ArcHydro Adjoint Catchments Issue

[PeytonLisenby]

Hello,

I have seen several posts on a few different forums about the 'Adjoint Catchments' tool in ArcHydro running and completing successfully, but without creating any polygons, or fewer polygons than it should have created, with blank attribute tables and/or zero shape areas. As far as I can tell, there is no solution to this problem. There are only a few hints and tips that go toward hoping the problem goes away. I am posting this question here in order to hopefully add some specificity to this issue, or at least determine if the issues I am currently having are truly a lost cause. If this question has already been answered somewhere, I apologize for the duplication, but I would like to be directed to the solution.

I have a somewhat large data set (a 1m resolution, ~3000 km^2 catchment DEM) and I am attempting to run the 'adjoint catchments' tool. I have had no trouble in the preceding steps of running the 'fill sinks', 'flow direction', 'flow accumulation', 'stream definition', 'stream segmentation', 'catchment grid delineation', 'catchment polygon processing', and 'drainage line processing'. I currently have a catchment polygon file and a drainage line file with almost 900,000 attributes, each, and this is based on a 1km^2 river threshold used for the stream definition tool.

I have run the 'adjoint catchments' tool several times now with varying results. It always says it finishes successfully; however, it either creates a blank 'adjoint catchments' feature class, or one that has anywhere from 200 to 3000 polygons where some have zero shape area. Sometimes the tool only creates polygons for the northern-most catchments of my DEM (I once got it to complete ~ 70% of the catchments...until ArcGIS crashed). Invariably, the tool progress shows that in 'Pass 1' the tool evaluates about 340,000 of the 900,000 catchments; however, the Done:NotDone ratio reads as ~ 340,000:140,000.

I have addressed some usual suspects for this tool failure - turned background geoprocessing off, updated to the latest version of ArcHydro, cleared my temporary files, set sensible target locations, made sure my processing extent is correct. I have tried working within a save .mxd and an unsaved, untitled .mxd, along with setting different workspace processing locations. I am running ArcGIS 10.2.0 with ArcHydro ver. 10.2.0.197. I have plenty of RAM (16 GB, even though ArcGIS can only manage to use 4 GB), and plenty of free disk space.

The only way I have found around this issue is to go back to the 'Stream definition' tool and set a higher river threshold. I changed mine from 1 km^2 to 10 km^2. I then repeated the intermediary steps and ran the 'Adjoint catchments' tool again...and it worked fine (this time there were only ~ 44,000 catchments to process). The problem is that I need to maintain a 1 km^2 river threshold for my particular project. Does the 'Adjoint Catchments' tool have a processing limit on the number of polygons it can handle? Can I not work with more than 100,000 catchment polygons? Is there away around this apparent size limit?

Apologies for the long question, but I hope there is enough information here to accurately portray the issue at hand.

Thanks! 

[DanPatterson_Retired]

I am sure you have seen the history and the similar conclusions you have arrived at (last thread in this post)

a problem about using "adjoint catchment processing"

Perhaps there is a fundamental flaw in the processing design, if it was a memory issue, then you should get some warning or error.  Have you addressed this with the archydro team?

[PeytonLisenby]

Hi Dan,

Sorry for the delayed response. I have seen that previous thread.

I haven't addressed this issue with the ArcHydro team as of yet simply because I am not sure how best to do it.

As a side note, I read on another forum (and I can't remember where) that since the 'Adjoint Catchment' tool is used for calculating distance upstream, it doesn't matter what river threshold is used to ultimately create the 'adjoint catchment' feature class. Theoretically, one could create everything up to the 'catchments' feature class with a small river threshold and then separately create an 'adjoint catchment' feature class using a larger river threshold. Subsequently, this different 'adjoint catchment' (larger river threshold) feature class could be used in conjunction with the other grids and feature classes (with the smaller river threshold) to create drainage points, batch points, and delineate individual watersheds. I don't know if this is accurate, so I'll give it a whirl.

Cheers,

P

[Siri_JodhaKhalsa]

Exactly the same problem here, empty Adjoint Catchment feature class while Catchment is fine. Large domain, small stream initiation threshold.  I don't really need the Adjoint Catchment since I'm only interested in a particular set of pour point, but the batch watershed processing uses the Adjoint Catchment feature class as an input. Most of the basins defined using a set of 24 batch points look fine but some have clipped, linear borders.  Wondering whether the empty Adjoint Catchment feature class could be the cause.  But, who knows, the ArcHydro Tools have so many bugs. Anyway, starting over with smaller stream threshold yields Adjoint Catchments, but much bigger than I need.

[MarkBoucher]

Clipped, linear borders can be a sign that the processing extent was constrained to some domain (sometimes that of the BatchPoints dataset). 

Mark

[HannesZiegler2]

Not a solution to the problem, but your comment about the 'usual suspects' helped me resolve my own issues with running the Adjoint Catchment Processing tool, so I just wanted to say thank you!

[MarkBoucher]

The "usual suspects" are also discussed here: https://community.esri.com/message/45017#45017 

[BrendenMclane]

I've also had this issue with ArcHydro 10.5.0.104 and ArcGIS 10.5.1. Everything ran fine but the adjoint output was empty. I took the above advice and gradually increased my stream threshold from a low of 500 pixels (at 9ft resolution) to 3000 pixels, or 0.022576 sq km, until my adjointment catchments finally got populated. This resulted in about 190,000 catchments and around 90,000 adjoint catchments in my area of interest. In any case, the batch delineation tool now works without any of the strange linear cutting off the output basin. I am disappointed that ArcHydro can't seem to tolerate more catchments than that, but since the delineation tool will cut out the downstream portions of my last catchment, this still gets me what I need in terms of basin delineation.

Thanks for the advice!

[MarkBoucher]

Please, someone correct me if I'm wrong ...

... I believe Batch Point delineation should work no matter what size the stream threshold is. I believe the more Adjoint Catchments you have, the faster the batch processing will go. This is because the batch delineation processes only have to delineate the areas upstream of the Batch Points to the upstream edge of the catchment polygon they are in. The the rest of the delineation happens very quickly by simply selecting the next upstream adjointcatchment. So Adjoint Catchment process is like prepossessing the delineation to make the batch processing faster. It is also used in other tools, so you need to do it for those. 

So if you have a huge watershed and a large stream threshold (resulting in fewer adjointcatchment polygons), the batch delineation processes should take longer than if you have a smaller stream threshold (resulting in more adjointcatchment polygons). Until this issue gets addressed, increase your stream threshold (reduce the number of adjointcatchment polygons) until it eventually works.

I recall a story that when this was first demonstrated live at a conference, many of the attendees thought it was a joke, a trick, or not real. They couldn't believe it. They were used to delineation taking forever. Using the network analysis capabilities (necessitating HydroID, UpstreamID, NextDownID, etc), Arc Hydro is able to do these tasks relatively quickly.

Mark