Analysis of medium and small scale Digital Surface Models with respect to slope and aspect

The work is about preliminary analysis for DSM errors distribution evaluation.

Three digital models were taken in consideration: the SRTM DSM (step around 90m), the ASTER DSM (step around 30m) and the DSM provided by Lombardia Region (step 2m).
The analysis was performed in an area including the walled city of Como and part of the Como province around; in the chosen area a LiDAR dataset, provided by the Lombardia Region, was also available.

At first the LiDAR data were filtered in order to compute a DTM and a DSM characterized by a higher spatial resolution (step of 2m) so they were used as ground truth, considering that he nominal vertical accuracy of the dataset is about 20 cm.

All digital models were three-dimensionally readjusted using points from the LiDAR dataset to remove 3D effects with a Matlab software developed at the Geomatic Laboratory of the Politecnico di Milano.

Once removed three-dimensional effects from the LiDAR derived digital model, 33 classes were obtained from the combination of four slope classes and eight aspect classes (the last class corresponds to flat terrain).

The digital models from SRTM and ASTER as well as the Lombardia Region DSM were compared with respect to the 33 classes in order to evaluate, through two-way ANOVA (Analisys Of VAriance), the dependency on slope or aspect that may be due to survey geometry or to the way a model was generated.

Before and after the readjustment procedure the RMSE between the models and the DSM from LiDAR were computed to evaluate accuracy variations.

In the last part of the work possible DSM errors depending on terrain coverage were evaluated using terrain coverage information DUSAF obtained from Regione Lombardia.