Freshwater/Brackish Wetland

Wetlands were mapped principally during the supervised classification which utilised input from known wetland areas delineated by Dudgeon and Chan (1996). The resulting classification was also cross-checked with the sites identified by Lau (1998) for amphibian habitats.

Several difficulties arose when mapping wetlands. The spectral property of wet, or damp, vegetation is very similar to that of wetlands and discrimination between the two required input from technical specialists familiar with the area. This expert judgement allowed non-wetland sites to be distinguished and reclassified to other habitat types. The resulting areas were checked and found to be more accurate than the original classification. Another difficulty is that since wetland boundaries are expected to vary between the wet and dry seasons and the satellite imagery data set is not expected to account for this, boundaries of the mapped wetland areas are inexact. The habitat is thus considered to have a relatively low level of mapping confidence.

Fishpond/Gei Wai

This habitat type could not be mapped using computerised classification techniques. This is because gei wais are often indistinguishable from surrounding waterbodies, and fishponds do not always contain water as some of them may be drained during the period when imagery was collected. Therefore this habitat type was mapped primarily through recourse to the existing data (Table 4.2d). The results of the mapping were then checked against the aerial photos. The category is thus regarded as being mapped with a high level of confidence in terms of habitat type and boundary.

Natural Watercourse/Modified Watercourse

As watercourses may not have contained water on the date the imagery was collected, the imagery data were not relied upon for mapping of watercourses. Instead, 1:1K topographic data (Lands (1997 1:1K)) were used. The Lands Department data does not distinguish between natural and physically modified watercourses (e.g. drainage channels) and so manual input with reference to the 1:20K topographic data (Lands (1997 1:20K)) was carried out. Further adjustment of watercourses into the modified category will depend on the pollution status of the watercourses based on the existing information from the EPD River Water Quality Monitoring database. Despite the detailed Lands Department data, watercourses are considered to be mapped with medium accuracy since some of the junction areas for modified and natural watercourses cannot be clearly identified.

Mangrove

This habitat was mapped through supervised classification using existing information (Table 4.3b) and aerial photos as the supporting data sources. In Tam and Wong (1997), a total of 43 mangrove stands have been identified with 23 of them studied in detail. The boundaries of mangrove habitats identified from the literature corresponded closely with the imagery classification and this habitat is therefore mapped with a high level of confidence.

Rocky Shore

As the spectral properties of rocky shore are very similar to the categories of bare rock or soil, artificial rocky/hard shoreline, landfill and quarry, the habitat cannot be discriminated using computerised classification alone. However, natural rocky shore was mapped accurately through manual identification using existing topographic data (Lands (1997 1:20K) and (1997 1:1K)), aerial photos and the expertise of the Study Team. Mapping confidence is thus high. As the spatial extent of natural rocky shores is usually less than 20 m wide which is below the required mapping accuracy of the Study Brief (i.e. ± 20 m), the habitat was represented as a linear component on the habitat map.

Artificial Rocky/Hard Shoreline

Artificial rocky/hard shoreline were mapped primary based on manual identification using existing topographic data (Lands (1997 1:20K) and (1997 1:1K)) and aerial photos. This habitat type can often be easily distinguished from natural shores by its relatively straight coastline. The habitat is thus mapped with a high level of confidence. As all artificial rocky/hard shorelines normally have their spatial width less than 20 m wide which is below the required mapping accuracy of the Study Brief (ie ± 20 m), they were represented as a linear component on the habitat map.

Intertidal Mudflat

As intertidal mudflat area are spectrally indistinguishable from the surrounding waters, classification based solely on satellite imagery alone was not sufficient. As a result , this habitat was mapped primarily by manual digitising of identified intertidal mudflat habitats compiled through review of existing topographic data (Lands (1997 1:20K)), aerial photos and specialist input. As the landward side of mudflats is usually fringed by mangroves, locations of mudflats were checked against the existing 43 mangrove stands in Hong Kong as described above (Tam and Wong 1997). Those possible flat areas not indicated on the topographic data were cross-checked with the aerial photos. These steps improved the accuracy of the mapping to a medium level of confidence.

Sandy Shore

As sandy shore habitat cannot be discriminated from the bare rock or soil category using imagery classification, this habitat was mapped through recourse to manual identification of the habitats based on the existing topographic data (Lands (1997 1:20K)) and (1997 1:1K)) and the expertise of the Study Team. Aerial photos were cross-referenced as necessary. Using these techniques, sandy shore habitats are considered to be mapped with a high level of confidence. As the spatial extent of sandy shores, with the exception of beaches, is usually less than 20 m wide which is below the required mapping accuracy of the Study Brief (ie ± 20 m), the majority of the habitat were mapped and represented as a linear component on the habitat map. For sandy shores which have widths greater than 20 m, interpretation of aerial photographs was conducted to map these beaches as polygons which are spatial in nature.

Seagrass Bed

As areas of seagrass beds are difficult to be distinguished through satellite imagery classification and aerial photos, they were mapped primarily based on existing information provided by the ACFD Herbarium and published materials on seagrass ecology (Table 4.2d). Since information on the extent and location of existing seagrass beds are sufficiently comprehensive and up-to-date, the habitat is mapped with a high level of confidence.