Dear all,
A challenge I have found in finishing my Thesis is the following; I have several hydrographic transect profiles (point files with distance corresponding to depth) provided in Excell format. These include three different data sets, one being Multibeam Echosounder (MBES) or the ''truth'' and two datasets i am looking to compare to the reference MBES set, these are Landsat-8 and Sentinel-2 satellite derived bathymetry (see attached).
To describe the strength of the linear relationships between the different profiles i opted to use the Karl Pearson’s correlation coefficient r , with the guide that Evans (Evans, R.H., 1996) suggests for the absolute r values. Where; .00-.19 is “very weak” .20-.39 is “weak” .40-.59 is “moderate” .60-.79 is “strong” and .80-1.0 is “very strong”. In normal operations Im used to working with statistics between actual data layers and not specific transect profile samples.
The result is that all these profiles reached strong and very strong correlations. From a Hydrographers/visual point of view these terms seem too subjective or positive.
My question; is there a ''better'' way of quantifying the similaritie/differences between these 3 transects than Pearson's r?
Kind regards,
Leon
Since you assume that the data obtained from the Multibeam Echosounder (MBES) are the ''truth'', you can do residual analysis. Take the S2 measurement as example. The difference between the measured values of S2 and MBES is the residual. Ideally, the residuals should be approximately normally distributed around zero. If the residuals are not centered at zero, there is a bias in the S2 measurement. On the other hand, the standard deviation (or variance) of the residuals measures the relative "precision" of the S2 measurement. So, reporting the bias and standard deviation will provide a picture about the quality of the S2 measurements. The same residual analysis can be done for the L8 measurements.
- Badges
- Science topic