How to find vacant, developable land with GIS: a step-by-step method
"Where is the vacant land?" sounds like a question with one answer. In practice it is three questions: which parcels have nothing on them, which of those can actually be built on, and which of those suit what you want to build. A vacant parcel inside a floodway, or a strip of railroad right of way, is vacant and useless.
This guide walks through the method GIS analysts use to get from a county's full parcel file to a short, defensible list of developable sites, using real county layers as examples and finishing with a worked example from Nashville.
Step 1: decide what "vacant" means
Start with the county parcel layer (see where to find free GIS data for your city if you do not have it yet). There are three common ways to flag a parcel as vacant, and a good screen states which one it uses:
| Signal | How to test it | Where it goes wrong |
|---|---|---|
| Land use code | Filter the assessor's land use or property use field for vacant codes or descriptions | Codes describe the tax class, which can lag what is on the ground |
| Improvement value of zero | Building or improvement value equals zero, or total value equals land value | Small sheds, parking lots or recent construction may not match your idea of vacant |
| No building footprint | No building footprint intersects the parcel (city footprints or OpenStreetMap buildings) | Footprint layers can be out of date or incomplete in new subdivisions |
Counties that publish a vacant land layer
Some counties have already done this step, and their definitions are a useful model:
- Mecklenburg County, NC: a vacant land layer on Charlotte's GIS server starts from all tax parcels, removes golf courses, airport property, park property, submerged land, utility and railroad rights of way and cemeteries, and keeps only parcels with no building permits or building footprints. A count query in October 2026 returned 27,742 parcels.
- Cuyahoga County, OH: the county planning commission's residential vacant land inventory uses parcel geometry from the county fiscal parcel fabric "to identify suburban vacant residential lots and support evaluation of future residential development opportunities". It held 15,001 parcels in October 2026, each with municipal zoning, permit fields and a field recording why it was kept: 14,364 are vacant in the parcel fabric, and the rest were kept on the strength of a permit record or a county land bank (CCLRC) listing.
Step 2: remove land that cannot be built on
Next, take out land where building is prohibited, impractical or not the owner's call. Each constraint is an overlay against the vacant parcels:
| Constraint | Typical data | Notes |
|---|---|---|
| Floodplain | FEMA National Flood Hazard Layer | The 100-year floodplain is the Special Flood Hazard Area: zones A, AE, AH, AO, V, VE and their variants |
| Wetlands | National Wetlands Inventory or state wetland maps | Screening only: NWI uses a biological, not regulatory, definition |
| Steep slopes | Slope from a USGS 3DEP elevation model | Use the slope limit in your local code, and measure the share of each parcel above it |
| Parks and open space | City parks layer, OpenStreetMap parks | Also conservation easements where published |
| Rights of way and railroads | Parcel type or owner fields, OpenStreetMap rail | Often long thin parcels that pass an area filter |
| Cemeteries, golf courses, airports | Land use codes, OpenStreetMap | The exclusions Mecklenburg County applies |
| Water | Parcel layer, hydrography | Mecklenburg's exclusions include submerged land |
Drop the parcel, or erase the constrained part?
There are two ways to apply a constraint. Drop any parcel that touches the floodplain, which is conservative and simple. Or erase the floodplain from each parcel and keep the remaining net area, which keeps large parcels with a wet corner. If you erase, apply your minimum size test to the net area, not the original acreage.
Step 3: check zoning fit and minimum size
Join the city's zoning districts to the remaining parcels and keep those whose district allows the use you have in mind. Two practical details matter here:
- Codes vary by city. Zoning layers carry district codes (RS20, IR, AR2A) that mean nothing outside that city. Group them into general categories (residential, commercial, industrial, mixed use) before filtering. Cuyahoga's inventory does this with a generalized zone field next to each municipality's own code.
- Parcels cross district lines. A large parcel can touch two or three zoning districts, and a plain spatial join returns one row per district. Assign each parcel the district covering most of its area, or deduplicate, so each parcel is counted once.
Then apply a minimum size. Measure area yourself in a projected coordinate system rather than trusting an acreage field alone, and use net area if you erased constraints in Step 2. Shape matters too: a long, narrow parcel can pass an acreage test and still be unbuildable, so a minimum width or a compactness check is worth adding for small sites.
Step 4: look at ownership
Owner fields on the parcel layer turn a land screen into an acquisition list:
- Public land: filter owner names for the city, county, state, school district, housing authority or transit agency. For affordable housing, publicly owned vacant land is often the first list people ask for.
- Land banks: where a county land bank publishes its inventory, join it. Cuyahoga's vacant land layer already flags parcels listed by the county land bank.
- Assemblage: adjacent vacant parcels with the same owner can be dissolved into one site, which can turn several small lots into one that passes your size test.
- Recent sales: sale date and price fields, where present, show which parcels have just changed hands.
Step 5: check access to roads, transit and utilities
A developable parcel needs a way in and services to connect to. The usual layers:
- Roads and highways from OpenStreetMap or the state DOT, measured as distance to a highway or to the nearest road frontage.
- Transit stops from the agency's GTFS feed or the city portal, measured as straight-line distance or, better, a walking network service area.
- Water and sewer from the utility, where it publishes service area or main lines. This is the layer most often missing from public data, so treat it as a gap to fill with the utility rather than a reason to skip the criterion.
- Power for industrial sites: substations and transmission lines.
Step 6: rank what is left
After Steps 1 to 5 you have a set of parcels that pass every hard constraint. Ranking them is a separate step with separate logic: normalise each soft criterion (distance to transit, net acreage, land value, distance to a highway), weight them, and combine them into one score.
Keep the weights written down next to the result, so anyone reviewing the list can see why parcel 3 outranks parcel 4. For the full method, see site suitability analysis with a weighted overlay.
A worked example: seven dry parcels in Nashville
Here is the method applied to a real screen in NeuraCities. The question: in Nashville and Davidson County, which parcels are vacant, over ten acres, within half a mile of an interstate and outside the FEMA 100-year floodplain?
- 1Vacant: keep parcels whose land use description contains VACANT and whose acreage is over 10.
- 2Access: buffer OpenStreetMap interstates by half a mile and keep the parcels inside the buffer.
- 3Constraint: remove anything in the 100-year floodplain.
- 4Zoning: join the 5,992 zoning districts and count each parcel once, even where it straddles two districts.
| Land use | Acres | Zoning | Area |
|---|---|---|---|
| Vacant rural land | 41.5 | AR2A | Joelton |
| Vacant industrial land | 28.4 | IR | Goodlettsville |
| Vacant rural land | 21.3 | AR2A | Joelton |
| Vacant residential land | 18.1 | RS20 | Goodlettsville |
| Vacant rural land | 15.8 | AR2A | Whites Creek |
| Vacant rural land | 15.7 | AR2A | Nashville |
| Vacant rural land | 11.3 | R10 | Goodlettsville |
Seven parcels out of 286,920 is the point of a screen: a list short enough to look at one by one. Read the full story in Seven parcels in Nashville, if you want to stay dry.
Running the whole screen in NeuraCities
Each step above is a standard GIS operation (filter, buffer, overlay, spatial join, dissolve, weighted overlay), which is exactly what the NeuraCities GIS agent runs. You describe the screen in one message:
In Mecklenburg County, NC, find vacant parcels of at least 5 acres outside the FEMA 100-year floodplain, zoned for multifamily or mixed use, within 800 m of a light rail or bus stop. Flag the ones owned by a public agency, and rank the rest by distance to transit and acreage.
- The checklist comes first, so you can see which vacancy definition and which flood zones it will use before anything runs.
- The Data Agent finds the parcel, zoning, flood and transit layers, probes each one for coverage of the county, and attaches its source.
- The GIS agent runs the filters, overlays and joins with deterministic code, and the weighted overlay with the weights you state.
- You get a map, a ranked table you can export to Excel, and the candidate layer as a Shapefile or CSV.
If you have a better parcel file or a utility's service area, upload it and it takes priority. Save the prompt as a workflow with {{county}} and {{min_acres}} as variables to run the same screen in the next county.
What a GIS screen does not tell you
A screen narrows a county to a handful of parcels worth a closer look. It does not replace due diligence on the finalists:
- Title and easements: parcel maps are not the legal record of a boundary or of what is recorded against it.
- Wetland delineation: a screening layer does not establish jurisdictional wetlands; for work that could affect wetlands, the US Fish and Wildlife Service points to the local US Army Corps of Engineers regulatory office.
- Site conditions: soils, contamination, access permits and utility capacity need site-specific checks.
- Data currency: check the update date of the parcel and zoning layers. A rezoning or a new subdivision since the last refresh changes the answer.
Frequently asked questions
How do I find vacant land for development?
Start with your county's parcel layer and flag vacant parcels by land use code, a building value of zero or the absence of a building footprint. Then remove floodplain, wetlands, steep slopes, parks, rights of way and cemeteries, keep parcels whose zoning allows your use and that meet a minimum size, and rank what is left by access and other criteria.
How do I find vacant lots in my city?
Check whether your county publishes a vacant land layer (Mecklenburg County, NC and Cuyahoga County, OH both do). If not, download the county parcels and filter for vacant land use codes or parcels with no building value. Search "[county] parcels GIS" or "[county] vacant land GIS" to find the data.
What is a developable land analysis in GIS?
A screen that starts with all parcels, keeps the vacant or underused ones, removes land that cannot be built on (floodplain, wetlands, steep slopes, protected and public uses), tests zoning, size, ownership and access, and ranks the remaining sites. Each step is a standard GIS overlay, join or filter.
How do I tell if a parcel is vacant in GIS?
Use the assessor's land use or vacant flag, check whether the building or improvement value is zero, or test whether any building footprint intersects the parcel. Combining two of these is more reliable than one, because they often disagree.
Should I remove the whole parcel if part of it is in the floodplain?
It depends on the purpose. Dropping any parcel that touches the floodplain is simple and conservative. Erasing the floodplain and keeping the remaining net area keeps large parcels with a wet corner; if you do that, apply your minimum size test to the net area.
Can AI find developable land?
Yes, if the AI plans the screen and a GIS engine runs it. In NeuraCities you describe the criteria in plain English, and the agents find the parcel, zoning, flood and transit data, run the overlays and return a mapped, ranked list with the source of every layer.
More guides on site selection and land
Describe your criteria and get a ranked, mapped list of developable parcels with every source attached.
Run a vacant land screen