Repository navigation
Tmrt output seems dominated by SVF rather than actual sun/shadow pattern (isotropic sky)- possibly user error, would love your input #880
Unanswered
Benjamin-neu
asked this question in
Q&A
Replies: 2 comments 2 replies
|
Hi Benjamin, |
2 replies
|
Thank you very much for the valuable feedback. Thank you very much for your quick and precise help. |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Hi UMEP team,
First of all, thank you for this fantastic and freely available tool. As a Master's student in landscape planning, being able to run PET/Tmrt analyses without expensive software has been a huge help.
I'm currently testing the SOLWEIG workflow with QGIS 3.44.8 (Windows) before committing to it for my thesis, which will analyze pedestrian-level heat stress (Tmrt/PET) in two residential areas in Baden-Württemberg, Germany, as a basis for evaluating heat-adaptation measures. Before starting the real analysis, I wanted to do a small sanity-check test run, and ran into a result I can't explain, so I'd really appreciate your input on whether I'm missing something or whether this points to a genuine issue.
Setup:
Single 1×1 km DSM/DEM tile (official German LGL open geodata, ETRS89/UTM32, 1 m resolution) No CDSM, no land cover grid Minimal synthetic met-file (global radiation only, two timesteps), isotropic sky, UTC+2
What I observed:
Using QGIS's "Identify Features" on the raw Tmrt pixel values, a spot I know from the aerial photo to be in full midday sun comes out cooler (~20 °C) than a spot clearly in building/tree shadow (~24–27 °C), the opposite of what I'd expect.
What I tried to rule out myself (so you don't have to check these):
Merged multiple LGL tiles vs. a single unmerged original tile → same pattern either way
DEM in .xyz format vs. potential coordinate mismatch → extents match DSM exactly, cross-checked directly in the raster
Wall height/aspect rasters accidentally swapped → fixed, confirmed correct in the log, pattern persists
Reproducible in both SOLWEIG v2025a (UMEP for processing) and v2022a (menu-based UMEP), same inputs, same result
What I found that might be a useful clue:
I re-implemented the shadowing algorithm (shadowingfunction_wallheight_13) myself with a correctly computed sun position for the test time/location, and it does reproduce the expected sunlit/shaded pattern on this DSM. But across the whole raster (~1M pixels), this "correct" shadow pattern only weakly correlates with the model's actual Tmrt output (r=−0.55, wrong sign for what I'd expect).
Much more strikingly, Tmrt correlates almost perfectly with the SVF raster alone: Tmrt ≈ 29.24 − 10.20 × SVF, R²=0.998 (residual SD only 0.08 °C). It looks as if the actual per-timestep shadow/direct-beam contribution barely influences the final result here, while SVF (which doesn't depend on the sun's position at all) seems to dominate almost entirely.
I'm very open to the possibility that I've set something up wrong on my end. I'd just love to understand what, since I'd like to use this tool with confidence for my thesis. Happy to share the DSM/DEM tile, met-file and Tmrt output if that would help you take a look.
Thanks so much for your time and for maintaining this tool!
Best regards.
Benjamin
All reactions