniques and methodology used in a previous WRRI project and to establish a better understanding of the data accuracy requirements of water resource management agencies are described. The objectives of this research were: 1) to identify the data accuracy requirements of resource management agencies who create or utilize reservoir volume statistics, and 2) to compare those requirements against measurements of surface water area taken from Landsat reflective infrared (band 7) imagery and digital data tapes by digital and areal measuring methods. Project results show that the maximum amount of tolerance acceptable would be +20%. However, it was suggested that in all cases a +10% accuracy range would be more useful.

The results for objective number two indicate that consistent and reliable measurements of reservoir surface areas from Landsat satellite data are possible. it appears that a computerized digital measuring technique such as the Image 100 can be used now to make reservoir surface area estimations. It is also possible that measurements from the planimeter and image scales of 1:250,000 can be improved to meet the desired accuracy levels. Further work needs to be done to improve on the measuring techniques and procedures to provide data with a known variance and reliability. There is also room for newer and more advanced classification and measuring techniques to be evaluated. One such technique would be automated image classification using unsupervised statistical classifiers. Other image processing systems which are more economic to run and easier to operate than the Image 100 should also be investigated.

Project Nos. 3109216, 1345631