We have dual polarized radar data for a site with random natural vegetation and require to estimate the amount of vegetation biomass for each pixel in our imageries. Which models should we consider for the radar data? We additionally have RGB images of the scenes for some of the scenes and thus synergistic use of the images could also be considered. Unfortunately, we don't have any IR information.
Thanks Richard. I will be on the lookout for those publications. I will also post the approaches we eventually use for others to consider.
A combination of both SAR and MS is always a better choice for vegetation mapping/study. However, if you want to use only SAR then a temporal data set (time series) would be a better option rather than using a single date data.
Thanks Ruxandra. I have seen her impressive contributions and will be consulting these in addition to the works suggested earlier.
We looked into a multi-temporal approach using airborne L-band SAR data acquired every 2-3 days over a 3 weeks period. Frankly there was not much improvement when using such a dataset over single data retrieval. We have also tried retrieving biomass from different PolSAR decomposition components but the results were pretty similar with those obtained from backscatter alone.
The highest improvement in biomass estimation accuracy was obtained by increasing the mapping unit area. The mapping unit area was highly related to biomass RMSE ...and I mean here R2=0.92. We decreased the error from about 50% at 0.05ha plots to about 30% for 3 ha plots.
And something important, our average biomass is 60 t/ha, well inside the sensitivity interval for L-band data.
Once the paper is accepted for publication I will upload a version on RG.
Thanks Dr. Tanase, your comments are highly appreciated. I look forward to consulting your research once published. We will definitely use this knowledge you have generated.
- Badges
- Science topic
- Similar topics
- Geoscience
- Remote Sensing