Relative calibration of multitemporal Landsat data for forest cover change detection
Not for data collection
Publication information
OKM publication type
A3
Category
Artikkelit ja abstraktit
Sub category
Artikkelit tieteellisissä kokoomateoksissa
Type
Alkuperäisartikkeli
Refereed
Kyllä
Publication channel information
Title of journal/series
Remote Sensing of Environment
Internationality
Yes
Detailed publication information
Publication year
1999
Page numbers
1-11
Language of publication
Finnish
Co-publication information
International co-publication
No
Classification and additional information
Subject headings
, deforestation, land use change, change detection, relative calibration, reforestation,
Additional information
Abstract The aim is to investigate how well old Landsat data can be calibrated and utilized for monitoring past changes in forest cover status. The material consists of three MSS images and a TM image covering a period of 19 years. The classes of interest were 1) no change, 2) deforestation and 3) reforestation. Various relative calibration methods are first compared and their effects on the results of interpreting the "change image" are then studied. As it is sometimes impossible to locate unchanged areas for calibration, the use of both unchanged and changed training areas for calibration purposes is tested here. The change detection method is fixed to parametric supervised change image classification. The longest time difference between image pairs was 19 years and R2 with a time span of this order was around 40%. R2 for the models increased about 10% when time span was reduced from 11 years to 3 years. Smoothing methods can predict band-by-band regression very well, and R2 values are 10 - 15% higher than with the robust regression technique used. RMSE was slightly larger when multiple band models were used than with smoothing methods. When general calibration models were applied to unchanged data, the differences between the calibration errors were not statistically significant. Thus it is not very dangerous to use an entire target area for image calibration and include changed areas in the training data for calibration models. Models created from multiple occasions did not perform well in classification. A fair agreement with estimates and the ground truth was possible using four images. The best result was obtained when using three image pairs and multiple band robust regression calibration. Classification with calibration was only slightly better than without calibration, the difference not being statistically significant.,