It sounds like the answer should be NO.

You shouldn't expect global DEM with very high resolution and open access.

However, local DEM generated from high temporal resolution of optical (PlanetScope, Pleiades, WorldView, SPOT...) images is possible. The DEM generated from SAR images (TerraSAR, COSMO SkyMed, ALOS-2) is high resolution as well.

Pietro Milanesi This catalog is pretty cool, thank you!

Since costs go up exponentially as resolution increases, probably not. The orbital SAR technology has existed for decades to use very tight beam from multiple look angles and interferometry to probably get down to inches ( https://spacenews.com/umbra-15-centimeters/ ), but a DEM is not a simple artifact like a optical image, there is a massive amount of computation and error checking that has to happen, so nobody is going to expend funds to map vast areas for no reason, let alone keep it updated, especially when airborne Lidar is becoming vastly more available, DEMs and point clouds ( https://www.usgs.gov/3d-elevation-program ).

When you say 'out of date', what is the use case where that would matter? The current free DEMs are composites of all those you mentioned ( if you look at the source layer it will tell you which for any part of the raster ), and they update them as new data comes in and they refine algorithms ( NASA has some excellent papers on the history and preparation processes ). There are certainly 'not talked about' political and military considerations in play also.

Michael John Patrick thank you for your reply. In the case of 'out of date', I was referring specifically to SRTM and ALOS PALSAR data. We use these global DEMs, for example, to generate watersheds in the Horn of Africa. The problem is that due to the high dynamics of natural processes in this area, the relief changes very rapidly. Quite often, we find that the waterdhed in the current relief is shifted by several kilometres more than the generated one. Therefore, I wonder if there are any updates or new global DEMs being created so that we can better fit our modelling to the current relief.

Vilém Spálovský These a generic observations, since I am not familiar with the details of your project.

The first is that 'The Horn of Africa' is a somewhat large space so knowing the Area of Interest where you are noticing the inconsistencies, 'high dynamics' could be landslides on near vertical slopes or flashflood gullies an nearly flat terrain.

The second is terminology, people use 'DEM', 'SRTM', ALOS PALSAR' in a very generic sense, but the span of those is over twenty years now, with multiple versions being released with different processing levels, and all along various NGOs and other Communities of Interest creating their own derivatives, and redistribution of those through various other portals and products (especially hydrological adjusted DEMs using various algorithms). And it isn't only the data that is important, the metadata and documentation ( for example The 2020 NASDEM user guide https://lpdaac.usgs.gov/documents/1318/NASADEM_User_Guide_V12.pdf of note is section 5 NASADEM FINISHING) . This applies to both baseline already being used and the latest datasets that are being used for comparison.

Third, one simply can't get around reading, at what ever level of understand the documentation, Ex. (Note: NASA SRTM Version 3 is not to be confused with “SRTM Versions 3 and 4” produced by CGIAR as their own follow-on DEMs to the NASA JPL and NIMA/NGA produced SRTM Versions 1 and 2.).

Fourth the 'DEM' isn't just elevation data, it is a collection of datasets - slope, aspect angle, profile curvature and plan curvature products and various supporting information like source, scene count, error, etc. ( Table 2

NASADEM products and their groupings ). Any single one, or combination of these might be useful resolving a situation for a particular use case, depending if I was interested in ridge lines versus flow direction, surface convolution, etc.)

Fourth is the analytical processing pipeline, basically delineation change areas and the sourcing more recent data to fuse into the model, and making an assessment whether that affect the significant outputs. For instance, I've used sequential SAR scenes crudely georeferenced to identify change areas ( The ex. is ALOS Palsar, but Sentinel is even better ), and if justified you could create a newer DEM for just those areas ( not easy). Or maybe make an academic request from AirBus Defence for that limited area, or maybe some other research project or commercial firm like a mining company has flown Lidar in the area.

But you can't get around understanding the data you are using, especially if the resolution is coarse compared to the features. Even definition of 'height' of a given pixel is questionable - is it the max, min, an average, calculated with surrounding pixels, etc. A the SRTM return from very flat area incised with narrow and deep gullies might create a dataset that appears to noise pixel to pixel, but by looking at aspect or curvature would reveal correct delineation.

Michael John Patrick Thank you for your very detailed and helpful reply.

Vilém Spálovský You might also look into: FABDEM (Forest And Buildings removed Copernicus DEM) is a global elevation map that removes building and tree height biases from the Copernicus GLO 30 Digital Elevation Model (DEM). https://data.bris.ac.uk/data/dataset/25wfy0f9ukoge2gs7a5mqpq2j7

Paper at Article A 30m global map of elevation with forests and buildings removed

Michael John Patrick Yes, thank you. That might be helpful.