35 layers.

Population change for every county in the 50 states and DC, 2020–2024, split into natural change (births minus deaths) and net migration.

Every other layer tells you what a county is; this tells you which way it is going, which decides whether the school, the clinic and the grocery are still there in fifteen years. All 3,109 counties are individually clickable, and each balloon separates natural change from migration — they often point opposite ways. Shrinking on deaths but gaining migrants is an ageing county being replenished by newcomers, and is a good sign; growing births with people still leaving is not. The ramp is brown-to-blue rather than red-to-green on purpose: growth means services survive and also means neighbours and rising prices, so which end you want is your call.

Source: Census Population Estimates vintage 2024, TIGER/Line geometry

Median depth of catalogued water wells, by US county and by Canadian census division — what it actually costs to reach water.

The cost layer for water. Water Rights tells you whether you may use it and rainfall tells you how much arrives; this tells you what it costs to reach it, which on raw land is often the single largest expense of making a parcel habitable. At $30–60 a foot, a 90 ft well and a 650 ft well differ by $20,000–35,000 before a pump. Read the 25–75 spread, not just the median — within-county variation frequently exceeds the difference between counties, because valley alluvium and upland bedrock are different aquifers that happen to share a courthouse. Canadian coverage is British Columbia, Alberta, Ontario and Quebec only; water-well records there are provincial and no national register exists.

Source: USGS NWIS national groundwater site inventory

The 18 western states where water is prior-appropriation or hybrid — buying land on a creek does not give you the right to use the creek.

Only the difficult states are drawn, split into 9 pure prior-appropriation and 9 hybrid. Unshaded is riparian — the good case, though eastern states increasingly permit large withdrawals too. This is the layer most people do not know to look for until after they have bought.

Source: State water law doctrine

Moderate, High and Very High wildfire hazard potential. Unshaded means low risk.

Turn off Moderate to see only the serious ground. Reads fire likelihood, not destructiveness — the Southeast lights up for frequent low-severity pine fires, which is not the same problem as the Western crown-fire risk.

Source: USFS Wildfire Hazard Potential 2023

Mapped earthquake ground motion (Ss, MCE_R short-period, in g) — the figure building codes are written against.

New Madrid carries the highest mapped ground motion in the contiguous US — 3.17 g against 2.02 g for Los Angeles County — and it is nowhere near a plate boundary.

Source: USGS seismic design maps, ASCE 7-16

EPA radon zones 1–3, predicted indoor screening levels.

All three zones draw by default. Untick Zone 3 (and usually Zone 2) so only Zone 1 shows — then shaded means avoid, like everything else.

Source: EPA Map of Radon Zones

Share of each county's mapped soil that USDA classifies as prime farmland, under the strict definition.

Scott County AR at 12% is woodlot and pasture; Caldwell Parish LA at 60% is ground you could farm. Strict USDA definition only — 'prime if irrigated' is excluded, which is why California reads 0% across all 58 counties. Alaska is excluded entirely: SSURGO carries no farmland classification there, so every borough computed as a false zero.

Source: USDA NRCS SSURGO via Soil Data Access

Frost-free days in eight bands, on a diverging scale that goes red at both ends.

Too long is as wrong as too short. Draws everything by default — leave ticked only the bands you would actually accept and it behaves like the other filters.

Source: NOAA 1991–2020 Climate Normals

Distance from each county's most remote corner to the nearest hospital with an emergency room.

Metro rings tell you how far you are from a city; this tells you how far you are from a trauma bay. Counties are banded by their worst corner rather than their centre, deliberately — a centroid sits near the county seat, which is exactly where the hospital is, so ranking by centroid puts three quarters of the country inside 15 miles and says nothing. Balloons name the nearest facility and flag Critical Access hospitals, which are capped at 25 beds and stabilise-and-transfer rather than treat. Straight-line miles, so treat it as a floor on the real drive.

Source: CMS Hospital General Information, Census gazetteer 2024

Share of households with a broadband subscription, by county — national median is about 85%, and rural counties run from under half to nearly everyone.

The modern version of the access question: can you earn a living from this land without driving somewhere to upload a file. It measures what households PAY FOR, not what the line outside could deliver — so a low number means no service, unaffordable service, or an older population that doesn't buy it, and those are very different problems. Used anyway because the alternative is worse: FCC availability data is self-reported by the carriers who benefit from claiming coverage. To resolve a specific road, call the local electric cooperative — in rural America they're often the fibre provider.

Source: Census ACS 2023 5-year, table B28002

Effective property tax rate by county — median taxes paid as a share of median home value. Ranges from about 0.15% to 3.6%.

Nothing else in this map costs you money every year. Purchase price is a one-off; this is the recurring bill, and over a thirty-year hold it frequently exceeds the purchase-price difference between two counties you were choosing between. It matters most exactly where this map sends you — cheap land in a high-rate county isn't as cheap as it looks, and a county losing population with a fixed cost base has one obvious lever to pull. Worth reading alongside County Trajectory. Agricultural and timber exemptions aren't modelled and can cut the bill sharply.

Source: Census ACS 2023 5-year, tables B25103 and B25077

How much of each county is public land you can actually walk onto — open access acreage, not the percent-federally-owned figure every other map shows.

Banded on open-access acreage from PAD-US, so federal wetland easements over private farms do not read as land you can hunt. The balloon names who manages it — BLM, National Forest and a state wildlife area are very different neighbours. High bands cut both ways: recreation and a neighbour who will never subdivide, but a thin private market and the commonest reason a cheap western parcel is landlocked.

Source: USGS PAD-US over Census TIGER land area

Distance rings around 70 US metros, scaled by population: 50 miles at 250k up to 200 miles at 8.5M.

54% of all US counties fall inside a ring, so read these as "how deep in" rather than pass/fail. The edge of a ring is not a wall.

Source: US Census metro population

The same population-scaled rings for 21 Canadian cities.

Context for the northern border. These are cities, not metros, so the Toronto area draws five nested rings and Montreal three. Toggle independently of the US rings.

Source: Statistics Canada

Population-scaled rings for 59 Mexican cities.

Context for the southern border, and the reason the border zone reads the way it does. These are cities, not metros, so greater Mexico City draws seven nested rings.

Source: INEGI

Population-scaled rings for 15 cities across Central America and the Caribbean.

Rounds out the hemisphere view. These are cities, not metros — Guatemala City, Mixco and Villa Nueva are three rings over one urban area. Off by default unless your search reaches that far south.

Source: National statistical agencies

Statewide building-code adoption in three bands.

Deliberately ambiguous. "Not Resistant" means cheaper and freer to build off-grid, and also less regulated construction around you. Decide which direction you want before switching it on, or leave it off.

Source: FEMA Building Code Adoption Tracking (BCAT)

Every US state, Canadian province and Mexican state scored 0–100 on twelve statutory features — permit to purchase, rifle bans, magazine limits, carry regime, preemption.

Five bands, all on a 0–100 scale. Texas, Alaska and 22 other jurisdictions sit at 0; California and New Jersey top it at 100, with New York, Massachusetts and Hawaii at 94. Canada scores 88 and Mexico 72 — both regulate federally, so every province and state inside them carries the same score. Laws that are enacted but not yet in force count here, and the balloon says which ones and when they bite.

Source: State statute, scored on twelve features, current to August 2026

Every US state, Canadian province and Mexican state scored 0–100 on seven requirements — approval, testing, parent qualification, notice, records, subjects, hours.

Five bands, all on a 0–100 scale. South Carolina tops it at 100 — scored on its district route, the only one that runs through the state rather than a private association — with New York at 81 and Pennsylvania and Georgia at 76. Texas asks almost nothing at 10, and 38 jurisdictions require nothing at all. Canada regulates provincially so the spread there is real: British Columbia and Ontario want a letter once a year, Québec wants a learning project the ministry reviews plus its own examinations, at 71.

Source: State statute, scored on seven requirements, current to August 2026

Köppen classes A–E, all 25 sub-classes, across North America.

Background context rather than a filter. Leave off unless you want the backdrop.

Source: Köppen-Geiger classification

Three hand-drawn regions: the southern border, Tornado Alley, and the Gulf and South Atlantic hurricane corridor.

Approximate by design and labelled as such — a sanity backdrop, not a data layer. Leave off unless you want the outline.

Source: Hand-drawn

Share of dwellings in each Canadian census division where a 50/10 Mbps fixed service is available — 97.9% nationally, but a floor of 0% in Nunavut and northern Labrador.

The Canadian counterpart to Broadband, and deliberately NOT the same measurement — that layer counts US households that PAY for a subscription, this one counts Canadian dwellings that COULD buy 50/10. Availability always reads higher, so never compare the two numbers directly. Canada publishes no subscription figure below the province, so the choice was availability at 293 census divisions or subscription at 13. Each balloon carries its province's subscription rate for calibration and splits wired from fixed wireless — the split that decides whether a rural address gets a real connection. Satellite excluded.

Source: ISED National Broadband Data joined to the 2021 pseudo-household distribution; provincial subscription from Statistics Canada table 11-10-0228

Whether CONAGUA has any water left to concession, for each of the 653 aquifers Mexico is divided into — 275 are already in deficit.

Mexico's answer to Water Rights, and the first thing to check south of the border. Groundwater there is federally concessioned, so an aquifer published as overdrawn means no new concession is being issued at all — land above it may be unusable for anything that needs water, whatever the seller says. Banded by published annual availability as a share of the aquifer's own recharge.

Source: CONAGUA published aquifer availability (disponibilidad media anual)

Population change by municipio, 2010 to 2020, from the two Mexican censuses.

The municipio is Mexico's county analogue and there are 2,469 of them, so this is genuinely county-grain rather than state-grain. A decade of change rather than the four-year window used for US counties, because Mexico runs a decennial census with no annual estimates — the two are not on the same scale and should not be compared band for band.

Source: INEGI Censo de Población y Vivienda 2010 and 2020

Share of occupied dwellings with an internet connection by municipio — 52.1% nationally, but a median municipio of just 24%.

The same measurement as the US Broadband layer — dwellings that actually have a connection, not availability — but on a completely different distribution, so do not read the two on one scale. The US county median is about 85%; Mexico's municipio median is 24%, and 465 municipios are under 10%. Bands are cut to Mexico's own spread, not America's.

Source: INEGI Censo de Población y Vivienda 2020

Every installation, range and training area the Department of Defense publishes a boundary for — 819 sites and 37,627 square miles, drawn as the fence line rather than a dot on a city.

The neighbour layer, and the one to turn on before you fall in love with a parcel. A boundary here is a property line and nothing more: what actually reaches a homestead — jet noise, low-level training routes, artillery surface danger zones, restricted airspace — runs past the fence by miles and is published nowhere as a single map, so read every polygon as the visible edge of something larger. Bands are by area, because area is what decides whether a base is a neighbour or an employer: the median site is 448 acres, and 38 sites hold 83% of the acreage. Every site also carries a dot at its centre, sized by band: a boundary shrinks as you zoom out and a 40-acre armoury is sub-pixel from any useful altitude, so the dot is what makes the small sites findable — switch all five bands on, pull back, and every military site in the country is on screen at once. Two flags are worth the click. 221 of these installations are named in 31 CFR Part 802, which puts a nearby land sale to a foreign buyer under CFIUS review — it does not touch a US buyer, but it narrows who you can resell to, and several states have since written their own foreign-ownership bans off the same list. And 32 sites are not active: caretaker, excess or closed. Cheap land beside a closed depot or ammunition plant usually has an environmental restoration programme attached, which is worth finding before the price convinces you.

Source: DoD Military Installations, Ranges and Training Areas (MIRTA), via Esri US Federal Data

Blast rings and idealized fallout plumes for an 800 kt surface burst on each of the ten largest North American metros — Mexico City to Atlanta — scaled from Glasstone & Dolan.

The distance layer: how far from the ten obvious targets is far enough. Each city carries a marker whose balloon lists every radius, five blast rings (fireball, 20 psi, 5 psi, third-degree burns, 1 psi) for a surface burst and again for an optimum-height air burst — the air-burst set is what NUKEMAP shows by default and roughly doubles the 1 psi circle, but produces no early fallout — and four fallout plumes, the unit-time reference dose rate at H+1 for 1000, 100, 10 and 1 R/hr, drawn as the Glasstone & Dolan Table 9.93 cigar for a 15 mph westerly. Because the plume goes wherever that day's winds aloft send it, each contour also carries a wind envelope: the ground it sweeps as the wind swings 60 degrees either side of the prevailing direction. Read the rings as the zone where the city itself is the hazard and the plumes as the zone where the wind is: at 100 R/hr an unsheltered person collects roughly the median lethal dose in four days, at 10 R/hr no acute illness but far past any peacetime limit, at 1 R/hr the concern is crops, pasture and open water rather than people. The 1 R/hr plume is unticked by default because it runs about 700 miles. Everything is the envelope of one steady-wind day over smooth ground for one warhead — real plumes bend, fall in hot spots and would come from many bursts at once — and the geometry is good to about a factor of two. Regenerate for a different yield, burst height, fission fraction or wind with the flags on the generator.

Source: Glasstone & Dolan, The Effects of Nuclear Weapons (1977) §2.127, Fig. 3.73, §7.96, Table 9.93; OTA, The Effects of Nuclear War (1979)

The same blast rings and fallout plumes for the ten largest US military installations by population and the ten largest by acreage — Fort Bragg to Twentynine Palms.

Companion to the Cities layer and to Military Installations, for the buyer who has already put distance between themselves and the metros. Twenty targets in two groups: the ten most populous installations (Bragg, Cavazos, Lewis-McChord, San Antonio, Norfolk, Campbell, Moore, Bliss, Lejeune, Pendleton) and the ten largest by DoD boundary acreage from the Military Installations layer (the Nevada and Utah ranges, White Sands, Wainwright, China Lake, Bliss, Goldwater, Yuma, Dugway, Twentynine Palms). Aimpoints are the cantonment or main airfield, because a warhead goes where the people and aircraft are rather than to the geometric centre of a bombing range — so the Nevada range is drawn on Nellis, at the edge of Las Vegas. Same rings for both burst types, same four plumes and envelopes, same assumptions and the same factor-of-two caveat as the Cities layer, and the same 15 mph westerly, which is the usual case over the continent but not the only one. No Canadian or Mexican base makes either list. Missile fields and bomber bases are a different target set: this layer is the biggest garrisons, not the most strategic ones.

Source: Glasstone & Dolan, The Effects of Nuclear Weapons (1977) §2.127, Fig. 3.73, §7.96, Table 9.93; OTA, The Effects of Nuclear War (1979)

Canada's Emergency Government Headquarters network: the Diefenbunker at Carp, the six regional bunkers, and the zone, municipal, relocation and relay sites that wired them together.

The continuity-of-government network built 1959-63, mapped in eight tiers from the four-storey Central Emergency Government Headquarters at Carp down through the six REGHQs (Nanaimo, Penhold, Shilo, Borden, Valcartier, Debert) to zone and municipal headquarters, federal relocation sites and the dispersed relay stations. Read it as history rather than as shelter: all but Valcartier are demolished, sealed, flooded or in civilian hands, and two of them are open to the public as museums. Every placemark carries a Precision field and a Corroboration field, because the sourcing is uneven - a 'site' coordinate is published for the structure and drops on it, while a 'town' coordinate means the building has never been publicly identified and the pin sits on the settlement. Six pins were confirmed against independently mapped OpenStreetMap bunker footprints; three were snapped to them. Source conflicts are recorded in the balloons rather than silently resolved, including one peer-reviewed paper that wrongly states Manitoba's bunker was never built. Most zone headquarters, all NUDET reporting posts and the DEW, Mid-Canada and Pinetree radar lines are deliberately absent - undocumented by location, or a different programme.

Source: Wikipedia, Emergency Government Headquarters and Diefenbunker; Parks Canada National Historic Site record; Dave's Cold War Canada; Diefenbunker: Canada's Cold War Museum; Beyond the Diefenbunker: Canada's Forgotten Little Bunkers, Material Culture Review; Perth Historical Society; OpenStreetMap

The US continuity-of-government estate: Mount Weather, Raven Rock, Cheyenne Mountain, the Greenbrier's congressional bunker, the Presidential Emergency Facilities and FEMA's regional bunkers.

Companion to the Canada layer, organised by who each facility was for rather than by rank, because the American programme is several overlapping ones rather than a single chain. The civil backbone is the Federal Relocation Arc: every cabinet agency kept an A team for a shelter under its Washington headquarters, a B team for Mount Weather, and a C team for a dedicated site 20-30 miles out, with authority passing down as each layer was lost. Alongside it sit the military command bunkers, one bunker for Congress, the black-budget Presidential Emergency Facilities disguised as water towers, and FEMA's regional operations centres. The sharpest contrast with Canada is that most of these are still open - Mount Weather, Raven Rock, Cheyenne Mountain, Offutt and six regional centres are active federal facilities, and Denton ran the Oklahoma City and Hurricane Harvey responses from the rooms it was dug for. Precision is flagged per placemark and matters more here than in Canada: these are live installations, so a published coordinate marks the gate or the campus, not anything underground. Missile silos, SAGE direction centres and the radar lines are excluded as separate programmes, and only four of the 9-to-75 Presidential Emergency Facilities have ever been named publicly, so only four are mapped.

Source: Wikipedia (Federal Relocation Arc, Presidential Emergency Facility, Mount Weather, Raven Rock, Cheyenne Mountain, Project Greek Island, Fort Ritchie); Center for Land Use Interpretation, Bunkers Beyond the Beltway; Zuckerman, The Day After World War III (1984); Rotenstein FCC filing (2004) and RPPN paper (2010); FAS nuclear forces guide; OpenStreetMap

The overlap: remote US counties where a real connection can still be bought, scored 0–100 as road-network remoteness multiplied by what share of RURAL locations can buy fibre and at what speed.

The layer for anyone who needs to keep an income while living a long way out. It is not a broadband map — layer 11 is that — it is a map of where remoteness and bandwidth actually coincide, and it screens out everything that is not remote rather than ranking it low. Remoteness is drive time on the real road network from USDA's Frontier and Remote codes, not distance on a map. Connectivity is FCC availability graded by technology and speed tier, so gigabit fibre outranks plain fibre outranks fixed wireless — and it reads the RURAL rows only, because rural fibre builds routinely stop at the county seat and a county-wide average would let the town flatter the country around it. Satellite is excluded from the score on purpose: it is available essentially everywhere on this map, so counting it would flatten the layer to one colour. It is printed in every balloon instead, alongside the 67 counties where under half of rural locations can buy any land-based service at all and a dish is the real answer. Every balloon also names the local electric utility's ownership, because what actually predicts a fast remote county is whether the incumbent is member-owned. Read a high score as “worth a phone call”: these are providers’ own claims about what they could deliver, not measured speeds.

Source: USDA ERS Frontier and Remote Area Codes 2020 vintage (released March 2026); FCC Broadband Data Collection fixed broadband summary by geography, data as of 31 December 2025; US Census ACS 2024 5-year table B28002; EIA Form 861 (2024); Census gazetteer 2024 and CBSA delineations

Total annual precipitation across North America in ten bands, on a 5‑arc‑minute grid — drawn where it falls rather than averaged to a county.

The layer that says whether a parcel’s water arrives from the sky or from a pump, and the one to read before well depth and water rights. It is a grid rather than a county map because a county mean is meaningless in mountains — Fresno County runs from 6 inches on the valley floor to 50 on the Sierra crest, and the average of the two describes nowhere in it. Twenty inches is the line the whole map is built around: below it dryland farming is a gamble and the water right is the asset, above it rain‑fed pasture and garden work in most years. Forty inches is the other hinge, where the problem inverts and keeping water out of a building, a road and a septic field costs more than getting it in. It is precipitation, not rain: every band includes the water equivalent of snow, so read it alongside layer 21 to tell an 18‑inch snowpack from 18 inches of thunderstorms. Each band switches on and off by itself, which makes it a filter as much as a backdrop. Every value was checked against the 14,991 NOAA station normals inside the window and the layer description gives the result region by region, including the two places where the grid and the gauges disagree.

Source: CHELSA V2.1 bio12, 1981–2010 normals at 30 arc‑seconds (CC0); checked against NOAA 1991–2020 US Climate Normals; Natural Earth land and lakes

The typical cold night — the mean daily minimum of the coldest month — across North America in ten bands on a 5‑arc‑minute grid.

The layer that decides what a house has to survive, what a pipe has to be protected from, and which fruit and pasture live outdoors without help. It is NOT a USDA hardiness zone and is not labelled as one: a zone is the average annual extreme minimum, the single coldest night of a typical year, which is far colder than the coldest month’s average night and not a fixed distance from it — the gap is small on a maritime coast and very large in a continental interior. Read a band here as the winter a place has, and a zone map for what a plant must survive on its worst night. Checked on every build against the 7,092 NOAA 1991–2020 monthly normals inside the window; the layer description gives the result by region and says which way to lean, because the source runs about 3 °F warm on winter minima and the description explains exactly why.

Source: CHELSA V2.1 bio6, 1981–2010 normals at 30 arc‑seconds (CC0); checked against NOAA 1991–2020 US monthly Climate Normals; Natural Earth land and lakes

The typical hot afternoon — the mean daily maximum of the warmest month — across North America in ten bands on a 5‑arc‑minute grid.

The cooling-load layer, the livestock-shade layer and the can-I-work-outside-in-August layer. It is a typical afternoon, not a record and not an engineer’s design temperature: half the days of the hottest month are hotter than the band you are standing on, and the 1% design dry-bulb runs several degrees above that again. It says nothing about humidity, which is the whole difference between 95 °F in Phoenix and 95 °F in Houston — layer 33 is the nearest proxy the set has. Checked against the same 7,092 NOAA monthly normals as layers 34 and 36; the source runs about 1.6 °F cool on summer maxima, so the description says to lean hot and shows the measurement it is leaning on.

Source: CHELSA V2.1 bio5, 1981–2010 normals at 30 arc‑seconds (CC0); checked against NOAA 1991–2020 US monthly Climate Normals; Natural Earth land and lakes

Mean annual air temperature — day and night, summer and winter together — across North America in ten bands on a 5‑arc‑minute grid.

The backdrop of the three temperature layers: it says what kind of climate a place has rather than what any season does, and it is the number that tracks heating and cooling over a whole year. Its own weakness is the interesting part — an average hides the range, and 45 °F in coastal Oregon and 45 °F in eastern Montana are the same number and nothing like the same place, one mild all year and the other swinging 60 °F between its coldest and warmest months. Read it with 34 and 35, which are the two ends it averages. It is also the most accurate of the three: against 7,092 NOAA monthly normals the median cell is within half a degree, where the two extremes are out by 1.6 and 3.3 °F in opposite directions.

Source: CHELSA V2.1 bio1, 1981–2010 normals at 30 arc‑seconds (CC0); checked against NOAA 1991–2020 US monthly Climate Normals; Natural Earth land and lakes