I've spent the last week reviewing ESRI Videos, Help Documentation and read through a great portion of Konstantin's book "Spatial Statistical Data Analysis for GIS Users".
I have a few questions to validate my understanding of the geostatistical tools and workflows and how I can use them to model uncertainty propagation from SRTM DSM (Digital Surface Models) for hydrological modelling:
- Is there anyway of producing equally probable realisations using the Empirical Bayesian Kriging using the PREDICTION_STANDARD_ERROR output?
The reason for looking to see if its possible to create equally probable realisations from the Empirical Bayesian Kriging interpolation is that its already creating numerous simulations and determining the best fit semivariogram based on the input points. I'd have to convert the SRTM DSM into points to be able to model the uncertainty within the SRTM DSM based on ground control points from a LiDAR Survey.
- Is it possible to using the Empirical Bayesian Kriging PREDICTION as input to Gaussian Geostatistical Simulations to generate equally probable realisations of my DEM's?
The reason for seeing if its possible to use the output PREDICTION raster from Empirical Bayesian Kriging as input to Gaussian Geostatistical Simulations is that it would contain the least error in converting the SRTM DSM to a Geostatistical Layer
- Is there any workflow in getting around the current limitation of the raster cell size (2049 X 2049) ?
My hydrological watershed study areas are definitely going to be larger than 4.2 million cells
Summary:
I'm trying to make use of the tools within Geostatistical Analyst to accomplish my goals of generating equally probable realisations of the SRTM 30m , 90m and hydroSHEDS that I then can use to generate watersheds, stream networks and longest flow paths . The results from each simulation will be used to generate a probability distribution of the derived watersheds, stream networks and longest flow paths in order to quantify and visualise the uncertainty.
Your advice and guidance to achieve the following will be appreciated.
Regards
Peter Wilson
