usgs-r/drb-network-prep

Code repo to prepare the river network for modeling in the Delaware River Basin

drb-network-prep's Issues

Omit divergences from crosswalk table for reaches within `drb_segs_spatial`

Even if omit_divergences = TRUE, some reaches in p2_drb_comids_all_tribs represent divergent reaches (where STREAMORDE != STREAMCALC). It appears that these reaches are associated with PRMS/NHM segments included in drb_segs_spatial. Add a step to get_intersecting_nhdplus_catchments() to exclude these reaches when building the crosswalk table if omit_divergences = TRUE.

> tar_load(p1_nhdv2reaches_sf)
> tar_load(p2_drb_comids_all_tribs)
> p2_drb_comids_all_tribs %>% 
+     mutate(COMID = as.integer(COMID)) %>% 
+     left_join(sf::st_drop_geometry(p1_nhdv2reaches_sf[,c("COMID","STREAMORDE","STREAMCALC")]), by = "COMID") %>% 
+     filter(STREAMORDE != STREAMCALC) %>% pull(PRMS_segid) %>% 
+     unique()
 [1] "1256_1" "2137_1" "2138_1" "233_1"  "249_1"  "2757_1" "2758_1" "2765_1" "2766_1" "2767_1" "2797_1" "354_1"  "355_1"  "396_1"  "593_1" 
>

Add a target for the NHDPlusV2 catchment polygons

Create a gpkg file containing the NHDPlusV2 catchment polygons for the DRB. Related to #11.

hru's without hru_ids in GFV1_catchments_edited

Jake mentioned the need for universal ids in hru processing tasks (USGS-R/drb_gridmet_tools#11 (comment)). I was looking into this for the gridmet data. In the GFv1_catchments_edited.gpkg file, there is an hru_id column which seems like would serve the purpose of more "universal" identifier. A couple questions I have about this:

1. Is this hru_id_nat?

2. I noticed that for 6 of the 867 hru's, there is no hru_id. These are for the three segments that have been split (3_1/3_2, 51_1/51_2, 8_1/8_2). I'm wondering if we should still assign the original hru_id to the *_1 segments and maybe to the *_2 segments too.

Look into using nhdplusTools helpers instead of traverse_nhd()

I'm wondering if the code to build the NHM - NHDPlusv2 crosswalk table could be made simpler and thus more readable by replacing traverse_nhd() with some helpers from nhdplusTools. Due to idiosyncrasies in the NHM v1.0 network this may not necessarily be a better solution to what we've got, but probably worth considering.

460 nhd flowlines do not have catchments in DRB

I was looking at the nhd catchments/flowlines in the DRB. Since these will likely be added to this repo, I wanted to make a note of something I found.

Of the 12970 nhd flowlines in the DRB, 460 do not have an associated catchment.

These 460 are all very small. Some cross over catchment boundaries.

Here are a couple samples (the red ones do not have catchments)

Here'e an overview of the lengths of the segs w/ and w/o catchments:

See here for code: https://gist.github.com/jsadler2/504067b814c2d2601e1241ddf3ad7a16

Omit special handling steps when identifying contributing catchments

Omit special handling steps when identifying comid_cat for the subset of segments defined in drb_segs_spatial. See #15 for further details.

This means omitting:

• drb_segs_spatial from _targets.R
• get_intersecting_nhdplus_catchments() in 2_process/pair_nhd_catchments.R
• pair_nhd_catchments in 2_process/pair_nhd_catchments.R

add hru_id_nat to GFV1_catchments_edited

To make GFV1_catchments_edited more useful, we should add the hru_id_nat in addition to the hru_id

Check crosswalk table for zero-area COMIDs

There are PRMS segments where comid_down, comid_seg, or comid_cat are zero-area COMIDs (i.e., with AREASQKM = 0). From the drb-inland-salinity-ml, it would be ideal if all COMIDs had attributes. Check these instances of zero-area COMIDs and if possible, omit these from the crosswalk table.

Scope out differences between NHGFv1 and NHGFv2

In FY22, PUMP-ML projects used NHGFv1 to represent the river network fabric. Now that NHGFv2 has been released, how do these datasets compare in the DRB?

Add network adjacency matrix based on NHDv2 network

Add a function/target that computes the network adjacency matrix for the DRB NHDPlusV2 network, likely borrowing heavily from the calc_dist_matrices function in delaware-model-prep.

Missing NHD catchments as a result of omitting divergences

When I started working on #17, I noticed that using p2_drb_comids_all_tribs results in some notable "holes" in the network, i.e. places where we don't identify that COMID as one for which to grab the catchment polygon. Some of these result because of the way we use special handling to identify comid_cat's for reaches within drb_segs_spatial. But it also seems that divergent reaches have associated catchments. I don't think I appreciated that by omitting divergent reaches in the crosswalk table we'd also be impacting the catchments over which we may want to calculate HRU-scale attributes (e.g. as in the inland salinity project).

Here's an example, where the white "holes" are catchments associated with divergences.

Not all fixed HRU geometries are valid

Not all the geometries within p2_GFv1_catchments_edited_sf are valid (e.g. all(st_is_valid(p1_catchments_edited_sf)) returns FALSE). See this PR in the inland salinity repo for further details.

Add seg_id_nat to crosswalk table

Retain attribute seg_id_nat (in addition to regional segment/DRB segment identifiers) when writing crosswalk table.

Some unexpected COMIDs being returned in comid_cat for segments within `drb_segs_spatial`

I noticed some unexpected COMIDs are included in comid_cat for at least one of the segments within drb_segs_spatial. We currently identify these COMIDs using the NHD catchments that overlap the original PRMS HRU, so these extraneous COMIDs are being returned from the lines below in get_intersecting_nhdplus_catchments:

    # Download flowlines and VAA for COMIDs in nhd_cats_intrsct
    nhd_flines_intrsct <- nhdplusTools::get_nhdplus(comid = c(nhd_cats_intrsct$featureid),
                                                    realization = 'flowline') %>%
      # reformat variable names to uppercase
      rename_with(.,toupper,id:enabled)

Here's an example for PRMS subsegid 396_1: