The fact that surveys cannot provide 100% lake coverage, EcoSound utilizes a geostatistical procedure called kriging that analyzes various spatial properties of the data and models and their relationships. These models are used to predict vegetation biovolume at unsampled locations and create a uniform map. Kriging creates a smooth surface and actual data points may be slightly higher or lower than estimated values. Lake vegetation changes constantly throughout the year, so smoothing out peaks and valleys is important.
Furthermore, the fluid properties of water subject to wind, waves, and currents create a sampling environment where repeatability of survey results in less than a 1 m2 area is unrealistic in most circumstances.
Consequently, we employ kriging as a statistically robust way to characterize the general nature of the bottom during the sampling trip. Users can increase the accuracy of output maps by driving slower and travelling along closely spaced transects. In these cases, actual depth and vegetation values in sampled locations will be preserved in the output grid. As neighbor points become more distant or variable, smoothing will increase.
To minimize interpolation error, by default, map outputs are not generated past 25 m of your track. Users can either collect more data in these “blanked” areas or increase the buffer by editing your trip in BioBase. For more information, refer to the Trip reprocessing section or the Merge trips section described in this manual.
The default kriged vegetation heat map has been tested in a variety of cases and should produce a robust snapshot of biovolume. However, researchers may want to have a closer look at the spatial relationships of the data and do their own modeling. This can be done by exporting the comma delimited point data along their GPS track from EcoSound and then import it into any third-party GIS or statistical software.