Remote Sensing and Wetland Ecology: a South African Case Study

Abstract

Remote sensing offers a cost efficient means for identifying and monitoring wetlands over a large area and at different moments in time. In this study, we aim at providing ecologically relevant information on characteristics of temporary and permanent isolated open water wetlands, obtained by standard techniques and relatively cheap imagery. The number, surface area, nearest distance, and dynamics of isolated temporary and permanent wetlands were determined for the Western Cape, South Africa. Open water bodies (wetlands) were mapped from seven Landsat images (acquired during 1987 - 2002) using supervised maximum likelihood classification. The number of wetlands fluctuated over time. Most wetlands were detected in the winter of 2000 and 2002, probably related to road constructions. Imagery acquired in summer contained fewer wetlands than in winter. Most wetlands identified from Landsat images were smaller than one hectare. The average distance to the nearest wetland was larger in summer. In comparison to temporary wetlands, fewer, but larger permanent wetlands were detected. In addition, classification of non-vegetated wetlands on an Envisat ASAR radar image (acquired in June 2005) was evaluated. The number of detected small wetlands was lower for radar imagery than optical imagery (acquired in June 2002), probably because of deterioration of the spatial information content due the extensive pre-processing requirements of the radar image. Both optical and radar classifications allow to assess wetland characteristics that potentially influence plant and animal metacommunity structure. Envisat imagery, however, was less suitable than Landsat imagery for the extraction of detailed ecological information, as only large wetlands can be detected. This study has indicated that ecologically relevant data can be generated for the larger wetlands through relatively cheap imagery and standard techniques, despite the relatively low resolution of Landsat and Envisat imagery. For the characterisation of very small wetlands, high spatial resolution optical or radar images are needed. This study exemplifies the benefits of integrating remote sensing and ecology and hence stimulates interdisciplinary research of isolated wetlands.

Keywords: Envisat; Landsat; Wetland monitoring; wetland distribution and density; wetland ecology.

Figure 1.

Number of delineated wetlands per km². Sun symbols indicate images taken during summer. Diamonds indicate the cumulative amount of rainfall over twelve months before the date of image acquisition (mm/year).

Figure 2.

Fraction of detected wetlands belonging to different size classes (in ha) in summer and in winter, with indication of the standard deviation.

Figure 3.

Visualisation of the distance of each pixel to the nearest wetland in the winter of 2000 (left) and in the summer of 2001 (right); light shading = long distance, dark shading = short distance to the nearest wetland.

Figure 4.

Details from the cross-classification of the maximum likelihood Landsat ETM+ classification of June 2002 and the Envisat ASAR classification of June 2005. The left part shows a region in the East of the study area containing few small to medium-sized wetlands; the center inset illustrates classification results over a vast area with medium-sized to large wetlands, located in the center of the study area; the right part covers a very large salt pan near the Atlantic coast.

Figure 5.

Number of detected wetlands in the study area used for comparison (see materials and methods), as a function of wetland size class (ha) for the Envisat ASAR image of June 2005 and the Landsat ETM+ image of June 2002.