NHD Surface Water & Section 404 Screener

Give it a list of site coordinates. Get back, per site, the answer that drives a Clean Water Act call: how far is the nearest perennial stream, versus the nearest intermittent one, versus the nearest ephemeral wash — plus the mean annual flow on that reach, the nearest lake, reservoir or swamp/marsh, the stream order and drainage area of the receiving water, the HUC8/10/12 the site drains through, and a HIGH / MEDIUM / LOW Section 404 review flag that tells you which feature drove it and why.

Data comes from the USGS NHDPlus HR MapServer (hydro.nationalmap.gov/arcgis/rest/services/NHDPlus_HR/MapServer) — keyless, public-domain, no login, no scraping of a rendered page.

---

Why the perennial / intermittent / ephemeral split matters

Most surface-water tools dump every NHD flowline within a buffer and leave you to sort it out. After Sackett v. EPA (2023) the jurisdictional question turns on relatively permanent flow with a continuous surface connection, so this actor does the sorting — and reports the evidence.

FCode	Regime	Raises the 404 flag?
46006	Perennial stream	Yes
46003	Intermittent stream	Yes
46000	Stream/river, unspecified	Yes
55800	Artificial path (mapped centerline of a river or lake)	Yes
46007	Ephemeral wash	No
33600 / 33601 / 33603	Canal / ditch / aqueduct / stormwater	No
42800–42824	Pipeline	No
42000–42003	Underground conduit	No
46800	Drainageway	No

Polygons are treated the same way. Lakes, ponds, reservoirs, swamp/marsh, estuaries, bays, foreshore, rapids and submerged streams raise the flag. Washes (FType 484), playas, ice masses, ditch polygons and engineered structures — levees, bridges, dams, spillways, lock chambers, flumes, water intake/outflow, mapped inundation areas — do not. They are still reported, in their own columns, so you can see what was excluded and why.

Flag bands: HIGH = a jurisdictional feature within 100 ft (30.48 m). MEDIUM = within 300 ft (91.44 m). LOW = everything else.

A site genuinely far from mapped water is a legitimate result: you get a row with null distances, LOW, and jurisdictional_likelihood: UNLIKELY — not an error and not a missing row.

The flag can also be null, and that is the point

section_404_review_flag is null whenever a contributing NHD layer did not answer for that site. A LOW or MEDIUM verdict asserts that nothing jurisdictional lies within 100 ft; if the waterbody layer never responded, that negative was never verified, and shipping it as LOW would be a confident wrong answer on a determination that carries liability. A HIGH is different — a positive finding cannot be reversed by missing data — so a HIGH stands and is marked positive_despite_partial_coverage.

Every row carries screening_status, layers_ok, layers_unavailable, a layer_*_status per NHD layer, and section_404_flag_reason in plain English. Filter on screening_status = "complete" before acting on any negative finding. If the load-bearing flowline layer is down for every site, the run fails rather than emitting rows — nothing is emitted and nothing is billed.

---

Who buys this

• Phase I ESA consultants — the surface-water and adjacent-waters section of an ASTM E1527-21 report, with a citable reach code, the NHD feature date, and a date-of-source stamp for every row.
• Wetland delineators — nearest_swamp_marsh_dist_m is NHD's wetland proxy and the strongest single signal that a formal delineation is in scope; the per-regime feature census (perennial_features_in_radius and friends) is what you price the field effort from.
• §404 / Clean Water Act permitting teams — first-pass triage of which parcels need a delineation and an approved jurisdictional determination before a Nationwide vs Individual Permit path is chosen. mean_annual_flow_cfs is the number a relative-permanence argument is actually built on.
• Solar, BESS and data-center siting teams — rank hundreds of candidate parcels by 404 exposure before spending money on field work. The wash and ditch exclusions matter enormously in the arid south-west, where washes are the most common mapped NHDArea feature.
• Land developers and title / diligence shops — flag creek-adjacent parcels across an acquisition pipeline in one run.

---

Example input

{
  "assets": [
    { "lat": 40.0143,  "lon": -105.2829,  "label": "Boulder CO - creekside redevelopment parcel" },
    { "lat": 29.99091, "lon": -89.93323,  "label": "New Orleans East LA - swamp/marsh adjacent site" },
    { "lat": 33.42931, "lon": -111.98414, "label": "Tempe AZ - parcel inside a mapped desert wash" },
    { "lat": 35.2,     "lon": -115.9,     "label": "Mojave NP CA - upland solar reference site" }
  ],
  "radiusMeters": 1000,
  "maxResults": 100,
  "includeNonNetworkFlowlines": true
}

assets is required — the run is rejected before it starts if it is missing, not an array, or empty, so nothing is consumed. lat/lon are decimal degrees (latitude/longitude and lng are accepted as aliases). radiusMeters is clamped to 50–8000, default 1600.

Those four prefilled sites deliberately cover the range of outcomes: a creekside parcel 5.2 m from a perennial reach carrying 102.7 cfs (HIGH, LIKELY); a Louisiana site sitting inside an NHD swamp/marsh (HIGH, wetland-driven); an Arizona parcel sitting inside a 2.54 km² mapped desert wash, which comes back LOW because washes are reported but are not jurisdictional after Sackett (the nearest actually-jurisdictional feature is the Salt River at 101 m, just outside the 300 ft band); and an upland Mojave parcel whose only nearby feature is an ephemeral reach 258 m away (LOW, UNLIKELY).

Example output row (trimmed)

{
  "asset_label": "Boulder CO - creekside redevelopment parcel",
  "lat": 40.0143,
  "lon": -105.2829,
  "search_radius_m": 1000,

  "section_404_review_flag": "HIGH",
  "section_404_flag_status": "complete",
  "jurisdictional_likelihood": "LIKELY",
  "jurisdiction_basis": "A perennial reach lies 5.2 m (17.2 ft) from the site, inside the 100 ft band. NHDPlus reports a mean annual flow of 102.69 cfs on that reach, which supports a relatively permanent flow characterisation under Sackett v. EPA. A canal or ditch lies 364.4 m away and was excluded from the flag.",

  "nearest_jurisdictional_source": "flowline",
  "nearest_jurisdictional_name": "Boulder Creek",
  "nearest_jurisdictional_dist_m": 5.2,
  "nearest_jurisdictional_dist_ft": 17.2,

  "nearest_flowline_name": "Boulder Creek",
  "flow_regime": "perennial",
  "nearest_flowline_fcode_label": "Stream/River: Hydrographic Category = Perennial",
  "nearest_flowline_dist_m": 5.2,
  "mean_annual_flow_cfs": 102.689,
  "mean_annual_velocity_fps": 1.6366,
  "stream_order": 6,
  "drainage_area_sqkm": 336.052,
  "total_upstream_length_km": 705.632,
  "reachcode": "10190005000085",
  "nearest_flowline_data_date": "2012-02-18",

  "nearest_perennial_dist_m": 5.2,
  "nearest_intermittent_dist_m": 412.6,
  "nearest_ephemeral_dist_m": null,
  "perennial_features_in_radius": 4,
  "intermittent_features_in_radius": 2,
  "canal_ditch_features_in_radius": 6,

  "nearest_waterbody_name": "Varsity Lake",
  "nearest_waterbody_type": "Lake/Pond (perennial, closed lake)",
  "nearest_waterbody_dist_m": 839.6,
  "nearest_waterbody_is_jurisdictional": true,
  "nearest_swamp_marsh_dist_m": null,

  "huc8": "10190005",
  "huc10": "1019000504",
  "huc12": "101900050406",
  "huc12_name": "City of Boulder-Boulder Creek",
  "huc12_type": "Standard",
  "tohuc": "101900050601",

  "screening_status": "complete",
  "layers_ok": 5,
  "layers_unavailable": null,
  "distance_method": "exact_geometry",
  "error": null,
  "source_url": "https://hydro.nationalmap.gov/.../MapServer/3/query?...",
  "retrieved_at": "2026-07-28T18:24:57.974Z"
}

---

Output fields

All 94 fields are listed below — every field this actor declares, named in full. Distances are reported in metres and in feet; the _ft column is the exact same measurement, converted, so a permit form and a US survey can both be filled straight from the row.

Your site, echoed back

Field	Meaning
asset_label	Your own label for the site, echoed verbatim so results join back to your parcel list
lat	Site latitude in decimal degrees (WGS84/NAD83); null on a row whose input coordinate was unusable
lon	Site longitude in decimal degrees (negative in the USA)
search_radius_m	Radius in metres actually searched around the site for flowlines, waterbodies and water areas

Verdict and audit trail

Field	Meaning
section_404_review_flag	HIGH / MEDIUM / LOW, or null when a contributing layer did not answer
section_404_flag_status	complete, positive_despite_partial_coverage, indeterminate_partial_coverage, indeterminate_probe_failure, not_screened
section_404_flag_reason	Why the flag was suppressed or qualified; null on a clean row
jurisdictional_likelihood	LIKELY / POSSIBLE / UNLIKELY / INDETERMINATE — a screening opinion, not a determination
jurisdiction_basis	Full reasoning: which feature, at what distance, with what flow, and what was excluded
nearest_jurisdictional_dist_m	The distance the flag bands were actually applied to, in metres
nearest_jurisdictional_dist_ft	The same distance in feet
nearest_jurisdictional_source	Whether a flowline or a polygon drove the flag
nearest_jurisdictional_name	Name of the feature that drove the flag
nearest_jurisdictional_type	Type of the feature that drove the flag

Nearest flowline

Field	Meaning
nearest_flowline_name	Nearest reach of any type; falls back to reach code — never blank
nearest_flowline_ftype	Decoded NHD FType of that reach
nearest_flowline_fcode	Raw numeric NHD FCode
nearest_flowline_fcode_label	The verbatim USGS FCode description, straight from the layer's coded-value domain
flow_regime	perennial / intermittent / ephemeral / artificial_path / canal_ditch / pipeline / underground_conduit / …
nearest_flowline_is_jurisdictional	Whether that reach is of a type that can raise the flag
nearest_flowline_dist_m	Distance to that reach in metres
nearest_flowline_dist_ft	Distance to that reach in feet
nearest_flowline_layer	3 (network, carries NHDPlus VAA) or 4 (non-network, no flow/order/drainage data exists)
mean_annual_flow_cfs	NHDPlus qama — the relative-permanence proxy
mean_annual_velocity_fps	NHDPlus vama
reach_slope	Smoothed NHDPlus slope for the reach
stream_order	Strahler stream order
drainage_area_sqkm	Total upstream drainage area
total_upstream_length_km	Arbolate sum — total upstream network length
is_headwater	Network start — nothing upstream
reachcode	NHD reach code, to cite in a permit or delineation
nearest_flowline_gnis_id	GNIS identifier of the named reach
nearest_flowline_length_km	Length of the mapped reach
nearest_flowline_data_date	Vintage of the mapped feature
nhd_resolution	Source resolution of the mapped feature

Flow-regime distances and census

Field	Meaning
nearest_perennial_dist_m / nearest_perennial_dist_ft	Nearest FCode 46006, metres and feet
nearest_intermittent_dist_m / nearest_intermittent_dist_ft	Nearest FCode 46003, metres and feet
nearest_ephemeral_dist_m / nearest_ephemeral_dist_ft	Nearest FCode 46007 — reported, never flagged
nearest_canal_ditch_dist_m	Nearest ditch or canal — reported, never flagged
perennial_features_in_radius	Perennial reaches within the radius
intermittent_features_in_radius	Intermittent reaches within the radius
ephemeral_features_in_radius	Ephemeral reaches within the radius
canal_ditch_features_in_radius	Canals and ditches within the radius
pipeline_conduit_features_in_radius	Pipelines and underground conduits within the radius
flowline_features_in_radius	All flowlines within the radius
waterbody_features_in_radius	All waterbody polygons within the radius
surface_water_features_in_radius	Every surface-water feature within the radius — the density signal a delineation scope is priced from

Waterbodies and wetlands

Field	Meaning
nearest_waterbody_name	Nearest polygon of any type — which on an arid site is often a wash
nearest_waterbody_type	Decoded type of that polygon
nearest_waterbody_dist_m / nearest_waterbody_dist_ft	Distance to it, metres and feet
nearest_waterbody_area_sqkm	Mapped area of that polygon
nearest_waterbody_fcode	Raw numeric NHD FCode of the polygon
nearest_waterbody_fcode_label	Verbatim USGS FCode description for the polygon
nearest_waterbody_purpose	NHDPlus purpose code (tailings, settling, sewage treatment, cooling, evaporator, disposal, …)
nearest_waterbody_reachcode	Reach code of the polygon, where NHD carries one
nearest_waterbody_data_date	Vintage of the polygon
nearest_waterbody_layer	Which source layer the polygon came from (NHDWaterbody or NHDArea)
nearest_waterbody_is_jurisdictional	Whether that polygon can raise the flag
site_inside_waterbody	The site coordinate falls inside the polygon
nearest_jurisdictional_waterbody_name	Nearest polygon that can raise the flag
nearest_jurisdictional_waterbody_type	Its decoded type
nearest_jurisdictional_waterbody_dist_m	Distance to it in metres
nearest_swamp_marsh_dist_m	Nearest NHD Swamp/Marsh (FType 466) — the wetland-delineation trigger
nearest_swamp_marsh_name	Name of that swamp or marsh, falling back to its NHD identifier
nearest_lake_pond_dist_m	Nearest lake or pond
nearest_reservoir_dist_m	Nearest reservoir
nearest_reservoir_type	Reservoir type: tailings pond, settling pond, sewage treatment pond, cooling pond, evaporator, disposal and so on
nearest_wash_playa_dist_m	Nearest wash or playa — reported for transparency, deliberately excluded from the flag

Watershed

Field	Meaning
huc8	Subbasin — what most permit forms ask for
huc10	Watershed
huc12	Subwatershed
huc12_name	Subwatershed name
tohuc	The HUC12 this subwatershed drains into
tohuc_is_terminal	Whether the outlet is terminal (OCEAN, CANADA, MEXICO)
huc12_type	Standard, Closed Basin, Frontal, Multiple Outlet, Water, Island, Urban, Indeterminate
huc12_area_sqkm	Subwatershed area
huc12_noncontributing_area_sqkm	The part of it that does not drain to its outlet

Coverage and QA

Field	Meaning
screening_status	complete / partial / failed
layer_network_flowline_status	ok / unavailable / not_requested for NetworkNHDFlowline
layer_nonnetwork_flowline_status	Same, for NonNetworkNHDFlowline
layer_waterbody_status	Same, for NHDWaterbody
layer_nhdarea_status	Same, for NHDArea
layer_wbdhu12_status	Same, for WBDHU12
layers_ok	How many of the five layers answered
layers_unavailable	Which ones did not
distance_method	exact_geometry, ring_probe_bound, no_features_in_radius, unavailable
unmapped_codes	Any NHD code USGS returned that this actor does not decode — surfaced, never dropped
error	Null on a clean row; reason or PARTIAL: notice otherwise
source_url	The exact query used
retrieved_at	UTC run timestamp

---

How the distances are computed

Exact point-to-polyline and point-to-polygon distance from the geometry USGS returns, in WGS84, with an equirectangular projection anchored at your site (sub-metre accurate over kilometres). A site inside a polygon is distance 0.

If NHD ever returns features without geometry, no exact distance exists. The actor then falls back to attribute-only ring probes at exactly 30.48 m and 91.44 m to decide the band honestly, sets distance_method to ring_probe_bound, and leaves every *_dist_m field null — because a probe gives you a bound, not a distance, and reporting a bound as a distance would be a fabricated number.

The projection trap, and the live assertion that guards it

NHDPlus HR answers in Web Mercator (wkid 102100) by default. Without an explicit outSR=4326 every coordinate comes back as metres-from-Greenwich, and every distance derived from it is wrong by orders of magnitude while looking entirely plausible. An offline test cannot catch that, because the fixture would have been shaped from the bad data.

So every geometry-bearing response is checked at runtime, on three independent signals. If any of them fails the run fails loudly rather than emitting distances:

1. the declared spatial reference is 4326 — the real signal, and the only one of the three that also catches a near-identical datum swap such as NAD83 / 4269, where every coordinate still looks correct;
2. every vertex is a possible WGS84 coordinate by magnitude (longitude within ±180, latitude within ±90). Web Mercator is metres, so an x of ≈ −11,721,894 fails this instantly and unambiguously;
3. each returned feature's own bounding envelope is within a metre allowance (3× the search radius, 5 km floor) of the queried point — which catches a response describing the wrong place.

Coastal note (fixed 2026-07-28, build 1.1.7). Signal 3 used to be "every returned vertex must be within about a degree of the site", and that is simply false on an open coast. NHDArea layer 8 models the Gulf of Mexico as a single 364,494-vertex polygon spanning lon −89.861 … −88.013 / lat 29.250 … 30.433 (10,021 km²), so a perfectly valid WGS84 response at a Biloxi MS waterfront parcel contained the real vertex [−89, 29.25] — 130 km south of the site — and the run hard-failed with a false "projection drift". Every open-coast, bay, sound and large-lake site was unscreenable. A large polygon legitimately has distant vertices; what it may not do is have its entire extent sit far from the point the buffer was drawn around, which is what the envelope test now asserts. The Mercator case is caught by magnitude, where it always belonged.

Four more live checks run once, before any site is screened, and fail the run if they trip:

1. every layer id still maps to its expected layer name (a USGS renumbering would otherwise make the actor query FlowDirection and call it "flowlines");
2. the flowline FCode subtype domain still carries 46003 / 46006 / 46007, the codes the whole perennial / intermittent / ephemeral split rests on;
3. a fixed canary query at Boulder Creek still returns features, in 4326, near the canary point, with NHDPlus value-added attributes attached — so a silent loss of qama / streamorde / totdasqkm surfaces as a red run instead of a column that quietly renders null on every row;
4. NHDArea FType 484 (Wash) still exists, since the non-jurisdictional wash handling depends on it.

---

Use as an MCP tool

This actor is available over the Apify MCP server, so an AI agent (Claude Desktop, Claude Code, Cursor, or any MCP client) can call it directly. Point your MCP client at https://mcp.apify.com and expose the actor malonestar/nhd-surface-water-404-screener. A prompt like "screen these twelve parcel coordinates for Section 404 exposure and tell me which need a wetland delineation" becomes one tool call returning structured rows. Field-level descriptions in the output schema are written for an agent to read, and jurisdiction_basis gives the agent the reasoning to quote back.

---

Reliability

• Every request has a 45 s AbortController timeout and retries with exponential backoff on 408/425/429/5xx.
• Missing or empty assets is rejected by the input schema before the run starts, so nothing is consumed.
• Per-layer failures are recorded on the row, not just in the log, and a negative verdict is suppressed rather than asserted over missing data.
• If some assets resolve and others fail, the failures come back as rows carrying an error, a null flag and screening_status: "failed" — good rows are never discarded, and a failed site never reads as clear.
• The run fails outright when nothing can be answered at all, and then nothing is emitted and nothing is billed.
• The 2000-record page cap is handled by following exceededTransferLimit with resultOffset pagination.

---

FAQ

How do I find the nearest perennial stream to a property? Pass the property's latitude and longitude in assets. The row's nearest_perennial_dist_m is the distance in metres to the nearest NHD FCode 46006 reach, and nearest_flowline_name / reachcode identify it.

What is the difference between perennial, intermittent and ephemeral streams in NHD? NHD encodes it in the FCode: 46006 perennial (flows year-round), 46003 intermittent (flows seasonally), 46007 ephemeral (flows only in direct response to precipitation). This actor returns the distance to the nearest of each, and only the first two can raise the review flag.

Do I need a Section 404 permit for my project site? This actor gives you the screening inputs, not the legal determination. A HIGH flag means a relatively permanent water is within 100 ft and a wetland delineation plus an approved jurisdictional determination is the normal next step. Only the US Army Corps of Engineers can make the actual call.

Are ephemeral streams jurisdictional after Sackett v. EPA? Generally no — which is exactly why this actor reports ephemeral distances separately and does not let them raise the review flag. The same applies to ditches, washes and playas.

Why did a desert site come back LOW when it is sitting in a wash? Because a wash (NHD FType 484) is an ephemeral feature and is generally not a water of the United States after Sackett. The wash is still reported — see nearest_wash_playa_dist_m, nearest_waterbody_type and nearest_waterbody_is_jurisdictional — it just does not raise the flag.

What does mean_annual_flow_cfs tell me? It is NHDPlus's modelled mean annual flow for that reach in cubic feet per second, and it is the best openly available proxy for the relatively permanent flow test. A reach carrying 100 cfs supports a very different argument from one carrying 0.07 cfs. It is null when the nearest reach came from the non-network layer, which does not carry NHDPlus attributes at all.

Why is section_404_review_flag sometimes null? Because one of the NHD layers did not answer for that site, so a LOW or MEDIUM would be asserting a negative that was never verified. Check layers_unavailable and section_404_flag_reason, and re-run when the USGS service is healthy.

Which HUC12 watershed is my site in, and where does it drain? huc12 and huc12_name give the containing subwatershed from the Watershed Boundary Dataset; tohuc gives the HUC12 immediately downstream, and tohuc_is_terminal tells you when the outlet leaves the HUC network entirely. huc8 and huc10 are derived for the permit forms that ask for them.

How do I batch-screen a portfolio of parcels for wetland risk? Put all of them in assets in one run — one row comes back per parcel, including parcels with no nearby water. Sort by section_404_review_flag and nearest_swamp_marsh_dist_m, and filter to screening_status = "complete".

Is the USGS NHDPlus HR data free to use? Yes. It is US Government public-domain data served without an API key. This actor charges for the compute and the transform, and every row carries the exact source_url used so any number can be re-derived.

---

Pricing

Pay per result: $12 per 1,000 results. One result = one screened asset row. Rows emitted for assets that failed validation or could not be screened carry an error field and a null flag, so you can see exactly what was charged for. If nothing could be screened, the run fails and no result events are charged.

---

Related actors

Section 404 exposure is one layer of a site screen. These share the same [{lat, lon, label}] input shape and join on asset_label:

Actor	What it adds
fws-wetlands-proximity-screener	USFWS National Wetlands Inventory polygons — the authoritative wetland mapping. NHD's Swamp/Marsh is a coarse proxy; NWI is what a delineation is scoped against. Run both.
epa-impaired-waters-303d-screener	Whether the receiving water you just identified is CWA §303(d) impaired, and for which pollutant — which drives TMDL and discharge-permit conditions.
epa-contaminated-site-screener	Superfund, RCRA, USTs, NPDES and enforcement history for the ASTM E1527-21 recognised-environmental-conditions section of the same Phase I.
site-due-diligence-bundle	All twenty diligence layers for one coordinate in one call — contamination, wetlands, flood, seismic, habitat, historic, levee, karst and more — with a composite site score and a go / caution / no-go verdict.