dem_error_study

Code and results of Hugonnet et al. (2022), Uncertainty analysis of digital elevation models by spatial inference from stable terrain. 🌐 🗻

The dataset is available at https://doi.org/10.5281/zenodo.7298913.

Below a short guide to: perform the uncertainty analysis of your own DEMs, retrieve the case study datasets, reproduce the processing steps with the case studies, reproduce the figures and tables of the paper.

Uncertainty analysis of your own data with xDEM

The methods of this study are implemented in xDEM, in routines that can be used to perform uncertainty analysis of your own DEM data.

There are three basic examples with:

• A single-line pipeline to estimate an elevation error map,
• A single-line pipeline to estimate the spatial correlation of errors,
• A single-line pipeline to propagate elevation errors spatially.

Additionally, there are three advanced examples with:

• The detailed steps to estimate and model the heteroscedasticity of elevation errors,
• The detailed steps to estimate and model the spatial correlation of elevation errors,
• The detailed steps to standardize elevation differences to use stable terrain as a proxy.

The full documentation of xDEM can be found at https://xdem.readthedocs.io/.

Note at the date of 07.11.22: xDEM is still in development (version 0.0.7), and its documentation in construction. Compared to version 0.0.6 used in this repository, several changes were added to xDEM late 2022, including:

• Construction of error pipelines that combines all steps,
• Addition of existing gallery examples and streamlining of existing ones,
• Minor fixes and improvements of routines.

Retrieve the case study datasets

The dataset consists of:

1. Nearly-simultaneous Pléiades–SPOT-6 elevation differences at the Mont-Blanc massif, the Pléiades DEM used as a reference for alignment and deriving terrain attributes, the SPOT-6 DEM (.tif, ~1 GB) at 5 m posting and the forest mask derived from ESA CCI (.shp, ~5 kB) available at https://doi.org/10.5281/zenodo.7298913.
2. Nearly-simultaneous ASTER–SPOT-5 elevation differences at the Northern Patagonian Icefield, the ASTER DEM used as a reference, the quality of stereo-correlation out of MicMac, and the SPOT-5 DEM at 30 m posting (.tif, ~150 MB) available at https://doi.org/10.5281/zenodo.7298913.

Reproduce the processing steps with the case studies

Setup environment

Most scripts rely on the code assembled in the package xDEM which in turns relies on SciKit-GStat. Some routines also rely on GSTools. You can rapidly install a working environment containing those packages and their dependencies with the environment.yml file, located at the root of the repository, using conda:

conda env create -f environment.yml

Note: Due to continuous development changes, xDEM is set to v0.0.6 in the environment file to exactly reproduce the processing steps of the paper.

How to use

Scripts for reproducing the processing steps are located in case_study_montblanc/ or case_study_npi/. Those are generally quite short as they use one-liner routines of xDEM. Some computations (e.g., simulation of correlated field) are performed only in the figures/ scripts. In the future, those might be integrated in xDEM.

While all scripts are commented, the details on the functions used are available through xDEM's documentation, SciKit-GStat's documentation and GSTools's documentation.

Reproduce the figures and tables of the paper

Scripts for reproducing the figures and tables are located in figures/. These scripts also depend on the environment file environment.yml as they rely on Cartopy and Seaborn in addition to Matplotlib. In some occasions, the figure scripts duplicate the processing steps done in case_study_montblanc/ or case_study_npi/ for plotting purposes (e.g., violin plots require the full distribution of samples, not only the binned estimates of dispersion).

For plotting figures of your own data, xDEM provides simpler plotting tools of binned data and variograms (see example gallery).

Enjoy! 🌋