Evaluation of the suitability of Landsat, MERIS, and MODIS for identifying spatial distribution patterns of total suspended matter from a self-organizing map (SOM) perspective

Abstract

Three different empirical models based on Landsat, MERIS, and MODIS sensors were used to retrieve TSM concentrations and construct long-term TSM image datasets during the period from 2003 to 2011. These datasets were utilized to evaluate the suitability of Landsat, MERIS, and MODIS for identifying spatial distribution patterns of TSM concentration based on a self-organizing map (SOM) algorithm in the Pearl River Estuary (PRE). 12 spatial distribution patterns of TSM in the PRE were identified from long-term TSM data of each sensor based on a 4 × 3 SOM array. The expected frequency of occurrence of any given pattern and relative frequency of occurrence of each pattern were calculated and analyzed. The factors that influence the variability of TSM and the spatial distribution patterns of TSM were explained by the contribution of human activities, precipitation, and characteristics of satellite sensors. Our results showed that similar spatial distribution patterns of TSM were extracted from three different TSM image datasets. Each pattern presented a strip distribution from northeast to southwest, and the values of TSM decreased regularly from northwest to southeast. However, patterns identified from Landsat and MODIS TSM images presented some anomalous patterns, which were inconsistent with local circumstances and previous studies. High (low) level spatial distribution patterns of TSM in the SOM array were associated with a variety of high (low) values of TSM concentration in the PRE. Frequency analyses of each pattern demonstrated that notable differences of TSM concentration existed in the PRE from 2003 to 2011. These findings can assist us to better understand the dynamics of TSM concentration and the performance of different empirical models established by Landsat, MERIS, and MODIS for identifying spatial distribution patterns of TSM concentration using the SOM algorithm.