SRTM and Aster Elevation Data

What's going on with SRTM and Aster global elevation models?

SRTM digital elevation model V003: https://lpdaac.usgs.gov/dataset_discovery/measures/measures_products_table/srtmgl1n_v003

Aster Global digital elevation model v002: https://lpdaac.usgs.gov/dataset_discovery/aster/aster_products_table/astgtm_v002

To conceptualize the lab exercise last Monday (February, 26), it might be a good idea to check out ArcGIS' hydrology toolset http://desktop.arcgis.com/en/arcmap/10.3/tools/spatial-analyst-toolbox/an-overview-of-the-hydrology-tools.htm (Maria)

In my group we struggled a bit to conceptualize the elevation models in relation to possible errors. When we analyzed the aster elevation data we noticed some channels had higher elevations as they went downhill, which confused us because it should be just the opposite. Our error raster for the SRTM data then perfectly matched up with what we saw as possible errors in the aster elevation model. The error raster for the aster elevation model had high levels of possible error in a similar area but it did not match as well as the SRTM data. We did not get to spend much time comparing the channels between the two data sets because we were so stuck on how the errors did not match up as we would have expected. We did not come to any conclusions as to why the errors played out that way but perhaps it has something to do with how the data is collected differently between the two data types. Unfortunately, we did not have enough time to investigate possible explanations. -Hannah K. ^I was just reading about how the ASTER can display large elevation inaccuracies on local levels (which was discussed on the provided link on ASTER) and then continued to research how this could potentially influence the channel network data that we mapped. I found an article that discussed the ways in which the horizontal and vertical spatial resolution may be insufficient for more localized and detailed channel networks. Perhaps this variability relates to strange elevations you were noticing. (Casey)

While researching the differences between SRTM and Aster data, I stumbled upon a press release from NASA's JPL in California. This release is brief, but brings to light some of the overlying issues regarding releasing informational data to those in developing countries. The press release states that the release of NASA's SRTM 30 m data will "empower local authorities" in Africa. The release ends by detailing that "training workshops on SRTM data are planned for users in Africa." While the sentiment behind releasing the data and training users to be proficient in SRTM is not intentionally negative, it highlights the ignorance that many possess. Attempting to level the playing field by releasing data is positive, yet providing a few training sessions for only those fortunate enough to attend will not be beneficial for a wider audience. These trainings provided by NASA and USAID seem slightly colonalist in nature, which seems to counteract the desire to "empower local authorities" by creating a hierarchy between those teaching and learning, and those who were able to attend the sessions and those not. Press release here: https://www.jpl.nasa.gov/news/news.php?release=2014-321. - Miranda

When overlaying our final channel outputs on satellite imagery in QGIS, we found the results of the Aster DEM to represent actually existing channels better than the STRM data. While the Aster data seemed to follow the paths of channels quite accurately with decent precision in most cases, the STRM data had many cases that did not match the real world. Especially in areas of more level topography, the STRM DEM yielded a series of fragmented channels that did not compose a network of flows like the Aster channels. In comparing the methodology behind each DEM, Aster uses a finer grain (60 by 60 kilometer) single-scene DEM compared to the much broader (225 km wide) STRM swaths. In reading the description for each data source, this is one possible artifact we believed may have contributed to lesser accuracy in our STRM data, but we would like to investigate other possible causes in further detail. (Oliver) --We had similar findings, and I think you may be onto something when you attribute some error to the STRM's broader grain than Aster. Is this the same issue that was discussed in chapter 4 of People and Pixels with the Amazon Rainforest satellite/demographic data overlays? I remember reading that in an ideal world, there would be a level of compatibility between layers that would confirm that the independent and dependent variables were "linked at the level of the decision unit involved." Could this possibly be a similar case? If I read the chapter correctly, it was saying that one of the major causes of error in the Amazonian deforestation project was the fact that they had to reconfigure satellite data to conform to larger/more irregular geopolitical boundaries that dictated the demographic data. Although that is not what we are doing here with the channels (I think?), this idea of different grain densities (is this the way we talk about grains?) could be important as we look at possible error sources. (Matia) Yes, it seemed like resolution is a major source of error. Maybe it's also contextual on the terrain/biome analyzed. Some other people on the internet have been saying ASTER has large artifacts such as peaks in flat terrain that is difficult to correct. This paper shows the SRTM is more accurate in the context of the forest of the Malaysian Borneo than ASTER. Furthermore, maybe we should use a tool to remove noise in the topographic data. (Steven)