Custom Data Solutions

Custom work starts where the catalog stops: a class we don't publish, a year that has already passed, or a deliverable that isn't a standard data file. The projects below range from a single distribution line buffered a hundred feet on either side, to two years of land cover across an entire state.

Anywhere

in the US - a single parcel to nationwide

Any year

that has passed - we're not clairvoyant (yet)

Anything

that the human eye can see on a map

Six ways a project goes custom

1. A class we do not publish.

Seven land cover classes answer most questions. Yours might need an eighth. We have built classifiers for silvicultural harvest, and for managed farmland as distinct from unmanaged ground. Features you already label by hand can become training data.

2. A year that has already passed.

Change needs two dates. Our models run against older imagery, and for most of the country we hold archives back to 2005, so you can have several snapshots of the same ground built to one specification.

3. Imagery and resolution you specify.

For a single area of interest we can use a newer vintage the moment a state releases it, leaf-off imagery for a cleaner read beneath canopy, at the highest available resolution. Or use licensed commercial imagery where data freshness justifies the cost.

4. Geometry cut to your boundaries.

Your area of interest can be a corridor buffered along a line, a watershed, or every parcel in a jurisdiction. The same data can arrive seamless, tiled, or clipped to each boundary as a separate file.

5. Statistics summarized to your geographies.

Most teams want numbers more than polygons. We tabulate against boundaries you define, from a single parcel to a statewide set of assessment areas, with change included where more than one year exists.

6. Finished cartography.

Some projects end in a map rather than a dataset. We produce map series at volume, with legends, index sheets and calculated areas, to a template you approve first.

Selected projects

Teams across sectors use EarthDefine data to answer spatial questions. Read about these projects to get a broad idea. Visit the Stories section from the menu above for more details.

A state forestry and mapping agency

A state can describe its forests from almost any dataset. Describing what has changed needs two dates built to one specification by the same models, so that the difference between them is the forest rather than the method.

They had a current canopy assessment and needed the land cover underneath it, for two years five years apart. They also needed ground that has been logged, which our seven class schema does not carry. We built a classifier for harvest and delivered it alongside the standard classes rather than folded into them.

The land cover went out as roughly 2,500 seamless GeoTIFFs across both years. Underneath that, the part their analysts open: vector summaries rolled up to the geographies the agency reports against, including 14,392 square miles of wildland urban interface, 40,769 square miles of state regions, Forest Inventory and Analysis units, planning district commissions, and every HUC12 watershed in the state. Numbers per boundary, without opening a raster.

If your mandate is written in terms of change, and the class it depends on does not exist yet, those are one project rather than two.

An engineering firm assessing vegetation risk for electric utilities

Knowing which trees touch a line is the easy problem. Knowing which are tall enough to reach one if they came down takes a height for every tree, the shape of its crown, and a bare earth surface, so that pole elevations and tree elevations can be compared honestly rather than as raw numbers from two different reference frames.

This firm came to us with a narrower question. Their utility client had already paid for a drone LiDAR flight over one distribution line, and they wanted to know whether our tree heights could stand in for the next one. The test ran along that corridor, buffered 100 feet either side. Ten months later their next area of interest covered most of a thousand square miles in a different state.

What surfaced along the way mattered more than the heights. Reviewing a land cover sample, their engineer told us something we had not considered. Impervious surface tells a utility what cannot burn and where a truck can work, and the tight edges of agricultural fields show where the ground is managed and watered. The space between those two is where the poles are. Nobody irrigates that strip and nobody mows it, so the vegetation grows taller and drier than anything around it, sitting directly beneath the conductors. Our seven class land cover called that strip herbaceous, and called the adjacent hayfield herbaceous too, because both are non-woody vegetation and that is what the class means. No schema we knew of drew the line where he needed it, so we built one that separates managed farmland, hayfields and pasture included, from everything else.

If the variable that drives your risk model falls between two classes instead of inside one, that gap is buildable.

A stormwater technology company

Telling a homeowner to manage their runoff only works if you can show them their own roof, their own driveway and their own trees.

This company had lot size and building footprint from a parcel provider. Everything else would have to be built: impervious surface split into buildings, roads and other surfaces, 2-foot contours, roof and driveway slope, tree points with heights joined to the parcel, and a per parcel table down to percent impervious cover and hydrologic soil group. They started with a trial on a short list of parcels, which is the sensible way to start. We quoted the full build across two municipalities, about 2.4 square miles and 5,000 parcels, and the quote is where the project stopped.

The month of questions that came with the trial was worth more than the trial was large. Packaging turned out to be harder than the data, because their front end could either mask one seamless layer per town or ingest a clipped file for every parcel, and since the source data is identical either way we priced both and offered to talk to their engineers. Precision was the second question, and a fair one, because square footage was going to appear on a screen in front of the person who owns the driveway.

Standard national imagery is flown leaf-on, so a driveway under a maple in July is not visible from above, and the fix is high resolution leaf-off imagery. The third question was how to get driveway square footage tied to the parcel identifier. We could not do that at the time. We can now.

A global aerial survey and mapping firm

When a utility settles crop damage with a farmer, the map is the conversation. It has to be legible to someone who has never opened GIS and defensible to someone who has farmed that ground for 30 years.

Across two construction seasons we produced 1,140 finished PDF maps covering eight wind projects and 520 square miles. Each sheet covered one PLSS section, a nominal square mile, for one landowner, with damage delineated by crop type, acreage calculated, and the easement identifier and landowner name on the page.

Built over our land cover and impervious layers and with a great deal of hand digitization on top, most of the effort went into judgment rather than classification. We agreed the rules on sample sheets before producing at volume: access roads and turbine pads come out of the damage total, slivers of roughly an acre or less stay in. The firm came back for the following season.

If what you owe someone is a finished map rather than a dataset, we produce those at volume as well.

A lake and pond management company

A company that manages small lakes and ponds needs one number quickly, which is: how much water is actually there?

They only wanted waterbodies mapped and had had no use for the other six land cover classes, and our nationwide product was still in development, so we priced a single state as a test run. New state imagery landed while we were scoping, which meant the delivery could come from imagery two years newer than the version we first quoted. It also meant retraining, more of it than we expected, and we missed our first date. When they asked whether we were still on track, we gave them the new date rather than shipping something weaker on the old one.

If you need one class out of seven and would rather not pay for the other six, that is a smaller and faster project, because each class runs as its own model.

Do you need canopy assessments for your project?

Much of our custom work includes urban tree canopy stats for cities, states, and nonprofits: from single campuses to nationwide. For repeatable projects like these, check out our Tree Map page.

How a custom project runs

Scoping call.
What you are trying to decide, and what the data has to support.
Written estimate.
Area, classes, attributes, imagery source and vintage, format, delivery and timeline.
Sample or paid trial.
A subset of the area, so you can test ingestion and we can confirm definitions before either of us is committed at scale.
Production.
Against the approved sample, with questions raised as they arise.
Delivery.
Your formats, seamless or tiled or clipped to your boundaries, by download or API. On larger projects we can split payment, half at the start and half once you have reviewed and accepted the data.

What helps us quote quickly

Location & boundaries

Your AOI and any boundaries you'd like statistics summarized to

Features & time period

Which features you need, any labeled data you already have, and the year(s)

Format & delivery preferences

Your preferred data format, coordinate system, delivery method, and timeline.

Your area of interest in any form that you prefer: (a) region as a shapefile, .kml, or geodatabase, (b) coordinates as latitude & longitude or (c) addresses
The features you need, defined in whatever way you define them, including the edge cases
Any labeled data or manual work you have already done
The year or years you need, if the current one is not the only one
The boundaries you want statistics summarized to
Format, coordinate system, and how the data reaches your system
Your timeline, and whether an early subset would help

Nonprofit and academic pricing applies to custom work as well as to our standard products.

Tell us what you need

Describe what you are trying to map and we will tell you whether it is a standard product, a custom build, or not possible yet. All three answers are useful.

Browse standard products