Delineating the mountain cloud forest of Taiwan by means of topographic cloud immersion with moderate resolution satellite data and ground based observations
Final Report Abstract
Mountain cloud forest (MCF) is one of the most endangered ecosystems on earth due to deforestation and a global decrease in low cloud frequency. At the same time, it plays an important role as an ecosystem service provider and offers habitats to a wide range of endemic species. For its protection and further exploration, global efforts are needed to map this ecosystem. An area-wide mapping of this ecosystem has also been claimed in the Cloud Forest Agenda of the UNEP World Conservation Monitoring Center, but was still lacking. Therefore, the aim of this project was the development of a satellite-based approach to map MCF in Taiwan. It is based on monthly maps of ground fog frequency since a major criterion for MCF is the frequent and persistent cloud immersion. Existing remote sensing techniques for detecting ground fog are not applicable to Taiwan. This is due to assumptions on the clouds microphysics that are not valid for the low stratus clouds in Taiwan that are mainly of advective-topographic nature. Therefore, a new method for the detection of ground fog in mountainous areas (DOGMA) has been developed. DOGMA is based on a negative correlation between terrain height and the cloud optical thickness of a cloud in areas where the cloud touches the ground. It operates on the basis of MODIS data, which have been sharpened to a resolution of 250 m using the high resolution MODIS channels 1 and 2. For 10571 MODIS daytime scenes between 1 January 2013 and 31 December 2014 over Taiwan, areas with ground fog were detected using DOGMA. From this data maps of ground fog frequency were calculated for each month. To create an area-wide map of MCF in Taiwan a raster data set was compiled from the DOGMA ground fog frequency maps, parameters derived from a DEM and Landsat 7 imagery. A random forest model trained with plot data from the National Vegetation Database of Taiwan was used to create a map of MCF conditions from these data set. Based on the trained model suitable mountain cloud forest conditions for the entire Taiwan in a resolution of 250 m per pixel were mapped. The resulting map was validated using an out-of-bag approach with data from the National Vegetation Database of Taiwan. It has been shown that the DOGMA-based ground fog frequency is suitable for the mapping of mountain cloud forest conditions by simply applying a threshold value. However, the inclusion of other variables by the random forest approach significantly increases the recognition quality. The map of the mountain cloud forest conditions was finally combined with a forest map derived from Landsat data to produce an area-wide MCF map for Taiwan in 250 m resolution. The map of MCF distribution created in this project is valuable benefit for conservation management as well as for research on MCF in Taiwan. The achieved knowledge gain comprehends two new approaches (including respective scientific software) to detect cloud immersion with (i) time lapsed cameras and (ii) satellite data, as well as the first high resolution MCF map of Taiwan as the basis for manifold future research (MCF development under environmental change) and application activities (e.g. water supply, nature conservation).
Publications
-
(2017) Mapping the montane cloud forest of Taiwan using 12 year MODIS-derived ground fog frequency data. PloS one 12 (2) e0172663
Schulz, Hans Martin; Li, Ching-Feng; Thies, Boris; Chang, Shih-Chieh; Bendix, Jörg
-
(2014): Automatic cloud top height determination in mountainous areas using a cost-effective time-lapse camera system . Atmospheric Measurement Techniques 7, 4185 - 4201
Schulz, M.; Thies, B.; Chang, S. & Bendix, J.
-
(2015): Frequency of low clouds in Taiwan retrieved from MODIS data and its relation to cloud forest occurrence. Remote Sensing 7, 12986-13004
Thies, B.; Groos, A.; Schulz, M.; Li, C.; Chang, S. & Bendix, J.
-
(2016): Detection of ground fog in mountainous areas from MODIS (Collection 051) daytime data using a statistical approach. Atmospheric Measurement Techniques 9, 1135 - 1152
Schulz, M.; Thies, B.; Chang, S. & Bendix, J.
