Global Digital Elevation Models, Model Error Propagation and Uncertainty

Introduction

This is a continuation of the Kilimanjaro project. For the Kilimanjaro project, I showed the steps needed to model hydrology and channel networks of the Mt.Kilimanjaro in northern Tanzania in SAGA using SRTM data.

This week, I plan to model the hydrology and channel networks of Bali and Lombok islands, Indonesia. In this project, I plan to delve deeper into:

• using batch processing algorithms for SAGA tools in order to automate processing tasks;
• drilling into sources of error and compare elevation models (SRTM and ASTER) and their resulting hydrological models for error, error propogation, and uncertainty;
• using some new SAGA tools for our analysis;
• visualizing data error in the elevation models in QGIS.

Study area

Pura Ulun Danu Bratan Temple, Bali

Mount Rinjani, Lombok

Bali(5,780 square kilometers) and Lambok(4,514.11 square kilometers) are two small islands adjacent to each other located in the Lesser Sunda Islands of Indoensia. Although the two islands are small, both islands are highly populated with a total population of around 8 million residents. Both islands are volcanic in the North and rather flat in the south. The beautiful landscapes from both islands have attracted tons of tourists to visit. Bali, specifically, has become the cradle of Indonesian Tourism, with popular beach resorts, grand hotels and a focus in arts and crafts all over the island. Lombok, on the other hand, is less touristy, although Mount Rinjani, located in the north of the lombok island with an elevation of 3,726 meters, is the second highest volcano in the country and one of the most popular volcanos for hiking and viewing in the country.

Interestingly, the two islands are also different significantly both culturally and ecologically. The two islands are seperated by the Wallace Line, an ecological transitional zone between Asia and Australia species. Moreover, although Bali is mostly Hindu, Lombok is highly Islamic. The contrast in landscapes, cultures and ecological zones have made the two islands among the most interesting corners in the world.

Preparation step

Data Collection

For this project, I am going to use both DEM models:

• ASTER Global Digital Elevation Model, distribute by NASA EOSDIS Land Processes DAAC, https://doi.org/10.5067/ASTER/ASTGTM.003.
• NASA Shuttle Radar Topography Mission Global 1 arc second. 2013, distributed by NASA EOSDIS Land Processes DAAC, https://doi.org/10.5067/MEaSUREs/SRTM/SRTMGL1.003.

Besides the DEM models, I am also going to use the .NUM files for these DEM data, which can help explain errors and differences between data sources by telling us the sources of information for each location.

All data are dowloaded from earth data.

Software

The softwares used for the project are SAGA 6.2 and QGIS 3.8.

Creating SAGA grid files for the .NUM files

In order to use the .NUM files in SAGA, we can use this supporting python file to create a set of SAGA grid files. The python file imports a binary file as a SAGA grid, assigns the WGS 1984 geographic coordinate system to the grid and mirrors the grid.

Batch scripts

A bat file is a batch processing file that can run a series of commands in Window’s command shell. Since each SAGA tool can be run as a command, I wrote a bat script file that processes all the relevant SAGA tools needed for hydrology and channel network analysis.

Additional step for study region with ocean or large lakes

The region of Bali and Lombok contains large amount of ocean surface, which may make it impossible for SAGA to resolve the flow direction algorithm. To fix this problem, I reclassified all elevation of zero or below to nodata by using the tool Reclassify Grid Values (Grid-> Tools->Reclassify Grid Values).

Analysis

Classification of .NUM files

For my analysis of error, I need to visualize the number grids of the .NUM file of ASTER and STRM. The ASTER and STRM have different value meaning and categories according to the online documentation of ASTER and SRTM.

Classified ASTER .NUM file:

Classified STRM .NUM file:

Grid difference

Using the Grid Difference (Grid->Calculus) tool, I can subtract one set of input from the other. I have used the tool twice to calculate the grid difference between the elevation model of ASTER from that of SRTM, and the Flow accumulation of ASTER data from that of SRTM.

Elevation Model(ASTER-SRTM):

Flow accumulation(ASTER-SRTM):

3D models

SRTM 3D Model:

ASTER 3D Model:

Visualizing potential errors and causes

I tried to use QGIS to help visualize the potential errors that I may encounter.

Elevation/Steepness

Steep/High elevation region

Aster and SRTM data seem to be more consistent at higher elevation and steeper lands than at lower elevation and flat lands, although both SRTM and ASTER data did not seem to pick up some river sections at steep regions close to their origins. Several factors may lead to this problem. If there is a rather thick canopie along the river bank, then the setallite imagery may not be able to pick up the elevation of the river bed. Moreover, the river bed might be steep enough and the setallite imagery was not able to pick up the steep evelation of the riverbed.

Flat/Low elevation

There is a lack of sufficent data at extremely flat coastal areas, for example, in Southern Bali, although ASTER seem to give a more detailed, consistent, and comprehensive data at flat regions than SRTM. A possible cause that mess up the hydrological analysis in flat regions is the irrigation System. Bali in specific, is famous for its rice irrigation system. Called Subak, the Balinese water management system has not only been crucial to the Balinese agarian society, but is closely tied to Balinese tradition, religion, and Tri Hita Karana philosophy. The Water Temple of Pura Ulun Danu Batur (shown in Study area section), for example, is a part of the Balinese Subak System. Indeed, the Balinese Sabuk system was enlisted as a World Cultural heritage site in 2012.

Despite water management system’s importance to Balinese culture and society, we should be aware that this long cultural tradition and artifical ecosystem complicates the hydrology in the region.

Rice Terrace in Bali

Water body

Both Aster and SRTM data seem to mess up inside a water body, for example, at the crater lake. This may be caused by the fact that the program identify lakes as a piece of flat land and because water routes through water bodies are kind of random. The program encountered here may be related to the problem that I had earlier with the ocean. However, as long as the points of entry and exit from the lake are consistent, then there should not be a big problem because after all, we can visualize a polygon water body layer on top of the rivers layer in this case, and the total volume and direction of water flow will still be consistent.

Flow accumulation

From the flow accumulation Grid difference data, both ASTER and SRTM data seem to be more consistent when the flow accumulation is small than when the flow accumulation is large. This makes sense because a small mistake may accumulate to a large one as flows accumulate.

Back to Main Page