Remote sensing for assessing wetland-groundwater interaction in the Kogelberg Biosphere Reserve

[摘要] ENGLISH ABSTRACT:The Table Mountain Group (TMG) Aquifer System is a regional fractured aquifer systemwith a large potential as a source of future water supplies in the Western and Eastern Cape.This system is currently under consideration for large-scale water abstraction. Many terrestrialecosystems, however, are dependent on these groundwater resources for survival.Exploitation of ground water resources at a rate exceeding the rate of natural recharge wouldresult in a lowering of the water table and the drying up of seeps.The main objective of this study was to determine if satellite remote sensing data can be usedfor the detection of groundwater-dependent wetlands, and secondly, to use multi-temporalimagery for estimating seasonal changes experienced in wetland communities in relation tosurrounding vegetation. The Kogelberg Biosphere Reserve, situated approximately 30km tothe east of Cape Point in the Western Cape, South Africa, was selected for investigation. Toaccomplish the objectives, three Landsat 7 ETM+ images (path/row: 175/84) captured on 22September 2001, 18 May 2002 and 23 September 2002 were acquired. Image fusion of themultispectral bands (30m resolution) with the panchromatic band (15m resolution) provided15m multispectral images for analysis purposes. Geometric correction, radiometricnormalisation and atmospheric corrections was performed in order to ensure pixel-levelcomparability between images. Once comparability between images was guaranteed,vegetation indices and tasselled cap components were derived to provide threshold values ofmoisture stress indicators and productivity estimations of wetland communities in relation tosurrounding non-wetland communities. Additionally, change vector analysis on thesetransformations provided the ability to detect and assess the seasonal changes experienced bythese communities during an annual cycle. The results of these transformations werecombined in a rule-based image classifier in order to assist in estimating the seasonaldependency of observed wetland communities.The ability to use Landsat 7 images and the abovementioned image processing procedures toidentify wetland communities with a high probability of groundwater interaction wasdemonstrated with a high degree of accuracy (78%). It is recommended that future studiesconcentrate on increasing classification accuracies, while focusing on incorporating thesetechniques into a remote monitoring system for assessing the impacts of groundwaterextraction on the groundwater-dependent wetland communities.

[发布机构] Stellenbosch University