Detecting trends in seagrass abundance using aerial photograph interpretation
Aerial photographs have been widely used to locate the presence of large estuarine plants for over three decades. For example, intertidal saltmarsh and mangrove are easily located in aerial photos and subtidal seagrass can be mapped as long as water clarity is adequate. Saltmarsh, mangrove and seagrass are known to have value in nurturing and sustaining estuarine fish. For coastal-scale monitoring projects, photos remain the data source of choice due to their low cost, but the technology used in their analysis has evolved considerably over the past 20 years. This improvement in analytical technology is important when historical and current day maps of the distribution of estuarine plants are being compared.
In 1985, an inventory of seagrass distribution in 130 NSW estuaries was done by applying simple photogrammatic methods (the camera lucida technique) to aerial photographs. This involved the use of an desktop system of optics to trace relevant elements from an aerial photo to a base map. If the 1985 maps are to be used as a baseline against which to assess trends in cover of seagrass over the ensuing decades, their accuracy with respect to the current accuracy of the new technology needs to be determined. Therefore, relevant historical photographs of selected estuaries were re-examined using a modern geographic information system (GIS). GIS analysis involves the scanning, rectification and digitising of air photos, with the digitising done to a magnified image on the computer screen.
Four estuaries were selected to have their historical photos re-examined. These were Port Hacking, St Georges Basin, Bermagui River and Merimbula Lake. Camera lucida produced a larger estimate of seagrass area than GIS in all four locations: 8% (12.9 ha) greater at Port Hacking, 243% (502.2 ha) for St Georges Basin, 15% (5 ha) for Bermagui River and 20% (38.0 ha) for Merimbula Lake. The consistent but moderate discrepancy in Port Hacking, Bermagui River and Merimbula Lake was attributed mainly to systemic differences between camera lucida and GIS techniques, such as the capacity of the latter to magnify images of the aerial photos. Magnification overcomes the tendency of the camera lucida technique to overestimate area by amalgamating patches of seagrass into continuous meadows. The large discrepancy at St Georges Basin was attributed to operator error; that is, the operators who mapped in 1985 and 2003 had widely differing interpretations of the aerial photographs, particularly of deepwater seagrass beds and species of seagrass with low canopy height, such as Halophila spp.
A twenty-year trend was calculated for each estuary using either the 1985 camera lucida or their revised GIS values as the starting points. Management intervention depended greatly on which of the two initial data points were used. In St Georges Basin, a catastrophic decline (65%, 553.9 ha) occurred based on the camera lucida / GIS comparison, but only a small decline was seen (14.7%, 51.8 ha) when GIS was used to assess historical as well as current photos. This result indicates that the evaluation of change in seagrass abundance using remote sensing needs quality control mechanisms to reveal errors in historical maps and prevent misinterpretation of trends.