Quiet Corners

Quiet Corners Field Maps

What it does

Quiet Corners Field Maps turns a rectangle you draw, or one or more USGS 7.5-minute quadrangles you pick, into a topographic map of that area, printed on ordinary Letter paper. The sheets are indexed so they tape together edge to edge, and every sheet carries the retrieval date of each source it was built from. Forest Service road-use seasons are drawn over the USGS Topo base. The PDF is assembled in your browser: no upload, no account, no server of ours, though the map services you load tiles from still see your requests (privacy page). It is at proof stage, so read Status before you rely on a printed sheet.

Open field maps

Choose an area

You build a print set on the Quiet Corners map itself, under “Print field maps of this area” — the same map that shows the county ranking, because there is only one. Once an area is drawn, a strip above the map holds it: Edit goes back to drawing, Undo restores the area you just replaced (once — a second Undo does not go further back), and Clear print area removes it outright, taking the saved recipe with it — and, for an area that came from a shared link, the link’s own area settings in the address bar. Clear leaves the map where it is and every layer as you had it, and it holds even if you reload straight away or the saved map setup was slow to read. Paper orientation, sheet overlap, the UTM grid and print quality live under Options, a disclosure closed by default at the foot of the print builder, and default to Letter portrait, 0.5 in overlap, UTM 1 km grid and 300 dpi unless you open it and change them.

Picked quadrangles are listed under the area controls, each with a Remove button and a note of what removing it saves, and clicking a quadrangle on the map toggles it. This page is where the sources, the printing instructions and the calibration procedure are written down; the privacy page says what the map services receive.

Text on the map itself is sized to be read outdoors on a phone: the Share view, Reset view, Use my location, starting-set and 3D buttons are 14px, the map credit line is 12px and wraps onto a second line rather than shrinking when it gets longer, the scale bar and hover labels are 13px, print-sheet ids are 12px, and the map key is 13px. A key row for something the area does not contain, or whose state is unknown, is shown in a muted grey at 80% strength rather than faded to 55%, so it stays readable (at least 4.5:1 against the panel) and still reads as the odd one out.

Starting sets

Above the layer switches are six buttons, each a named set of those same switches for one kind of outing. A button is a shortcut, never a mode: it ticks and unticks the boxes below and moves the two choices above them, you can see exactly which ones it moved, and every one of them is still yours to change afterwards. No layer reaches the map any other way. Before you press one, its own screen-reader description names which switches it would actually move — a switch already set the way the preset wants is left off that list, and the base map is named only when yours disagrees with it.

• Map only — every overlay off, on the topo base, except Road names and Named places & landmarks. Those two open on, and this preset leaves them as they are. The ground itself, with only road and place names over it.
• Trails & terrain — hiking trails, named places, and the hillshade relief surface, on the topo base. It answers what the ground does around a trail: which way it climbs, and what the named summits, gaps and ridges nearby are called.
• Land access — federal land designations, MVUM road and trail designations, Forest Service and BLM recreation sites, and the state land families (game lands, state parks, other state-managed areas) your drawn area covers. It answers which public and state-managed land reaches this ground and how to get onto it. State land rows only fill in once an area is drawn — draw one first if a switch reads “unavailable.”
• Compare counties — county boundaries and the Quiet Corners county ranking, the same context surface the county-score row on the dashboard turns on. It answers how one county compares to the ones around it. The ranking’s colour shading appears only while that row is on; with it off the dashboard map draws the county outlines alone, so the base map under them is not tinted.
• Night sky — the night light emissions surface, with the boundary lines out of the way and no base map under it. It answers whether tonight is worth the drive from here, and how far out you would have to go. Aurora and air quality are not switched on by this set; pick either one yourself under Land, sky & light context. Precipitation radar is deliberately not in this set: it reports what the beam met aloft in the minutes just past, which is not cloud tonight — the hourly conditions card is what answers the hours ahead — and it is a second full-screen picture over the first, which is the clutter limit below.
• Public land — national parks, forests, wilderness and refuges; Forest Service land ownership parcel by parcel; MVUM road and trail designations; hiking trails; Forest Service and BLM recreation sites; on the topo base. It answers where you may legally go and what is there when you arrive. A boundary still says nothing about access — see the section below.

Two buttons beside the presets, Undo and Reset, both start out disabled. Undo restores exactly the switches the last preset you pressed moved, once, and is disabled again afterwards — a second Undo does not go further back. Reset becomes enabled after you change a layer switch, the base map or the context surface by hand, or after a preset moves something. Changing the highlight or the hillshade light does not enable it. To put those two back, use Clear highlight and Reset light.

Once Reset is enabled, pressing it puts back every layer switch (the ones that open on, Road names and Named places & landmarks, go back on; all others go off), the base map (USGS Topo), the context surface (No context surface), the highlight, and the hillshade light.

Reset leaves alone your drawn area, the print recipe, a saved pin, and Map label size and Map label crowding.

Both buttons say what they did on the status line below the map, the same line every other action here uses.

What these sets are held to. Two of the three things worth measuring here are enforced by a test rather than by good intentions, and a set that exceeded either could not be built at all. The third, task completion, is not mechanical: it is watched rather than asserted, and what the code holds each set to is that it states the question it claims to answer.

• Load size. Each layer switched on is one query of its own against its service, and a query pages up to twenty requests. The public-land set is six layers, a worst case of 120 requests for one settled view; the night set asks the vector services for nothing at all, because its surface is a picture handed straight to the map. Six is the ceiling until a wider set has been measured.
• Clutter. At most one continuous-colour surface at a time, and that is now enforced rather than promised: Land, sky & light context is a choice of one — the steepness of the ground, night lights, radar, aurora, air quality, wildlife records, the Quiet Corners county ranking, or No context surface — exactly like the base map above it. It used to be four independent checkboxes, so a reader could stack four filled surfaces over each other and nothing stopped them; the light-emissions surface is drawn at 80% and the radar at 55%, and those two alone leave under a tenth of whatever is beneath them showing. No context surface is where the panel starts, and it stays a row of its own so turning them all off again is always one click. On the map key, the public-land set draws nineteen distinct marks and the night set draws one. The county ranking is the one surface here with no service behind it at all — it is the dashboard’s own local data, drawn on screen only, and it is never in the printable-vector fetch loop and never on the printed key: a screen-only layer switched on for the screen used to still be listed as a source on the sheet that never carried it, and the credit list is now built from what actually reaches paper rather than from what the reader had switched on.

Finding and checking a layer

The layer list grows with the map, so above it is a search box and an “Only layers that are on” toggle — the search narrows the rows shown by name or description as you type, the toggle hides everything not switched on, and neither ever changes a switch itself: turn off the search or the toggle and every row, on or off, is exactly as you left it. A line above the list states how many layers are on and how many are still loading. Each federal or state layer carries a small word next to its name — off, loading, ready, empty (fetched, nothing here), partial (the service stopped part way through), unavailable, or zoom in (too far out for this map to ask yet) — and its own “About this layer” details: where the data comes from, its licence and coverage caveats, and whether it has been fetched for the area drawn or not yet. A layer reading unavailable or partial gets a Retry button that asks that one service again, without re-fetching anything else or moving the map; unticking a layer’s switch is Clear — it removes the layer’s lines and its mark from the map key immediately, and the map itself does not move.

Finding a place

The finder is on the map itself, not in the Layers panel. Press Search places to open its box. As you type, it matches the towns Quiet Corners scores, with no request to any service. Results are grouped as Towns, Geographic names and Coordinates. A pair of coordinates is read as a point on the map. To look up a name in the USGS place-name record, press Search, or press Enter with no suggestion highlighted. That is the only time the finder asks the USGS, and it never does so as you type. If the USGS cannot be reached, the finder says so and shows the matched towns only.

Highlight

Two buttons in the Layers panel, Highlight trails and Highlight boundaries, thicken the lines of one drawn layer and dim everything else on the screen; Clear highlight restores exactly what was on screen before, never an approximation of it. A summary line states in words whatever the screen currently differs from the default view, and the map key dims and brightens its swatches the same way the map itself does, from the same state. None of this is a mode: switching a layer on or off, or drawing a new area, works exactly as it did before a highlight was ever pressed.

Advanced, closed by default under the buttons, exposes the same handful of knobs by hand: how much to dim everything not highlighted, how heavy the drawn lines are, how solid a filled polygon is, and a menu to spotlight any one layer directly rather than only trails or boundaries. Touching any of these releases a Highlight-button press back to a plain setting — there is no state where a highlight and a manual edit disagree about what is on screen.

Map appearance, closed by default under the map key, holds two things. Map labels holds the two label controls, explained under Names & landmarks below. The hillshade light controls — light azimuth and altitude, Sun now and Reset light — only mean something while Hillshade (adjustable light) is the chosen context surface, so they show only then; the rest of the time Map appearance says where to turn that surface on.

None of this reaches the printed sheet. The PDF is built from the print recipe alone — area, paper, scale, overlap, grid and the layers you checked — and the emphasis state above is never one of its fields; asking for it is refused outright rather than silently ignored. A screen highlighted for reading is not a page marked up for printing, and the sheet’s credits list is built the same way, from what was actually fetched for it, not from anything a highlight touched.

Three settings your device or browser may already be asking for are honoured here specifically because dimming is designed to reduce contrast for everything but the highlight, which a reader who has turned up contrast is asking never to happen: more contrast and Windows’ forced colors both cancel every dimmed opacity outright rather than trying to draw a softer version of it, and a highlighted button still marks itself without relying on colour alone. The reduced-motion setting costs nothing extra here, because a highlight or a clear takes effect immediately rather than fading in.

Terrain in 3D

Press 3D on the map to open a separate relief view of the area on screen. It shows the shape of the ground, built from a mixed-source elevation tile set (see the Source decision record below), with the USGS topo base and a hillshade on top. It shows terrain, not analysis. It does not compute shadows, line of sight or where you can walk.

Inside the panel:

• North up / reset view returns to the view the panel opened on, pointing north and looking straight down.
• Tilt up and Tilt down change the viewing angle, 10 degrees at a time, between flat and 60 degrees.
• The exaggeration checkbox (named Exaggerated terrain (1.5×) for screen readers) steepens the relief from the default 1× to 1.5×. Until you tick it, its visible label reads “Exaggeration: 1× (default) / 1.5× — exaggerated”. Once ticked, the label reads “Terrain exaggerated 1.5× (not true scale)” so the panel says the relief is steepened, and it goes back when you untick the box. Judge how steep the land really is from the unticked view.
• Close shuts the panel and frees everything it loaded.

What it draws from the 2D map: the layers you switched on that are lines or outlines are draped over the relief in the same colours. That means trails, MVUM roads and trails, recreation sites, and land and county boundaries. Named places are lettered on the relief at your Map label size and Map label crowding, and a name behind a ridge is left off rather than shown through the hill. While the Trails layer is on, each named trail is lettered on the relief too, on the part of it that is in view and not hidden by a ridge, and it shares the same room-making as the place names, so on a crowded view some trail names are left off until you zoom in.

What it does not draw: Road names stay on the flat map, and no road names are lettered in 3D. Camera layers that are on stand on the relief as pins in their flat-map colours. The raster surfaces (hillshade light, contour lines, slope, land cover, tree canopy, night lights, radar, aurora, air quality, wildfire, wildlife) cannot be draped. A list headed Layers in this view names the layers you have on that the panel can show. A line, outline or name layer is marked “shown”, or “loading in 2D first” if the flat map has not fetched it yet, and a raster surface is marked “Available in 2D”. Camera layers that are on are listed as “shown”. The list leaves out the reading rows, the extra sheets and the base map. It can still list Road names as “shown”, although the 3D view draws none.

Nothing for this view loads until you press 3D: no rendering library, no elevation tiles, no extra base tiles. The panel needs WebGL. With WebGL off, you get a message in place of the panel and the 2D map is unchanged.

Print

How to print the sheets at the right size, what the printed sheet carries, and how to check your printer.

Printing instructions

Print at Actual size (100%) and turn off any Fit or Shrink option. Each sheet carries its own page ID, and inside each grey overlap band the ID of the sheet that adjoins there with an arrow pointing at it. Lap two sheets so the arrows meet, slide them until the corner crosses in the band sit on top of each other, and tape. Tape each row left to right first, then join the rows. A set made from separate quadrangles can skip grid positions: the index page shows which sheets exist, and an edge with no arrow has no neighbour.

Road names

Road names is a layer switch in the same federal group as Named places & landmarks, but on by default, because a map with no road names leaves the reader guessing which line is which. It draws lettering along each road from USGS The National Map transportation data, on the screen only: it is never fetched for, or printed on, a sheet, never counts toward the “layers on” line above the list (that line is about what a sheet will carry), and carries its own word beside its name instead. Nothing is asked below zoom 10; interstates and other controlled-access highways appear from zoom 10, secondary highways from 11, connecting roads and 4WD tracks from 12, and local roads from 13. Highways print as route shields (I-26, US 74, NC 107) and other roads by name. After a pan or zoom the map waits 400 ms for you to stop, then asks once for the view you stopped on plus a margin around it, so a small drag inside that margin asks for nothing; a newer view cancels an older request, so the names on screen always match the ground in front of you. How many roads are lettered follows the same Map label crowding and Map label size controls as the other map names: a road is placed only where its label clears the town and orientation names, the named places and the camera pins already on the map, and where it does not fit it is simply left out, never drawn over them. That holds in either order: Named places drawn after the roads (switched on later, or their answer arriving late) send the road labels back through placement, and they step clear of the new names. The starting-set buttons never touch this switch: Road names is lettering on the map, not a topic, so no button turns it off and no button’s preview says it will; Reset puts it back on with the rest. If the service had more roads than one request returns, the row says More roads than shown; if a request fails it says Road names did not load and keeps the names already on screen rather than blanking the map — an unreadable answer is never presented as “no roads here”. A name is a published label, not a statement that the road is open, maintained or passable. While the switch is on, the map’s credit line names the source; switching it off removes the credit. In 3D, the road names appear over the relief too: open the 3D view with the switch on and it letters the same roads the screen map would for that view, in the same amber on a dark plate, at the same Map label size, turned to lie along each road as you see it (still readable, never upside down) and kept clear of the 3D view’s own credit line and of the place names there. It asks the same service the same way, once the camera has held still, and says the same things in its “Layers in this view” list: Road names: shown, Road names did not load, a plain “no road names found in this view” for an empty answer, or “zoom in to show them” below zoom 10. The 3D view is a snapshot taken when you open it, so to turn road names off there, switch them off on the screen map and open 3D again; the 2D map’s labels are untouched by a trip to 3D.

Names & landmarks

Named places & landmarks is a layer switch listed under the same federal group as the base map: summits, gaps, ridges, falls, springs, lakes, populated places and the rest of GNIS’s vocabulary, one point per name from USGS’s Geographic Names Information System. It is on by default, so a reader who touches nothing sees the names; untick it for a plain map, and a saved project that stored it off keeps it off. The USGS Topo base prints its own names, and these decluttered ones may repeat some of them. While it is on, the map asks for whatever area is on screen, with its own per-view pass like Road names: no drawn print area, no zoom floor beyond the lowest zoom a name can draw at (8) and no area ceiling gate it, a pan that stays inside the box already fetched asks for nothing, and a refresh that fails leaves the last labels on the map and says so. Panning or zooming out of that box asks again for the new area, and an answer for a rectangle you have since left is dropped rather than drawn, so the names on screen always match the area in front of you. A search that genuinely finds nothing prints “No named terrain found in this area” — a real answer, not a failure — and a search that did not finish (a timeout, an error) says so distinctly and never draws a silently incomplete set. The request follows the zoom: only the kinds of name that zoom can draw are asked for (places from zoom 8, peaks from 9, and so on), and when an area holds more names than one request carries, the most important kinds are fetched first and drawn, with a note naming what zooming in adds (“Showing places and peaks here – zoom in for creeks and lakes”). A capped answer still blocks the PDF, like any incomplete one. An answer that came back incomplete is still listed under Places in this view below, with a notice that it is incomplete — the map itself labels none of those names until a full answer arrives.

Screen labels declutter automatically as you zoom: only as many names as comfortably fit are lettered, and only a labelled name gets a marker on the map. Every name loaded for the area you’re looking at — labelled or not — is listed under Places in this view below the layer switch, with filters for the kind of place and a toggle for “Only places not labelled on the map”, so nothing the map declutters is ever hidden from you. Tab to a row, or tap it, and it draws that name’s marker and opens the same card a map pin opens — its kind, its state and county, its coordinates, the note that GNIS gives one published point per name and not every place the name is used, its approximate terrain elevation fetched from USGS 3DEP once you open the card, and buttons to save the place, copy the coordinates, or copy a shareable link to that place — without moving or re-centering the map. Saved places can also be downloaded as a GPX file, with Download GPX under Saved places, for a GPS app or another map.

Forest roads and trails open a card on a tap. With the forest-road or trail layers on, touching a drawn line (within about 12 px of it for a finger or pen, 6 px for a mouse) opens the same card a click does: its name, what kind of road or trail it is, the maintenance level or trail class the source gives, and the source’s credit. Where two or more lines pass within that distance, a short list of them appears first and you pick one; the map never picks for you. Under Places in this view, the forest roads and trails in view are listed in their own groups, along with a group named “Roads lettered on the map” for the named roads Road names has actually lettered; choose a row with the keyboard and the same card opens, and closing it with Escape puts you back on that row. The lettering itself is not a touch or tab target: a press on it reaches the map beneath.

Two controls set how names look. Both live in the Layers panel, under Map appearance › Map labels.

• Map label size sets how big every name the map draws prints: Normal (14 px), Large (18 px) or Extra large (24 px). It works on the flat 2D map and in the 3D view, and it covers town names, road names and named places alike.
• Map label crowding sets how many named-place labels show at once: Sparse, Standard or More. A numerous kind never crowds a scarcer one out entirely; each kind keeps a share of the slots.

Both are saved with your other display choices and carried in a shared link. Reset does not change them. Every name is lettered on a dark plate in a clear sans-serif, so it stays readable over a county colour, a road or a trail. The sizes follow your browser’s own text-size setting, and the room each name gets grows with it. Towns drop the Census’s class word, so the map says Sylva, not Sylva town.

A name on the printed USGS Topo base is not the same thing as a name this layer draws, and the base map’s own text cannot be decluttered the way this layer’s labels are: if the topo base already carries a name near a point this layer also names, the two may repeat on paper. See the Named terrain on the sheet section below for exactly how a name reaches a printed sheet.

A town this layer would name is left off screen if the dashboard has already lettered it. The same map also carries its own orientation labels — the towns Quiet Corners scores, and the larger places you would orient by — and a populated place here sharing one of those labels’ name and state is not drawn a second time under this layer’s separate rules, allowing for the raw Census suffix a town label can carry (Ashland town still matches Ashland). Screen only: its point is left out of both the map and Places in this view, but a printed sheet still draws straight from this layer’s own answer and is not decluttered against those labels.

Named terrain on the sheet

Named summits, gaps, ridges, falls, springs and other GNIS terrain names print as a small shape (one per kind, listed in the key) beside the name in physical type, sized and positioned from the sheet’s own scale — never from how far you have zoomed the screen. A name shares the same on-paper space as the road, trail and site labels already there: when a name’s text would land on top of something else, it is nudged to the next open spot near its point in a fixed, repeatable order, so the same area prints the same way every time. The point marker itself never moves — only its name can be bumped, and if no open spot is found nearby the name is left off and counted in the build record while its marker still prints, so you can see something is there even without a label crowding the sheet. Names print in the same typeface as everything else on the sheet; an accented spelling (Cañón, Étang) prints as GNIS spells it, and only a character that typeface cannot represent at all falls back to ?, the same fallback every other label on the sheet already uses.

Like the base map and MVUM (see Sources), an interrupted or capped fetch of GNIS names blocks the PDF rather than printing an incomplete set of names with nothing said — see “Named terrain (GNIS) must load completely before export” below.

Printing the key on its own

Print the key, under Options at the foot of the print builder, hands you a one- or two-page PDF of every symbol the layers you have switched on can draw, with the notices those symbols oblige. It is the same key a built set carries at its back — the same table, the same strokes — so it can be printed and read before a single map sheet is spent, and kept with the sheets afterwards. It lists what the layers can draw, not what a particular area contains; the on-page key is what narrows after Check sources. The two keys are deliberately not identical: the on-page key describes the map on your screen, so it drops the UTM grid and the sheet-overlap marks (which exist only on paper) and says of anything the current zoom is too coarse to draw that it appears when you zoom in.

Calibrating a printer

Before printing a set on a printer you have not used for these sheets, print a calibration sheet: either add ?calibrate=1 to the Quiet Corners map’s address, or open Options at the foot of the print builder and press Calibration sheet — both hand you the same one-page PDF instead of the map builder. Print that one sheet at Actual size (100%) with Fit and Shrink off, and it answers the two questions no test can answer from a file:

• Scale. Lay a ruler along either 10 cm bar. 9.9 cm instead of 10.0 means the printer prints 1% small, and the map’s scale bar will read the same way.
• Unprintable margin. The sheet carries nested rectangles 0.25, 0.20, 0.17 and 0.15 in from the paper edge, each labelled. The smallest labelled inset that printed complete on all four sides is the smallest inset that is safe on that printer; the sheets use 0.25 in today. Four corner marks sit on a real sheet’s frame corners, so a missing mark means the frame itself is being clipped.

Keep the sheet with the printer. It is good for every future job on that machine.

What a boundary does not mean

State game lands, parks and other managed areas are drawn as boundaries: a line marks where a designation begins and ends. It is not a statement that the ground inside is open to you, and it is not a statement that the ground outside is closed. Seasons, permits, camping, fires and vehicle access are the managing agency’s to state, not this map’s to infer, and they differ unit by unit. An empty result means the state’s service returned nothing for that area, which is not the same as “no public land here”. Confirm access and current closures with the managing agency before you travel.

Parcel boundaries are drawn from the counties’ own recording and tax maps, at 1:24,000 only. A parcel line is a record, not a survey: it is drawn to locate a property for taxation, not measured to establish where its corners are, and a recorded line and a surveyed line disagree routinely. It is not permission to enter. Coverage differs county by county, so a blank area means the county published nothing there rather than that the ground is unowned. No owner name, mailing address or occupancy field is ever requested — the catalogue records each state’s owner-bearing fields as a refuse-list, and every parcels layer this map offers has had its actual field VALUES audited, not just its field names, because a field named for a category can still hold people: North Dakota’s Ownership is declared with a nine-value domain and one county had filled it with individual landowners.

Mapped wetlands are neither a boundary nor a landholding, and the map draws them in their own colour for that reason. A wetland edge is approximate by construction — it is a line someone drew on an aerial photograph, which is why it is drawn broken rather than solid — and it carries no statement about access, ownership or what any law regulates. Where nothing is drawn, the layer’s own row says whether the ground was mapped and holds none or has never been mapped at all; those are different facts and they look identical on the map.

The same sentence is printed on the map key of every set that draws a boundary, because the sheet in your hand is where it matters.

On the screen each boundary family — National Forest, Wilderness, National Park unit, Wildlife Refuge, state game land, state park, other state-managed area — has its own colour and a light fill, so you can see where one begins and tell it from the next; the key beside the map shows the same colours. On paper they are all one blue, told apart by their line pattern, because a printed sheet is a grayscale halftone and a hue would not survive it.

Lengths and elevations on a card

A forest road or trail card states its length in miles (kilometres beside it), measured along the drawn line on the WGS84 ellipsoid, the same surface GPS uses, not on the flat screen. The number is the length of what is loaded in this view. When the whole road or trail is in the area the map fetched, the card says so plainly. When the line reaches the edge of that area, the source answered only in part, the preview drew only some of the lines of that kind, or the source gave the line no route id, the card reads “approximate — only the part loaded in this view” (or says why), because a length is never claimed whole on a guess. A map line is also only as exact as the source drew it: a card length is not a surveyed distance.

A place card’s elevation is the ground elevation at that point from the USGS 3D Elevation Program, a surface interpolated between lidar returns. The card names the source and says “vertical reference not stated by source” and “resolution not stated by source” because this site has not recorded either for the elevation samples. The same USGS service answers at whatever grid size a request asks for: its slope layer, measured here, gave a different steepness for one hillside at a 3 m grid than at a 270 m one, so a grid size would be a guess. “Unknown” means the source returned no data there, which is not the same as sea level.

Not measured: raster pixels, distance over the 3D surface (slope, switchbacks), and polygons larger than a hemisphere (measured as their smaller side).

Status

Proof stage. The engine builds real PDFs; what is outstanding is the physical verification — a ruler-and-tape test on printed sheets — which has to pass before the builder is offered as finished work. Until it does, the builder carries that caveat on its face, wherever it is hosted.

Sources

• USGS Topo basemap — National Map basemap tiles, the base map on every sheet, and the screen’s base unless you choose another.
• USGS Imagery Only basemap — National Map imagery tiles, an optional screen base on the same host as the topo. Measured 2026-09-12: cached through level 16, level 17 answers 404; the service states its data was refreshed June 2024. It shows ground cover as flown, on a multi-year cycle, not access.
• USGS Shaded Relief Only basemap — National Map shaded relief tiles, an optional screen base on the same host. Measured 2026-09-12: cached through level 13, level 14 answers 404, so closer in the same picture is enlarged and the layer row says so. It shows the shape of the ground, not what is on it.
• USGS 7.5-minute map index — National Map map indices, the named quadrangle outlines you can pick from.
• US Forest Service Motor Vehicle Use Map (MVUM) — EDW MVUM service, road and trail designations with seasonal dates. Tap or tab to a route and it says who may drive it and when — “Open to passenger cars Apr 1 – Dec 15”, or “Closed to passenger cars; open to high-clearance vehicles year-round”. A blank field is the inventory not answering, so it is never reported as closed and never as open; a route the source called seasonal without giving a window says the dates are not stated rather than reading as year-round.
• US Forest Service road inventory (Road Basic) — EDW RoadBasic service, the maintenance level and surface type of a forest road. MVUM answers whether you may drive a road; this answers what your vehicle would meet on it. Tap or tab to a road on the map and it says so in the inventory’s own words — “Crushed aggregate or gravel; suitable for passenger cars (level 3)”.
• USGS 3D Elevation Program (3DEP) — 3DEP ImageServer, elevation sampled along a drawn route for the optional summary page.
• Annual NLCD land cover — Annual NLCD Collection 1.1 ImageServer, what the ground was mapped as, as an optional screen surface. It is served from the USDA Forest Service’s imagery platform rather than from MRLC’s own host, which answered none of 33 probes over 19 minutes when this was measured — the product is USGS’s, the door it comes through is the Forest Service’s, and the credit names both. The year is always stated, because the service will not: ask it without one and it quietly answers its newest, which is 2024, and the same square of ground in NW Las Vegas reads Shrub/Scrub for 1985–2005 and Developed for 2006 on. The class numbers are published nowhere — the service’s own attribute table comes back empty — so the fifteen this map labels were recovered by exporting one extent twice, as raw values and as rendered colours, and pairing all 65,536 pixels; the sixteenth, Perennial Ice/Snow, did not occur in that extent and the key says its class number is unverified rather than passing it off as known. Every 30 m square is about a quarter-acre and holds one class, so a road or a creek narrower than that disappears into whichever class won the square. The lower 48 states only. There is no NLCD land-cover service here for Alaska, Hawaii or the territories, and the layer says the view is outside coverage rather than drawing it as bare ground — a distinction the picture itself cannot make, since an out-of-coverage export and an ocean export are byte-identical.
• NLCD percent tree canopy cover — TCC ImageServer, the modelled share of a 30 m square that tree crowns cover, on the same host and as the alternative screen surface — the two are radio buttons in one group, so one is drawn at a time. Its year is stated for the same reason and reaches 2025, one year further than land cover, so a card showing both dates each separately. The service’s own serviceDescription still advertises the older v2023-5 and contradicts its description; the catalogue of rasters settles it in favour of v2025-6, and the catalogue is what this map reads. Two values in the data are not percentages — 254 marks a non-processing area and 255 the background — and they are never averaged into one or shown as canopy. 70% canopy is not 70% shade on a trail: it is a view from above in one year, smoothed across a time series, and says nothing about the season or about what is under the crowns.
• USGS 3DEP terrain slope — the same 3DEP ImageServer, rendered through its Slope Degrees raster function as an optional on-screen surface. Steepness is shown in degrees and the key gives the percent-rise equivalent beside every band, because percent rise is tan(degrees) × 100 — 45° is a 100% rise — and the service labels no unit anywhere. The colours are this site’s own: the service publishes no class breaks for its ramps, so its picture cannot honestly be given a key. The export is asked for as png32 so that ground the elevation model does not reach arrives with a transparent alpha channel and is left undrawn, instead of collapsing onto the same value as level ground. Which way a hillside faces is read only where the slope is above zero — flat ground has no facing direction, and the service reports flat as 0, the same number as due north. It is arithmetic on a bare-earth surface averaged over a cell tens of metres across: not a statement that anywhere is safe to travel, and not a picture of sun or shade at any hour.
• USGS 3DEP shaded relief — the same 3DEP ImageServer, rendered through its named Hillshade Gray raster function, as an optional companion sheet. It shows the shape of the ground; the contour values stay on the topo sheet.
• USGS 3DEP contour lines — the same 3DEP ImageServer, rendered through its Contour Smoothed 25 raster function as an optional screen surface. One line per 25 metres of elevation (about 82 feet), so it will not match the feet-based lines on a topo sheet. They are drawn from a gridded model, not surveyed, carry no elevation labels, and each view is a live render that can take several seconds. Screen only: it replaces whichever other context surface was chosen and never reaches the printed PDF.
• USGS PAD-US public access — PAD-US Public Access FeatureServer, the access category a land manager filed for each protected unit your area touches, as an optional answer in words rather than as ink on the map. Four categories exist and the layer names all four in the service’s own wording — Open, Restricted, Closed and Unknown. Unknown matters most: the service’s own picture defines colours for only three of them, so a map drawn from its symbology silently drops the 43,047 records that carry Unknown, and here it is a fourth answer rather than a shade of Open. What a category is not, ever, is permission: it describes a management regime at whatever date it was filed, says nothing about camping, fires, hunting, driving or being there after dark, and a boundary still encloses private inholdings. The acreage shown beside a unit is that whole unit’s area and says so — six units touching a 491,877-acre rectangle carry 3,033,604 acres between them, because the polygons run far outside it, so no acreage is ever added up across an area. Outside the conterminous states PAD-US fills what it does not inventory with a single 105,513,892-acre background polygon whose category is Unknown; it is excluded from the request and dropped again on arrival, and the row says the inventory stops there rather than repeating it. The service publishes no version and no date on a record, so the row carries the date it was retrieved.
• EPA Level III ecoregions — the Omernik ecoregion service, read from a copy this site serves rather than from EPA on every visit. It names the kind of country a place is in — “Northern Lakes and Forests”, within “NORTHERN FORESTS” — drawn by analysts from geology, landforms, vegetation, climate, soils, land use, wildlife and hydrology together. It is a classification of the landscape, never a reading of it, and it says nothing about any particular acre. The data has not moved since December 2011, so there is no freshness to gain by asking EPA live, and the national request is refused outright unless it carries a generalization parameter — baking that into a committed file removes the failure rather than guarding it. The copy is 85 regions and 949,718 bytes, fetched only when the switch goes on. Conterminous states only: Alaska and offshore answer with nothing, and the row says the map stops there rather than that the ground has no character.
• USDA National Agriculture Imagery Program (NAIP) — NAIP imagery service, the optional aerial companion sheet.
• USGS National Map government units — govunits MapServer, federal designation boundaries: National Park System units, National Forests, designated Wilderness and National Wildlife Refuges, plus county boundaries from the 2025 TIGER/Line county file. County lines come from this service rather than the generalized outlines the dashboard’s county map uses: those sit a median of 250 metres from the real boundary and up to 2.4 kilometres at worst, which is most of an inch on a 1:24,000 sheet and in places nearly four. Bureau of Land Management is deliberately not drawn — its polygons are Field Offices, administrative jurisdictions rather than holdings, and the Moab Field Office alone spans 2,907 square miles of mostly non-federal ground.
• FEMA National Flood Hazard Layer — the NFHL map service, the regulatory floodplain, as an optional screen overlay. It is the one context surface here that is not about a moment: an engineering study modelled the 1%-annual-chance floodplain, FEMA adopted it for insurance and floodplain management, and it was drawn on a panel. So it never says where water has been or will be, and it is never drawn in the same breath as a weather alert. Its age belongs to the panel, not to the screen. The zone record carries no date at all — the effective date is a join to the FIRM panel underneath — and one 0.12° × 0.07° view measured panels adopted 2012-11-02 and 2016-12-08, four years apart, so the row gives the range rather than one date that would be false for most of what you can see. Zone X is not one thing: the 0.2%-annual-chance area, minimal hazard and areas behind a levee all arrive as X, and the subtype is what says which, so it is always shown. Where FEMA has published no study the layer draws nothing and says so — that is ground nobody assessed, never ground found to be low risk — and the map service is asked whether a study exists at all before it is asked for any zones, because zero zones over a rectangle means those two different things. A base flood elevation is printed only where the record states the datum it is above; 154 of the 159 zones measured carry -9999 in that field, which is the service declining to say. Dense views are refused rather than drawn: one 0.6° × 1.0° extent holds 29,016 zone polygons, so the layer counts first and asks you to zoom in above 1,500 rather than fetching fifteen pages of shapes nobody can tell apart.
• US Forest Service surface ownership — EDW SurfaceOwnership service, the Forest Service’s own Land Status Record System, parcel by parcel. A National Forest’s proclamation boundary encloses private inholdings; this layer says which ground inside it the Forest Service actually owns and which it does not. Dense by nature — 81 parcels in one quadrangle — so it is drawn as unlettered hairlines at every scale rather than left out. The record’s case name is the person the ground was bought from and is never requested.
• Bureau of Land Management recreation sites — BLM National Recreation Sites and Facilities, the 1,304 places BLM publishes as recreation areas. BLM land is the one federal estate this map drew boundaries for and gave you nowhere to stop; this is the layer that answers it. Each site carries the agency’s own directions and description — reduced to plain text, never rendered as the provider’s HTML — its own page on blm.gov, and the last-updated date that record states, which runs per record rather than per service. Opening a BLM pin — click it, or tab to it and press Enter or Space — shows those lines as a card you can read, scroll and shut with Escape; hovering still gives just the site’s name. Where a record says nothing, the card says the inventory did not say, never that there is no access. It is a list of recreation sites and nothing else: the source’s own night-sky language appears in 3 of the 1,304 records, so the layer makes no claim about what the sky does over any of them.
• US Fish and Wildlife Service National Wetlands Inventory — NWI Wetlands service, the mapped wetlands of an area, as an optional overlay. It is an inventory drawn from aerial imagery, not a survey and not a visit: a photo-interpreter read wetland vegetation, hydrology or geography in a picture taken in a stated year, and FWS says a margin of error is inherent and that inspection on the ground may revise any boundary. The year varies enormously and is read per place from the service’s own source layer — Horicon Marsh in Wisconsin was mapped from 1994 black-and-white photography at 1:20,000, Tonopah in Nevada from 2022 colour infrared — so the layer states the year for the area you drew rather than one national date, and says the scale is not stated where the record does not give a usable one. A polygon says nothing about access, ownership, whether the ground is wet today, or whether any law regulates it: FWS uses a biological definition that need not match Clean Water Act boundaries, and some farmed wetlands, seagrass and submerged aquatic vegetation are excluded from the programme outright. A blank area is two different facts, and the layer says which: ground that was mapped and holds none, or ground the inventory has never mapped — every unmapped area is in Alaska or the Pacific territories. A large area is refused rather than fetched: the service is asked how much it holds first, and past 4,000 wetlands nothing is requested, because asking anyway takes minutes and times out rather than failing.
• NOAA National Weather Service forecast — api.weather.gov, the hourly forecast for a selected place: temperature, precipitation chance, humidity, dew point and wind. Two steps, because a latitude and longitude is not a forecast: /points/{lat},{lon} names the forecast office and grid cell and hands back the forecast URL to ask. Temperature arrives in whichever unit the response states — the same hour answered 86 °F and 30 °C seconds apart depending on the request — so the unit is always read, never assumed. A forecast belongs to a grid cell rather than to a summit, so a valley and the ridge above it can share one. Wind is reported as the direction it comes FROM; an hour rounded to “0 mph” is the product’s rounding floor and is never reported as calm, and an hour with no stated direction says so rather than pointing north. Gusts are not in the hourly product at all — they come from the same office’s forecast grid (/gridpoints/{office}/{x},{y}), where every value covers a span of hours rather than one hour and each field starts and ends its spans in different places, so the spans are spread over an hour grid and matched by time before anything is read; a gust is reported as a peak rather than a sustained speed, and a forecast that states no gust is never read as no gusts. Cloud cover comes from the same grid, as the share of sky the forecast expects to be covered in that hour — stated by the service in percent, and read only when the service says percent, because a count of eighths printed as a percentage would report a covered sky as a clear one. An hour the grid does not state is reported as unknown, which is not the same as a clear sky, while a stated 0% is a real forecast and is shown as one. The hours are the place’s own: the card names the hour in the place’s zone, so on the night the clocks go back the two 1 AMs keep their own cloud values and on the night they go forward the hour that never happens is never printed. Past the end of the forecast’s own stated span the weather is simply not out yet, which the card says while the calculated sun and moon times still stand, and a forecast older than its refresh window says it is out of date rather than showing its numbers as current. Outside the United States and its territories the service answers “no forecast here”, which is a state the card shows rather than an error. The two cards fail apart: the place’s own clock zone arrives with the first response, the sun and moon times are arithmetic on that zone and owe the forecast nothing, so a forecast that fails after the zone has arrived takes only the forecast card with it and leaves tonight’s times standing. Reached automatically for a county or a town; for a selected named place, only when you ask, with that card’s own Before you go button — never on opening the card.
• NOAA National Weather Service alerts — active watches, warnings and advisories, the products in force at the selected place, on a card of their own. It is deliberately not part of the forecast card above: a forecast is a number for an hour, an alert is a product an office issued with a window of its own, and a failure in either leaves the other standing. Each alert shows its own issue and end times, stamped in the selected place’s time zone rather than yours, and links to the Weather Service’s record for that alert — the web field every single alert carries is the agency front page, identical on all of them, so it is never offered as detail about one. Nine alerts in ten carry no drawn outline at all: 342 of the 386 in force when this was measured were zone-based, so an alert names the zones it applies to instead of a shape, and a missing outline is the usual case rather than an empty answer. An alert whose own end time has passed while the page held the response stops being in effect and says so; one the office cancelled is withdrawn whatever expiry it still states; and a service that fails says it could not be reached, which is never rendered as “nothing in force”. Most of all, a quiet card means no National Weather Service product is in force here — not that the ground is free of danger. These are the hazards this one agency issues products for, which is not every hazard that exists. Reached automatically for a county or a town; for a selected named place, only when you ask, with that card’s own Before you go button — never on opening the card.
• GBIF wildlife records — GBIF occurrence search, counts of times somebody saw a living thing inside the area you are looking at and wrote it down, as an optional screen surface. It is asked as a single bounded count per area — limit=0 with facets, about 3.5 KB — never as a list of records, and that is not a performance choice: coordinateUncertaintyInMeters was present on 4,043 of 133,925 records (3.0%) in the cell this was measured against, so the precision of a record is usually unknown and nothing is drawn as a dot. The area is shaded and no location inside it is claimed. Only public-domain (CC0) and Creative Commons Attribution records are asked for; that filter is a complete whitelist over GBIF’s closed five-value licence enum, it is enforced by the server, and the row says how many records in the area it removed. Each contributing dataset is credited by its own published citation, resolved from the record’s dataset key — the dataset name field is absent on the largest publisher, so crediting from it would silently leave that publisher out. The count is an upper bound: two publishers can submit the same sighting and a count of records cannot tell them apart. Records whose location was deliberately widened — to protect a threatened species, or at the recorder’s request — arrive stating about 28 kilometres of uncertainty; they are counted and never placed. Most of all, this is a record of effort: two cells the same size measured 133,925 records and 827, a hundred-and-sixty-fold difference in recording and not in wildlife. An empty area means no records in this selection — not no animals, not poor habitat, and often just that nobody walked there with a phone. Nothing here says a species is present now, says where in the area it was, or says you will see it. No photographs are loaded: media carry their own licences, separately from the record holding them. GBIF data user terms.
• NOAA National Weather Service radar — MRMS base reflectivity, the national radar mosaic, as an optional screen overlay. It answers what is falling now, which the hourly card does not: an hourly number belongs to a grid cell and an hour, and a shower crossing one ridge is neither. What radar sees is what the beam met aloft, not what reached the ground — a beam clears an intervening ridge, and a light return can be dry air or birds. The service keeps a rolling window of about two hours and nothing older, so the layer always states the time the frame it is showing was valid and how long ago that was; past twenty minutes it says the picture is too old to act on rather than drawing it as current, and a service that does not state a time is reported as unknown. It is drawn on screen only and never on a printed sheet, because a two-hour window is out of date before the ink dries.
• NOAA aurora forecast — SWPC’s OVATION model, an optional screen overlay for the chance of aurora overhead. It is a forecast, and the payload says so itself: the observation time and the forecast time it carries measured sixty-six minutes apart, so the layer names the forecast time rather than describing the sky as it is now. It models the chance of aurora overhead and says nothing about cloud, so a high number and a clear night are two separate questions. Over the lower 48 it reads zero on almost every night — all 1,560 cells covering it were zero when the source was measured, with the first non-zero values near 65N — so “nothing forecast here” is the ordinary answer rather than a gap in the data. The whole-globe file is about 900 KB, so it is fetched only when the layer is switched on.
• EPA AirNow air quality — AirNow latest-hour AQI contours, the air quality category EPA published for the most recent hour it had. What the records hold is the whole of what the layer can say: a category from Good to Hazardous, and the hour it belongs to. There is no pollutant field and no index value in them at all, so the layer names the category and stops — it never names a pollutant and never states a number, because either would be invented rather than read. It is an observation of a past hour and never an outlook: every reading carries the hour it covers and how long ago that was, and one older than four hours says it is too old to act on rather than being shown as current. The contour surface is interpolated between monitoring stations, so ground with no monitor near it still takes a colour from what was measured elsewhere, and EPA publishes these hourly figures as preliminary rather than as the verified official record. United States only — 115 contours covered the whole country when it was measured, all sharing one hour, which ran 144 minutes behind the clock.
• NOAA smoke plume analysis — NESDIS Satellite Analysis Branch HMS smoke, an optional screen surface showing where an analyst outlined smoke in that day’s satellite images. It is smoke aloft — a column of air seen from above, not the air at head height — and it is deliberately a separate question from the air quality contours above, which measure a different thing at a different height and disagreed with it over the same ground when both were measured. The three classes are how opaque the smoke looks from orbit; NOAA says they roughly approximate 0–10, 10–21 and 21–32 micrograms per cubic metre through the column, and states plainly that no guarantee is made about the real density, so the layer reports them as approximations and never as a reading. Each plume carries the satellite image window it was drawn from, in UTC, and the row leads with that window rather than with the class. It is a daylight product: nothing is analysed at night, the first plumes of a day appear around local noon, and an empty map before then means the day has not been analysed rather than that the sky is clear — which the layer asks the service directly rather than assuming. North America only, screen only. NOAA publishes the same analysis as a KML file on its own file host, which serves no cross-origin header at all and cannot be read by a browser; this service is the same day’s analysis, feature for feature, and needs no archive opened and no KML parsed.
• NIFC wildfire perimeters — WFIGS current interagency perimeters, an optional screen surface showing boundaries mapped for active incidents. “Current” is the service’s name for itself, not a promise about any shape in it: each boundary was mapped at some earlier moment, and the median measured three weeks old with one in ten past two months — so the row states the age of every perimeter it draws instead of one “as of” for the layer, and a boundary whose mapping date the record does not state says so rather than being drawn as fresh. One record in the national set is dated the year 0926, which naive arithmetic would render as 401,770 days old; anything before 2000 is read as unknown. Prescribed burns are filtered out — planned, deliberate fire is not wildfire, and drawing one as the other would be a factual error — and incidents reported fully contained remain in the service, so a perimeter is not proof of active fire. The national set is 37 MB unsimplified and did not transfer at all when it was measured, so the shapes are always requested simplified to what the screen can resolve. NIFC states these data are dynamic and are not legal documents: nothing here is a closure, an evacuation order, or a statement about any road.
• NASA Black Marble night lights — GIBS WMTS tiles, the optional light-emissions context surface on the screen. It is a picture of light sent UPWARD from the ground, taken from orbit and averaged over a year, and the newest composite this service publishes is 2016 — so the year is part of the layer’s own name rather than a footnote. Two things it deliberately does not do. It never prints a Bortle class, a 1-9 rating or any score: upward radiance measured by a satellite and what a human eye sees standing underneath are different quantities, and mapping one onto the other would not be a unit conversion. And it never claims to describe tonight. Its finest detail is zoom 8, which is where the tile set stops — closer in, Leaflet enlarges the same picture and the layer row says so, because an enlarged tile looks exactly like real detail.
• BTS transportation noise — National Transportation Noise Map 2020 tiles, an optional screen surface for the lower 48 states. It is a model: the Bureau of Transportation Statistics’ estimate of aircraft, road and rail noise on an average day of 2020, as a 24-hour A-weighted level, and BTS says it should not be used to judge noise at an individual place or time. So it reads as a rough picture of where transport noise is a factor, never as the level at your campsite, and the map prints no decibel figure for a spot. Its own legend runs from 45 to over 90 decibels in bands; this map does not redraw those bands. Alaska and Hawaii are separate services and are not shown. Detail stops at zoom 12, coarser than the 1:24,000 print scale; closer in, the same picture is enlarged and the layer row says so.
• AWS Terrain Tiles (Terrarium DEM) — Mapzen/Tilezen elevation tiles on the AWS Open Data registry, the elevation the optional 3D terrain panel is shaped from. Fetched only after you press “3D”, never on page load. It is a mixed-source set (3DEP, SRTM, GMTED2010, ETOPO1 and others) with native tiles through level 15, so the Source decision record below says which source answers where.
• State land services — game lands, state parks, other state-managed areas and county parcel boundaries, one service per state, measured and recorded in state-sources.json, the one catalogue the map reads. Boundaries only; no owner or occupancy information is ever requested.
• Finding cameras — cameras show only where a camera list covers. What is shipped: road cameras (Caltrans in all twelve California districts, and state and city agencies in Kentucky, Texas, Kansas, Oregon, Utah, Minnesota, Indiana, Washington (Spokane, Kirkland and the state ferry terminals), Virginia, South Dakota, Louisiana and Colorado, under one Road cameras (state DOTs and cities) switch covering the US states); National Park Service park webcams nationwide (every webcam with a position, one list per state, read by map area); and HPWREN (California) and AlertWest (western states and Hawaii) cameras as links only, with no picture. Open Layers and map key and the Cameras block is the first thing under the search, ahead of every group, naming the five camera layers, each with its coverage and its status on one line beside its Go to. Go to beside a layer fits the map to it (road cameras open on California, park cameras at Great Smoky Mountains, camera links on the West, research and airport cameras on the lower 48 states) and opens the Cameras switches; it turns nothing on and requests nothing. Pins appear from zoom 7: zoomed out further, a camera layer reads Zoom in to see cameras, says how many cameras sit in the areas the view touches, and loads nothing; a view holding more than 300 cameras draws the 300 nearest its centre and says so; and a view with no cameras in reach says how far the nearest ones are. Each region shows its own status, and the states read apart: Layer off, Loading the camera list, Camera list unavailable (no area the view touches could be read; when only some could not, the report says how many areas did not load and the rest stay drawn), Outside coverage (the view never reaches the region), Zoom in (the region is in view but too wide to draw), None in this view (the region is in view but holds no camera there) and N cameras in this view. Searching Layers for camera, webcam, Caltrans, park, Tennessee, HPWREN, AlertWest or California finds the rows.
• Caltrans CWWP2 road cameras — CCTV status feeds, one per district (District 5’s is linked), an optional screen-only surface. The camera list is a snapshot of all twelve districts this site builds ahead of time, so turning the layer on asks only this site, one small file per area the map shows, never Caltrans itself; each camera’s card says its health is as of the last list check. It is never printed and never part of a PDF. A picture loads only when you open a camera, and its age is read from the image file’s own Last-Modified header — when Caltrans’s server last changed the file, never a guaranteed capture time — because the feed publishes no per-image capture time at all. A record Caltrans marks out of service is still shown, never hidden, and links to the picture and to Caltrans’s own documentation page rather than fetching anything. A camera’s card can be saved with Save camera: the saved place holds the name, the operator and the position only, never the picture or stream address, and Show on it switches this layer on and opens the card without a second request for the list.
• NPS park cameras (nationwide) — park webcam images, an optional screen-only surface. It is never printed and never part of a PDF. The camera list is a snapshot this site builds ahead of time, not a live request to the National Park Service, so opening the layer asks only this site; opening a camera then loads that camera’s current picture from the Park Service’s own host. The Park Service publishes no per-image capture time at all, so a picture’s age is never shown, only when the camera list was last checked. A camera the Park Service marks out of service is still shown, never hidden. The build snapshots every National Park Service webcam that has a position into a per-state list, and the map reads the lists by map area like the Caltrans ones. A source checkout without an NPS key ships only North Carolina’s list (Great Smoky Mountains); the deployed site has every state’s. Saved park cameras behave as Caltrans ones do. A camera that publishes no position is never drawn, so it has no card to save from; a saved camera that somehow carries no position (an imported list, say) cannot be shown on the map or checked for conditions, and its row says so.
• State and city road cameras (ArcGIS layers) — seven operators that publish their cameras as a public ArcGIS REST layer: the Kentucky Transportation Cabinet’s, the City of Arlington, Texas’s, the City of Lawrence, Kansas’s, the City of Salem, Oregon’s, Oregon DOT’s TripCheck, Utah DOT’s UDOT Traffic and the City of Kirkland’s. One generic reader serves all of them, asking each layer for its named fields only, one page at a time with a pause between pages, as a plain request with this site’s own user agent and nothing else; a source that answers only to a borrowed origin, a scraped token or a schema probe is left out, and so is a layer with no picture address, no position or only stale pictures. A camera is kept only when its operator marks it in service, it has a position inside the state’s own box (not padded; TripCheck also drops a camera whose name puts it in a neighbouring state, such as “I-5 Washington - Vancouver at Mill Plain” or “Westport - WA”), and its picture is an https address on the one host recorded for that operator (a second host, a plain http address or a doubled slash in the path drops the row). Each picture is checked when the list is built: most answered as an image to a plain request and were current by their Last-Modified header (the test used, because a layer’s own timestamp field can be months old while its pictures are minutes old), but a few per layer were missing, failing or months old, including some stand-in images, so a picture that answers 404, 410 or a server error, was last modified over seven days before the seed list was last refreshed, or has the recorded size of an operator’s stand-in logo (UDOT’s re-dated one), is left out. Those checks send one request per picture, so they run only when the seed lists are refreshed by hand; a site build leaves these lists as committed. A picture can still go stale between builds. Washington State Ferries terminal cameras come from WSDOT’s own ferry-map camera list (Cameras.ashx, a plain request with this site’s user agent); only the terminal pictures under /wsf/ are kept, so nothing overlaps a road-camera feed, each picture is checked before it is saved when the seeds are refreshed, and the cards link the WSF vessel watch. WSDOT’s travel information disclaimer has no copying ban, but WSDOT’s web privacy notice (Section H) says “we recommend that you contact us for permission to use any content contained on this site”, a recommendation and not a ban, so the usage terms stay unverified and WSDOT is credited with a link to its own page. Not used: Lexington, Kentucky (its terms of use say users may not modify, reuse, copy or distribute city website content without the express written permission of the City/LFUCG), King County, Washington (its terms of use say “You may not publish, display, distribute or commercially exploit any of the the website or app or the content therein without the prior written permission of King County”), Iowa DOT (its picture addresses are dated per-snapshot paths that go stale once saved) and Louisiana DOTD (video streams and 511 page links only). Each card links the operator’s public camera map. The recorded picture hosts are https://www.trimarc.org/images/milestone/, https://webapps.arlingtontx.gov/webcams/, https://assets.lawrenceks.gov/gis/traffic/, https://egov.cityofsalem.net/traffic-cameras/cameras/, https://tripcheck.com/RoadCams/cams/, https://udottraffic.utah.gov/1_devices/, https://its.kirklandwa.gov/CamView/, and for the ferry terminals https://images.wsdot.wa.gov/wsf/ (one exact host each, never a wildcard). Credit is the operator’s name, the usage terms are unverified, and no picture age is shown. Like the National Park Service list, these lists are snapshots this site builds ahead of time (one file per state), and the map reads them through the Road cameras (state DOTs and cities) switch, area by area like the Caltrans and NPS lists. Opening one of these cameras shows a link-only card (name, operator and a link to the operator’s camera page) and makes no request: the map does not load these pictures today.
• Statewide road cameras (CARS lists) and Spokane’s SRTMC — four state operators publish their cameras as a keyless public list: the Minnesota DOT’s (511MN), the Indiana DOT’s (511IN), the Kansas DOT’s (KanDrive), and the Colorado DOT’s (COtrip), and the Spokane Regional Transportation Management Center publishes Spokane County’s through its map query. A normal site build never asks any of them: it serves the committed lists, which were refreshed under --refresh-seeds and are the only ones whose pictures were checked. A refresh makes one plain request per list with this site’s own user agent and nothing else, no key. A camera is kept only when it is the credited operator’s own (Kansas keeps only the rows its list gives to Kansas DOT and drops those of the City of Topeka, the Kansas Turnpike Authority and KC Scout; Colorado keeps only those given to Colorado DOT or CDOT Cellular; Spokane keeps only SRTMC’s own pictures, not the WSDOT ones its map also shows), its list marks it public and does not mark it inactive, it has a position inside the state, and its picture is an https address on a host recorded for that operator: https://public.carsprogram.org/cameras/ (Minnesota and Indiana), https://kscam.carsprogram.org/snapshots/ (Kansas), https://cocam.carsprogram.org/snapshots/ (Colorado) and https://images.srtmc.org/ (Spokane’s own); the video address of a camera is never kept. A picture the list stamps as over a day old is dropped, and a refresh also asks for every picture’s headers (four at a time, with a pause between batches): a refused or missing picture, the operator’s offline placeholder (matched on size and etag together) and one whose Last-Modified is over a day old are left out. Colorado’s list is thinned to one camera per 0.025 degree grid cell (about 2.8 km, the lowest camera number winning) to keep the committed lists small. The usage pages of Minnesota, Indiana, Kansas, Colorado (COtrip, https://www.cotrip.org/) and SRTMC were read before these were added; each is a disclaimer with no clause against showing a picture. Iowa DOT’s terms forbid copying and republishing without its written consent, and New York’s 511NY and NYSDOT pages hold no terms text (the ny.gov disclaimer covers only links leaving the portal), so neither of those is included. Credit is the operator’s name, the usage terms are unverified, and no picture age is shown. Like the other lists, these are snapshots this site builds ahead of time (one file per state); the map reads them through the Road cameras switch, area by area, and opening one of these cameras shows the neutral card with its name, the operator and a link to the operator’s camera page, without loading a picture.
• Iteris and Socrata road cameras (VA, SD, TX, LA) — four more operators, read by two generic readers. Two publish a public camera list as GeoJSON on an Iteris content host: the Virginia Department of Transportation (VDOT 511, pictures from https://snapshot.vdotcameras.com/) and the South Dakota Department of Transportation (SDDOT 511, pictures from https://sd.cdn.iteris-atis.com/camera_images/; its camera list and its road-weather station list are both read, one pin per site, with the first still on the recorded host on the card). Two publish an open-data table through a Socrata API: the City of Austin, Texas (b4k4-adkb, pictures from https://cctv.austinmobility.io/image/) and the City of Baton Rouge, Louisiana (6z6u-ts44, pictures from https://511la.org/map/Cctv/; the table has no name column, so a camera is named by its code). Montana and South Carolina are not included: Montana’s terms forbid redistribution without written permission, and the SCDOT Disclaimer (https://www.511sc.org/static/disclaimer.html, read 2026-10-04) says its content may not be published, transmitted or displayed without express written permission. A normal site build never asks any of them: it serves the committed lists, which only a --refresh-seeds run fetches again. Every refresh request is a plain one with this site’s own user agent: no borrowed origin, no token, no GraphQL. Virginia keeps cameras the list marks active and Austin keeps only those marked TURNED_ON; the South Dakota and Baton Rouge lists carry no status, so every located camera with a picture on the recorded host is kept. Every kept camera has a position inside the state and a picture address on the one recorded host; the list a build could not fetch keeps the last committed file. Credit is the operator’s name (for Austin, the department as well) and a link goes to the operator’s own page. Terms pages read on 2026-10-04: no terms page was found for Virginia (its 511 site is an app shell and links only an SMS terms-of-service page; the vdot.virginia.gov footer reads only “© Copyright 2026, VDOT. All rights reserved”), so Virginia’s usage terms stay unverified and it is credited by name with a link to VDOT. South Dakota’s Disclaimer (https://www.sd.gov/kb_view.do?sysparm_article=KB0010221) says “Anyone may view, copy or distribute information found on the State’s website for personal or informational use without owing an obligation to the State”, which covers a free informational map like this one. For Austin, the pictures’ own operator site (https://austinmobility.io/, Austin Transportation and Public Works) links its Disclaimer to the City’s Open Data Terms of Use (https://datahub.austintexas.gov/stories/s/City-of-Austin-Open-Data-Terms-of-Use/ranj-cccq/), which says data “are offered free and without restriction” and that content created by City employees within the scope of their employment “is not subject to copyright protection”, and asks for credit “to both the City of Austin and the City Department that is the source of the cited data”, so the credit reads City of Austin Transportation and Public Works. The Baton Rouge/LADOTD disclaimer page has no copying clause and its dataset metadata says Public Domain (not relied on), so its usage terms stay unverified. Like the lists above, these are snapshots this site builds ahead of time, one file per state; the map reads them through the Road cameras switch, area by area, and opening one of these cameras shows the neutral card with its name, the operator and a link to the operator’s camera page, without loading a picture.
• HPWREN and AlertWest cameras (links only) — HPWREN’s camera list (UC San Diego; one pin per site, listing that site’s own working cameras, credited “HPWREN, UC San Diego”) and AlertWest’s camera list (one pin per camera outside California, credited “AlertWest”; California’s cameras belong to ALERTCalifornia and are left out, as are HPWREN sites’ cameras that other networks own). These two are links only: a pin shows the camera’s name, the credit and a “View on” link to the owner’s own page, and no picture, because HPWREN allows its pictures only for non-commercial use with credit and AlertWest licenses its pictures as CC BY-NC-ND, and this site chose not to show them. The site keeps no picture address for them, your browser loads nothing from either owner unless you follow the link, and the page’s security policy lists neither host. Like the National Park Service list, these lists are snapshots this site builds ahead of time, and the map reads them by map area under the Camera links (HPWREN, AlertWest) switch, loading nothing from either owner. The card’s link reads View on HPWREN or View on AlertWest.
• State road cameras (links only) — five operators whose public camera lists are read for pins and links only: Missouri DOT’s MoDOT Traveler Information, Maryland CHART’s, Delaware DOT’s, New Mexico DOT’s NMRoads and the New Jersey Turnpike Authority’s camera list. Each is read with a plain request carrying this site’s own user agent and nothing else. On 2026-10-04 each operator’s robots.txt was fetched, and the terms page was read where one is linked: the Turnpike Authority’s disclaimer, the State of Maryland’s and State of Delaware’s privacy notices (linked from CHART and DelDOT) and MoDOT’s privacy policy (linked from modot.org; no terms page is linked from traveler.modot.org) contain no ban on this use, and no terms page is linked from NMRoads. West Virginia DOT’s WV511 is left out because its disclaimer forbids republishing its material. A pin shows the camera’s name, the operator’s credit, a “View on” link to the operator’s public camera page (or its camera map), and no picture; the site keeps no picture or stream address for them, the usage terms are unverified, and the page’s security policy lists none of these hosts. A pin outside its own state, or within 30 m of another pin from the same operator, is dropped (an in-service pin is kept over an out-of-service one). Like the National Park Service list, these lists are snapshots this site builds ahead of time (one file per state, committed here and fetched again only by a --refresh-seeds run, so a normal site build never asks these operators). The map reads them through the Camera links switch, area by area; opening one shows the neutral card (name, credit and the owner’s link) without loading a picture.
• PhenoCam Network research cameras — camera list, a screen-only surface credited “PhenoCam Network (CC BY)”. Only a --refresh-seeds run reads the list (a normal site build serves the committed file and asks the network nothing); it keeps the United States cameras the network marks active and that sent a picture in the previous 30 days, as one snapshot file, which the map reads through the Research cameras (PhenoCam) switch. A camera’s card links to the network’s own page for the camera and its picture is the network’s latest image, hotlinked from phenocam.nau.edu, never stored here. The list’s contact names and addresses are never copied. A failed refresh keeps the last good file.
• AviationWX airport cameras — airport list, a screen-only surface credited with the service’s own line, “Weather data from AviationWX.org”. Only a --refresh-seeds run reads the airport list (never a picture; a normal site build asks the service nothing), and it keeps one pin per webcam at each listed United States airport, as one snapshot file, which the map reads through the Airport cameras (AviationWX) switch. A pin’s picture is the service’s 720-pixel image, hotlinked from api.aviationwx.org, never stored, and its card links to the airport’s own AviationWX page. A failed refresh keeps the last good file.
• USGS VolcView and AVO volcano cameras — VolcView ashcam list with the Alaska Volcano Observatory list, a screen-only surface credited per camera to its owner (“Alaska Volcano Observatory”, “USGS Hawaiian Volcano Observatory”, “FAA weather cameras”, “National Park Service”, “USGS Volcano Hazards Program”, each “via USGS VolcView” except “USGS Volcano Hazards Program”). Only a --refresh-seeds run reads the two lists (a normal build asks USGS and AVO nothing), with a plain GET that sends the build’s own User-Agent, never a browser’s, an Origin or a Referer, with a pause between requests. The AVO list names each camera’s organization: a known one (AVO, HVO, YVO, FAA) is credited, any other named owner (UAF, DGGS, WSDOT, private) drops the camera; a camera it does not name is credited by the page it links to, and one that links to a private or state owner (a ski resort, WSDOT) is dropped. A camera is kept once per webcamCode (the AVO list is a subset of VolcView’s: 396 of 460 on 2026-10-05, none only on AVO), and one with the same name within 50 m of another counts as the same camera. A camera with no picture, no position (the feed uses 0,0), a position outside US territory (the 50 states, DC, Alaska west of the dateline and the territories), or a newest picture over 24 hours old (or stamped more than 10 minutes in the future) is dropped, and every kept still is also checked with a HEAD request, at most 4 at a time (404, 403, a non-image or a Last-Modified over 24 hours drops it). Rights are unverified, credit plus a link to the camera’s VolcView page. USGS copyrights and credits (read 2026-10-05): “USGS-authored or produced data and information are considered to be in the U.S. Public Domain. While the content of most USGS websites is in the U.S. Public Domain, not all information, illustrations, or photographs on our site are. Some non-USGS photographs, images, and/or graphics that appear on USGS websites are used by the USGS with permission from the copyright holder.” The VolcView host publishes no robots.txt (404) and answers its API with 200. avo.alaska.edu has no terms or disclaimer page (its about page carries no copyright text and /disclaimer is 404) and its robots.txt is an HTML 200 page with no rules, so decision D1 applies to its credit: rights unverified, credit and link to the owner; its list is read for organization only. Pictures are hotlinked from volcview.wr.usgs.gov, never stored; the cameras show today under the Research cameras (PhenoCam) switch, which asks no host for anything when turned on; opening a pin loads its picture from volcview.wr.usgs.gov (avo.alaska.edu is asked only by the build, for its list). Cameras west of the dateline are written to their own files, the western Aleutians to volcview/ak-west.json and any in Guam or the Northern Mariana Islands to volcview/gu-mp.json, so no index file box crosses the antimeridian. In the Alaska box, east of -141, a point is kept only southwest of the panhandle border (Atlin, Stewart, Mount Edziza and Dease Lake, in British Columbia, are dropped). A failed refresh keeps the last good file.
• NOAA GLERL Great Lakes cameras — camera page, a screen-only surface credited “NOAA Great Lakes Environmental Research Laboratory (GLERL)”. Only a --refresh-seeds run reads the page and HEADs each still (4 at a time, with a pause; a plain GET and HEAD with the build’s own User-Agent), keeping a camera whose picture is under 24 hours old (the Muskegon cameras last updated in 2013 drop out). The station list and positions were copied from GLERL’s /metdata/ map on 2026-10-05 and are not re-read on a refresh, so a station GLERL adds later is skipped until the table is updated. Rights are unverified, credit plus a link to the station’s page. GLERL’s site notice (read 2026-10-05): “Information presented on these pages is considered public information and may be distributed or copied. Use of appropriate byline/photo/image credits is requested.” robots.txt disallows only /incoming/ and a few file types, not /metdata/. Pictures are hotlinked from www.glerl.noaa.gov, never stored; the cameras show today under the Research cameras (PhenoCam) switch, which asks no host for anything when turned on, and opening a pin loads its picture from www.glerl.noaa.gov.
• NOAA NDBC buoy cameras — not included. The station list (activestations.xml) answers, but the picture host www.ndbc.noaa.gov/buoycam answers 403 from its CDN to this build’s honest User-Agent. The build does not imitate a browser, so NDBC buoy cameras are dropped until NOAA allows the build agent or a public API (a follow-up).

Every source a set is built from is credited by name on its printed map key, derived from what the build actually consumed rather than written by hand.

Tap a recreation site and it also says how far the nearest unpaved road is — “0.7 mi from a mapped crushed aggregate or gravel road (straight-line)”. Three things that phrase is careful about: mapped means the roads fetched for the area you are looking at and no others; straight-line means exactly that, because nothing here calculates a route, a legal right of way or a driving time; and a road whose surface the inventory never stated is not counted as unpaved, so a site with nothing to say says nothing rather than promising gravel nobody looked up.

A road’s rating is the Forest Service’s record, not a report of current condition: a road can wash out, gate or be logged over without the inventory changing. Where no record matched the route, the road is drawn at its plain width and says nothing extra — an empty rating would read as “the inventory looked and found it ordinary”. Where two records disagree about the same route, the map says they disagree rather than picking one.

Honesty rules. MVUM is a designation dataset published forest by forest, not a live passability feed, and not every forest is included. Seasonal dates are shown only when the source supplies them: on screen when you select a route, and on the printed key under Seasonal routes, one line per route that reached the sheet. Routes the source called seasonal and gave no dates for are counted there too — listing only the dated ones would read as “these are the seasonal roads here”, and you would drive an unlisted one in February on the strength of it. A source that fails to load never disappears silently: it blocks the PDF, or you remove that layer explicitly. Each source prints its own retrieval or dataset date; there is no single composite “as of” date.

Source decision record

Which states are covered

50 of 51 states publish at least one land layer this map can read, measured 2026-09-06. Federal layers work everywhere, so a state with none here still gets a map.

Limits

Contiguous United States only. Letter paper at 1:24,000 or 1:48,000. PDF creation targets desktop Chromium and Edge first. This is not a navigation or safety authority: confirm current conditions with the managing forest or land office before you go. The 3D terrain panel needs WebGL; a browser with it turned off sees the same 2D map with a message in place of the panel, never a silent failure.

Keyboard and large text: Tab reaches the dock panels, the map controls and the place pins, and the pin you are on is never left hidden behind the map key or the zoom buttons. At 200% text the panel titles wrap onto more lines instead of being cut off, and at a 320 px width only the map itself scrolls sideways. The grey squares offered around a print area (“Extend print area here”) are for a pointer and are skipped by Tab.

Troubleshooting

The sheets do not measure true to scale. The print dialog scaled them. Choose Actual size (100%) and turn off Fit to page, Shrink oversized pages and any borderless option. A test sheet’s scale bar is the check: measure it against a ruler before printing the whole set.

“Check sources before building.” Nothing has been asked of the map services yet. Check sources first; the page will not build a PDF from sources whose state it does not know. Three kinds of layer are never among those sources, so switching one on never holds up the build: a screen-only layer, such as a context surface, which never reaches the PDF; an extra sheet (aerial, shaded relief, route summary), which the build fetches for itself and prints the map without if the service does not answer; and a state land layer that no state in your area publishes, whose switch turns off once Check sources has found which states the area covers. The line under Build PDF on this device names only the sources it is waiting on — the base map and each overlay you have switched on — so a layer that is off never shows up there as “not checked yet”.

Layers do not appear when I switch them on. Switch the layer on and stop moving the map: once the view settles the page asks for what is on screen and draws it. Two views ask for nothing, on purpose — zoomed out past the level the topo base appears at, and a view covering more ground than one query can honestly answer. Both say so on the page. Building the PDF always re-asks for the printed box itself, so a sheet never draws features fetched for somewhere else.

Which context surface am I looking at, and when is it from? Whichever one is chosen under Land, sky & light context — only one can be on at a time — and the map itself carries a line naming it and the time its data refers to. That time is the source’s own and never the moment it was fetched: the aurora surface shows the time it is a forecast for, radar and air quality show the time they were observed, and the night lights show their vintage (a 2016 composite, not tonight). The wildlife records are an aggregate over a span of years rather than a moment — a cell holds records from many different days — so that line says the time is unknown and the row beneath it names the year range the count covers. The steepness surface is arithmetic rather than a reading, and the national elevation model states no date for a given piece of ground, so its line says the time is unknown instead of borrowing the clock. The county ranking names no service and so states its own dashboard vintage rather than a fetch time, the same wording an inventory carries. A source that will not state its time says so on that line rather than being shown as current.

Can I turn the base map off, or see the aerial or relief on screen? Yes — under Base map in the Layers panel, choose USGS Topo, aerial imagery, shaded relief or no base map at all; the choice is for the screen only, and the tiles are drawn at full strength rather than dimmed. The PDF still prints the topo base on every sheet, with aerial imagery and shaded relief available as companion sheets under Extra sheets.

“USGS Topo must load completely before export.” At least one base-map tile did not come back. The build refuses rather than printing a sheet with a blank patch. Try again — the tile service rate-limits bursts — and if it persists, reduce the area or use 1:48,000, which needs fewer tiles.

An MVUM layer that reads unavailable, or a coverage that will not reach complete. The Forest Service service did not answer, or answered partially. MVUM is published forest by forest and not every forest is in it, so a genuinely empty answer for an area outside any published forest is not an error. If you want the base map regardless, turn the MVUM layer off explicitly and build without it — the page will not quietly drop a layer you asked for.

A layer says “partial” instead of “complete.” The map services answer a large rectangle one page at a time, and say in the answer whether more is waiting. A layer reads “partial” when the service still had more than it handed over, so what the page holds is a piece of that layer, not all of it. The build refuses on a partial layer for the same reason it refuses a missing tile: a truncated set is worse than none, because on paper it looks finished. Draw a smaller box and check sources again. Recreation sites and trails used to report themselves ready in exactly this situation — the sheet printed with trails missing and nothing said so.

“Named terrain (GNIS) must load completely before export.” The named-terrain query for your area came back partial, over the record cap, or timed out. The build refuses rather than print a set of place names with unstated gaps. Try again, draw a smaller box, or switch the names layer off explicitly to build without it.

The build is slow, or the tab runs out of memory. Every tile and every sheet is rendered in your browser, so a large area at 300 dpi is real work. The page shows an estimate before you start. If it is uncomfortable, cut the area, drop to 200 dpi, or use 1:48,000.

Nothing downloads at the end. PDF creation targets desktop Chromium and Edge first. A browser that blocks large blob downloads, or a private window with storage disabled, can stop the file being handed over.

The area came back slightly different after a reload. Recipes saved before September 2026 stored only the projected sheet extent, so restoring one rebuilds the box from the sheet grid and returns a slightly larger area. Redraw the box once and it will save and restore exactly from then on.

I removed a quadrangle and the page count did not change. A sheet prints wherever any selected quadrangle reaches it, so a quadrangle whose ground sits on sheets its neighbours also need saves nothing. The list says so before you remove it.

I pressed Clear print area and now there is no area after a reload. That is expected — Clear print area removes the saved recipe along with the area (and a shared link’s area settings from the address bar), so there is nothing left to restore, even if you reload the instant after pressing it. Undo brings back exactly what Clear removed, but only until you leave the page; after that, draw a new area.