What spatial resolution is required for mapping kelp "forests" with satellite imagers? Can kelp biomass be estimated using data from multispectral or hyperspectral imagers on satellites or aircraft?
On sailing through the kelp forest offshore of La Jolla and San Diego one notes a suppresion of whitecaping within the areas of kelp growth relative to the waters surrrounding them. Whether this represents mechanical damping of surface waves by the kelp blades , or is related to surfactant algal lipid release is problematic, but the difference in sea surface albedo may be observable both in optical and microwave imaging .
I'm not a specialist in marine RS, but if the kelp is not deep and the water sufficiently clear then I would think this can be done. The issue of spatial resolution is probably secondary to that of identifying the spectral signature of the kelp through the water and distinguishing it from other things like plankton. Which means you will have to live with the spatial resolution you can get.
The trouble is that InfraRed and Near Infra Red bands will be largely absorben by the water so methods for vegetation detection designed for land probably wont work, but a supervised classification for multiple sensors (optical, IR and perhaps even radar) could pick it up.
Russells suggestion of using the whitecapping as a proxy is a nice idea, but I wonder whether it would be consistent enough in terms of the relative difference with non-kelp areas for different bathymetry under different sea conditions?
A brief google suggests there is quite allot of work on this but it needs quite allot of background info such as bathymetry to distinguish the kelp from other things http://epic.awi.de/25251/
I suspect you could map x,y extent using satellite imagery. However, anything to do with volume (total biomass) would require other data. Perhaps some sort of an estimate based on bathymetry?
Hi there,
I agree with the previous comments. First, you should identify the depth of the kelp forest. So long it is located within the depth of penetration (DOP) of the image data that you will use, it should work. However, since each wavelenght has different DOP, as the water goes deeper, the number of usable band will be limited to the shorter wavelenghts.
The spatial resolution is also another issue. Depending on the nature of the kelp forest, at lower resolution, you may got a lot of mixed pixel. This can hinder the process of biomass estimation since the variation in kelp pixel is not only the function of kelp forest variation, but also due to the variation of other objects composing the same pixel. Also, if the kelp forest is narrow, using lower resolution may not be effective. my suggestion is to match the specification of the image that you will use with the environmental condition of the kelp forest.
Water quality in the environment should also be considered.
If the water is clear, then it is fine. But, if the water is turbid, it would be problematic since most downwelling irradiances will be scattered by the suspended material before reaching the kelp.
sunglint should be minimized, and you need infrared band to do that (See Hedley et al. 2005). Water column correction may or may not be necessary depending on the depth of the kelp forest.
TO perform biomass mapping, you can try to empirically model field biomass data and various image transformation i.e. visible-based vegetation index and PCA.
You can use the resultant regression function to convert image pixel value into biomass.
you can see my publication about seagrass LAI mapping as reference. Thank you.
- Badges
- Science topic
- Similar topics
- Geoscience
- Remote Sensing