Modelling water quality using thematic mapper data: Case of lake michigan

Publisher Website
Google Scholar
Abstract
Analysis of water qualify based upon prediction models that use discretely monitored data is affected by the locations of sampling stations and may not represent dominant water conditions. These models are also affected by the time lapse, sampling error, and atmospheric influences on the spectral data. This research investigated the impact of these three factors upon the predictive capability of water quality models using archived TM and archived ground based water quality data. Coordinate transformation was implemented to locate and identify ground sampling stations on the TM data products. Chavez's (1975) regression method was used to reduce haze effects. The impact of time lapse, sampling error and haze effects on the reliability of water quality predictive modelling is significant and, thus, it is unlikely that reliable water quality predictive models can be generated using archived TM and ground‐truth data. Our attempts to reduce atmospheric effects using Chavez's method were unsuccessful. Limitations intrinsic within ground‐truth data and remotely sensed data (especially time lapse) have profound effect on quantitative water quality assessment.