Accuracy assessment and validation of remotely sensed and other spatial information

TLDR

Since the 1970s, the importance of accuracy assessment for remotely sensed maps has evolved from limited scrutiny to mandatory minimum standards, while other spatial data lag behind in systematic validation. The paper aims to present a series of steps toward improved assessment and validation of spatial information and invites readers to evaluate their current position in this series. It outlines a stepwise framework for assessing and validating spatial data.

Abstract

This paper was presented at the conference ‘Integrating spatial technologies and ecological principles for a new age in fire management’, Boise, Idaho, USA, June 1999 Today, validation or accuracy assessment is an integral component of most mapping projects incorporating remotely sensed data. Other spatial information may not be so stringently evaluated, but at least requires meta-data that documents how the information was generated. This emphasis on data quality was not always the case. In the 1970s only a few brave scientists and researchers dared ask the question, ‘How good is this map derived from Landsat MSS imagery?’ In the 1980s, the use of the error matrix became a common tool for representing the accuracy of individual map categories. By the 1990s, most maps derived from remotely sensed imagery were required to meet some minimum accuracy standard. A similar progression can be outlined for other spatial information. However, this progression is about 5 years behind the validation of remotely sensed data. This paper presents a series of steps moving towards better assessment and validation of spatial information and asks the reader to evaluate where they are in this series and to move forward.