The model does not create the correct streams. There are sharp diagonal shards on the maps that should not be there.
InVEST-seasonal_water_yield-log-2026-06-04–10_20_51.txt (82.8 KB)
I have tried different DEM’s, higher and lower threshold flow accumulation (higher looks better but still incorrect). I have aligned all rasters, crs’s are the same.
How do I solve the issue?
Welcome to the forum @Livia !
Thanks for including your log file, it lets us know that you’re using the MFD flow direction algorithm, which creates somewhat different streams than D8. Have you tried the D8 flow direction algorithm? How do the streams compare with those produced by MFD?
It would also help to see an image of the stream network, which you can upload with a forum post.
Are the diagonal shards being created in areas that are very flat? It’s very hard to model streams in floodplains and other places that are very flat, with no elevation change to tell the model where streams would collect. So check that out too and let us know. Thanks!
~ Stacie
Hi, thank you for your response.
Using the D8 algorithm gives much smaller shards but also in more directions. TFA = 10000 in image below
The area is indeed very flat, not only where the shards are but the whole study area. Increasing the TFA has helped in reducing the shards but I am not sure if the maps are then still correct. The maps with a TFA of 200000 do not show any of the shards.
Thanks for the images @Livia , those help a lot.
If those areas are actually very flat, then i’m sorry to say that the model is going to create streams that are a bit funky. D8 makes multiple straight lines in flat areas, MFD makes wider streams. You can search this forum for previous discussions about this.
It’s useful to see where the LULC map says there is water, but the modeled streams are not based on the LULC map, only the DEM and the Threshold Flow Accumulation value provided as input.
If you want the streams to come out in a particular way, you can try “burning” them into your DEM using some other spatial source for stream network. Personally, i haven’t had much luck doing this well, but maybe others can recommend methods that work for them.
~ Stacie
Hi @swolny,
Thank you for your reply.
I did burn the streams into the DEM but it made no difference. However, I do not need to map the streams for my research so maybe it is not a problem. Are the quickflow, baseflow and available recharge still correct if I use really high threshold flow accumulation values? (100000+)
Best, Livia
Hi @Livia -
SWY uses the location of streams to determine a lot of things, so having them be at least reasonably located is important.
Most of the quickflow calculations do not use stream locations, except when the pixels are directly next to a stream.
Local recharge and baseflow are very much related to stream location. To calculate recharge the model routes water downslope, infiltrating and evapotranspiring along the way until it reaches a stream, then the calculations stop.
For more information about these calculations, see the SWY The Model section of the User Guide, particularly Figure 2 for a visual of how routing is related to several of the outputs.
~ Stacie
