When using data sets derived from remote-sensing methods, it is important to understand the level of accuracy associated with a particular product. Accuracy figures can refer to the accuracy of the calculated value when compared to the actual biophysical value or it can refer to the positional accuracy. In some cases both figures are provided. Accuracy statistics should be distributed with the data set. Unfortunately, this is not always the case. In some cases, accuracy statistics for a data set simply do not exist.
For data sets that represent categorized data, the statistics usually provide per-class and overall accuracy information. This is common practice with land cover data sets. For data sets such as elevation, values are given for horizontal and vertical accuracy. These values are usually given as a probability of being within a specified distance. Other data sets, such as those derived from MODIS data are validated by a team of scientists and in some cases the validation effort is incomplete or ongoing. When using these data sets, it is important to research the most current information available about the data set's accuracy. This information is often available on the Internet.
Different methods for reporting accuracy exist and this is an active research area. For example, new accuracy methods are being developed that provide information about the spatial distribution of the error. Methods are also being developed using fuzzy statistics to indicate the severity of the error instead of using the traditional approach ofnoting a value as either correct or incorrect.
Was this article helpful?