• MNDWI = (Green−SWIR)/(Green+SWIR), water where MNDWI>0
• Water-occurrence over clear observations (JRC construct)

Reliable (open water); corroborated by the Sentinel-1 SAR signal (cloud-robust)

• TWI = ln(a / tanβ), a = specific catchment area from D8 flow accumulation, β = local slope
• % of parcel above the saturation-prone threshold (TWI > 15)

• Relative wetness index
• Sensitive to DEM resolution
• A coarse DEM smooths micro-topography, so treat as a screening cue, not a drainage design

USFWS National Wetlands Inventory wetland polygon at the point (ArcGIS REST)

• US inventory flag only
• A jurisdictional wetland delineation requires a professional
• NWI can miss small/forested wetlands

• JRC Global Surface Water `seasonality` layer: per-pixel count of months (1-12) with open water in the layer year
• Parcel stat = % of pixels with water ≥1 month/yr (max months and permanent 12-month share reported in detail)

• 30 m Landsat-derived (archive 1984-2020
• Seasonality layer = calendar 2020). Detects OPEN water only
• Misses water under canopy and channels narrower than ~30 m

• Current US Drought Monitor category for the parcel's county (dominant D0–D4 class by exclusive area share
• County FIPS via the FCC census-block API)

• Weekly national product at county scale
• Current conditions, not a long-term water-supply determination

• Slope from Copernicus GLO-30 DEM (Horn 1981 gradient)
• Buildable = % of parcel under 8% slope

• Slope Reliable
• 'buildable' cut-off Indicative (planning convention). GLO-30 smooths <30-90m → use 3DEP LiDAR for US grading-scale

Local relief = max−min elevation, plus median absolute elevation, over the parcel from the GLO-30 DEM

~4 m vertical accuracy; surface (DSM) not bare-earth — prefer 3DEP LiDAR in the US

LandBenchmark Ruggedness Index (Riley et al. 1999): TRI = sqrt(Σ(z_neighbour − z_centre)²) over the 8-cell neighbourhood, mean over parcel

Screening index; correlates with slope. VRM (Sappington 2007) decorrelates if needed

• Slope from the DEM (Horn 1981). The parcel polygon is rasterised onto the DEM grid
• Cells are counted usable when slope is under the threshold AND they fall outside the FEMA Special Flood Hazard Area (NFHL layer 28, SFHA_TF='T') and outside mapped USFWS National Wetlands Inventory polygons. Usable acres = usable cells × cell ground area. Slope cutoffs are planning conventions (5/8/12/15%), reported as a curve rather than a single figure.

Screening estimate, not a site plan. The DEM is a gridded surface model (indicative, not survey-grade), NWI mapping is not a jurisdictional wetland determination, and FEMA zones are regulatory maps rather than observed flooding. Setbacks, easements, and local grading rules are not subtracted. Confirm with a licensed surveyor before relying on acreage.

• RUSLE topographic factor LS = (aₛ/22.13)^0.4·(sinβ/0.0896)^1.3 (Moore & Burch 1986), aₛ = specific catchment area from D8 flow accumulation, β = slope
• Relative terrain-driven erosion potential

• Topographic (LS) factor ONLY
• Rainfall (R), soil (K), cover (C) and practice (P) are not included, so this is RELATIVE terrain erosivity, not an absolute soil-loss rate
• DEM-resolution sensitive

• Topographic Position Index = elevation minus the local neighbourhood mean (Weiss 2001)
• Classed valley / mid-slope / ridge by ±1 standard deviation of TPI over the parcel

• Relative landform position at the sampled window scale
• Valley bottoms pool cold air/frost and water, ridges are wind-exposed
• A microclimate/siting cue, not a hazard determination

USDA soil texture class from SoilGrids 2.0 clay/sand/silt fractions, 0–30 cm depth-weighted mean

• Global 250 m model prediction (not a field survey)
• Heavy clay → shrink-swell & drainage issues, pure sand → low water/nutrient retention. In the US, SSURGO (see soil signals) is more authoritative

Soil pH in water from SoilGrids 2.0, 0–30 cm depth-weighted mean

• Global 250 m model
• Strongly acidic (<5) or alkaline (>8.5) soils limit crops and can affect concrete/rebar
• Confirm with a lab test

Soil organic carbon content from SoilGrids 2.0, 0–30 cm depth-weighted mean

Global 250 m model; a fertility/organic-matter proxy, not a nutrient test

Overpass nearest highway=* distance from the parcel centroid (haversine)

highway class = importance, NOT surface quality; legal/deeded access needs a title search

Distance to nearest EIA-mapped natural-gas interstate/intrastate transmission pipeline within ~3 km (ArcGIS FeatureServer distance query + local point-to-segment distance)

• Transmission-level network only
• A nearby line does NOT mean connectable gas service (that is the local distribution utility)
• Close proximity is a safety/easement consideration, and our <300 m flag is a screening cut in the spirit of PHMSA consultation zones, not a regulatory determination

USGS ASCE 7-16 seismic design values (S_DS design spectral acceleration and Seismic Design Category) at the point, Risk Category II / Site Class D

• US & territories
• Design-level ground motion for code compliance
• Not a site-specific fault or liquefaction study

USFS Wildfire Hazard Potential 5-class at the point (FSim + LANDFIRE fuels)

US only at 270 m — regional context, not a site-level fire assessment

USGS FDSN event service: count + largest + most recent M≥4.5 event within 100 km of the parcel centroid since 1975 (count endpoint + magnitude-ordered query)

• Observed catalog
• Completeness varies by region/era
• Hazard DESIGN values need the national code (US: ASCE 7 signal)

Haversine distance from the parcel centroid to the nearest Holocene volcano in the Smithsonian GVP WFS (bbox ±3°, widened to ±8° if empty)

• Proximity screen to Holocene volcanic centers
• Actual hazard footprints (ashfall/lahar) need national volcanic-hazard maps

• MODIS MCD64A1 monthly burned-area: burned pixel = Burn_Date > 0 (day-of-year of burn
• 0 = unburned, negatives = water/unmapped fill)
• Parcel stat = distinct burn months over the last ~6 years, plus % of parcel pixels ever burned

• 500 m resolution
• Reliably maps large burns but misses small, patchy or low-intensity fires
• Read as regional burn history, not a parcel-level burn determination

~10 yr of daily 2 m mean temperature and precipitation from ERA5 reanalysis (Open-Meteo Archive) at centroid → mean annual temperature + annual precipitation

~9 km reanalysis grid — regional context, not microclimate; WorldClim/station data refine

• ESA WorldCover 10 m `map` asset: dominant class % + top-3 class breakdown over the parcel from the verified class table (10 Tree cover … 100 Moss and lichen
• 0 = nodata)
• Tree-cover % reported in detail

• 10 m single-epoch class map (latest hosted year, 2021 v200)
• Mixed pixels and parcel edges misclassify
• Verify the dominant class against recent imagery

• Overpass proximity query: nearest landuse=industrial|landfill|quarry|farmland or man_made=works within ~500 m of centroid
• Nearest-nuisance distance (haversine)

• OSM completeness varies by region
• Absence ≠ confirmed clear. EPA FRS adds US industrial sites

• USGS PAD-US v4 protected-area polygons at the point and within 3 km (GAP status 1-3
• Fee, Designation and Easement categories), ArcGIS REST

• US only. GAP status 1-3 is USGS's own definition of land carrying a mandate for protection
• GAP 4 (most municipal parks) is excluded because it carries none. Proclamation boundaries are excluded
• They enclose private land. Distance is measured to a boundary generalized to ~5 m, so treat sub-10 m distances as adjacency, not survey.

USDA NASS Cropland Data Layer sampled at five points across the parcel (centre + quarter points) in 2008, 2015 and 2023. A tree or vine crop in any sample raises a legacy-residue flag, because orchards and vineyards were historically treated with lead-arsenate insecticide, whose lead and arsenic do not degrade and persist in surface soil.

• A screening indicator, never a finding of contamination
• Only a laboratory soil test can establish that. Five 30 m samples per year, so a small orchard corner can still be missed. The CDL record begins in 2008, while the lead-arsenate era ended around 1950: the period that created the risk is invisible to this dataset. Absence of an orchard here is NOT evidence of no orchard history
• Historical aerial photography and a records review are what settle that. The USDA CropScape service is intermittently unavailable
• When it does not answer, this signal is absent from the report and listed as unavailable rather than reported as 'no orchard'.

• Point-in-polygon and proximity queries against three federal designations: USFWS designated critical habitat (final), National Park Service National Register of Historic Places listings within 500 m, and EPA air-quality nonattainment areas. Each is reported as a possible trigger for environmental review
• Not as a compliance determination.

• LandBenchmark FLAGS triggers
• It does not certify compliance. Whether a project requires ESA Section 7 consultation, NHPA Section 106 review, or nonattainment New Source Review depends on the project, the federal nexus, and the permitting agency
• Ask a land-use attorney. Designated critical habitat does not by itself prohibit private activity without a federal nexus. The coastal zone is NOT screened: it is defined state by state and no free national service publishes it.

Landsat Collection-2 Level-2 surface reflectance, 1985–present, sampled in 5-year epochs. Each epoch uses the 3 least-cloudy scenes inside the same peak-canopy window, so epochs are seasonally comparable. Cloud, cloud-shadow, cirrus, snow and fill pixels are masked per-pixel from the Landsat QA band. Per epoch we take the per-pixel maximum NDVI = (NIR−Red)/(NIR+Red) (peak canopy) and the fraction of the parcel whose median MNDWI = (Green−SWIR)/(Green+SWIR) exceeds 0 while peak NDVI stays below 0.2 (standing water is wet AND unvegetated). A step of ≥0.25 NDVI or ≥0.15 water fraction between consecutive epochs is reported as an observed change. The cause is NOT inferred.

• Observed change, not an explanation. Clearing, excavation, fire, demolition, drought, harvest and ploughing can all produce the same reflectance step
• Imagery cannot distinguish them. 30 m pixels miss small features. Persistent cloud can leave an epoch with no usable observation, which is reported as a gap rather than as stability. The step thresholds are LandBenchmark's screening conventions, not published standards. A Phase-1 environmental records review is what identifies prior uses.