Sources of Uncertainty in Mapping Temperate Mangroves and Their Minimization using Innovative Methods

Estimates of temperate mangrove forest cover are required for management of estuarine ecosystems, particularly in areas experiencing rapid change in mangrove distribution. However, it remains challenging to obtain accurate estimates of temperate mangrove cover using remote sensing because of the unique physical features and environmental conditions of temperate mangroves. The objectives of this study were (1) to develop an improved image analysis approach for estimating temperate mangrove forest cover using remote sensing and (2) to test the new approach by comparing its accuracy and uncertainty with those of traditional image analysis. The study area (around 1500 ha) is located in the southern part of the Waitemata Harbour, Auckland, New Zealand. Landsat images and field surveys were used for mapping, and uncertainty was quantified using a Monte Carlo approach. This study showed that, using a traditional approach of mapping, misclassification was the highest source of uncertainty (up to 19% for dwarf mangroves and 16% for tall mangroves), followed by water column effects (up to 7% for dwarf mangroves and 5% for tall mangroves) and positional errors (up to 4% for dwarf mangroves and 5% for tall mangroves). Improved image analysis enhanced accuracy from 72% to 95% for tall mangroves and from 69% to 90% for dwarf mangroves. The improved approach minimized the overall uncertainty by up to 68% for tall mangroves and 57% for dwarf mangroves. Adopting this innovative approach to image analysis can improve accuracy of estimates of long-term trends in temperate mangrove forest cover. International Journal of Remote Sensing, Vol. 39, No. 1. Hal. 17-36 ISSN 0143-1161