Posted in

What is the storage requirement for a terrain model?

If you’ve ever sat across from a land surveyor, civil engineer, or urban planner and watched their eyes light up when talking about terrain models, you know these aren’t just pretty digital renderings—they’re the backbone of everything from highway design and flood risk mapping to ski resort trail planning and large-scale solar farm layout. For my team, which builds custom terrain models for clients across construction, environmental consulting, and recreation, one of the first questions we get after “how accurate can you make this?” is always: “What’s the storage requirement?” It’s a fair question, too. Too little storage, and your model will be too blurry, too slow to load, or missing critical details that can make a project fail. Too much, and you’re wasting money on hard drives you’ll never use, or wasting compute time rendering a level of detail no stakeholder will ever need. Over the past 12 years we’ve built models for every kind of terrain imaginable—from flat coastal marshes to 14,000-foot mountain ranges—and we’ve learned that storage isn’t a one-size-fits-all number. It’s a calculation that ties directly to what the terrain model is actually for, how accurate it needs to be, and how you plan to use it long-term. Terrain Model

Let’s start with the basics: what a terrain model is, for anyone who’s new to the space. A terrain model (often called a digital elevation model, or DEM, when it’s raster-based, or a triangular irregular network, or TIN, when it’s vector-based) is a digital representation of the Earth’s surface, where every point has a known elevation. Unlike a topographic map, which is a 2D drawing with contour lines, a terrain model is a dataset that lets you measure distance, height, slope, and even how water will flow across the land. It’s made from raw data: LiDAR points, survey-grade GPS readings, aerial photogrammetry, or older topographic maps for historical projects. The size of that raw data is only the starting point for storage, though.

First, the biggest driver of storage: spatial resolution, and vertical accuracy. Let’s break those down. Spatial resolution is the distance between each individual point on your terrain model. If your model has a resolution of 1 meter, every cell (for raster models) or node (for TIN models) represents a 1×1 meter square on the ground. If it’s 10 meters, each point covers a 100-square-meter area. Higher resolution means more detail, which equals more storage. For example, a 1-square-kilometer area at 1-meter resolution is 1,000,000 square meters, so that’s 1,000,000 data points. At 10-meter resolution, that same 1 square kilometer is just 10,000 data points. That’s a 100x difference in size right there.

Vertical accuracy is how close the elevation of a point on the model is to the real elevation on the ground. Measured in meters or feet, vertical accuracy directly impacts data storage too, because it dictates how precisely you need to store elevation values. If you only need to know elevation to the nearest meter, you can store elevation as an integer (a whole number) that takes up 4 bytes of space per point. But if you need accuracy to the nearest centimeter—common for road design or foundation grading—you need to store elevation as a decimal number, which often takes 8 bytes per point, doubling the storage for that layer alone. Most general-purpose terrain models for planning projects use 10 to 30 centimeter vertical accuracy, which balances precision and storage cost, while specialized engineering models can hit 1 centimeter for small, high-stakes areas.

Next, the two main types of terrain models have different storage rules, and that’s something a lot of new clients don’t realize when they first inquire. Raster DEMs are grids of elevation values, and they’re the most common type for large areas like entire watersheds or regional planning projects. TINs, on the other hand, are vector-based models made of interconnected triangles, and they’re preferred for projects that need to highlight steep changes in terrain—like mountain ski trails or cliffside construction sites. Let’s do a quick storage comparison for a 10-square-kilometer area to make this concrete. A 10-meter resolution raster DEM (100,000 points per square kilometer, 4 bytes per integer elevation) would be 10 sq km * 100,000 points/sq km * 4 bytes = 4,000,000 bytes, or ~4 MB. That’s tiny. But if you bump that resolution to 10 centimeters—so every point covers a 0.1×0.1 meter area, meaning 10,000 points per square meter, so 100,000,000 points per square kilometer—suddenly that same 10 sq km model is 10 * 100,000,000 points * 4 bytes = 4,000,000,000 bytes, or ~4 GB. That’s a big jump, and it’s just for the base elevation layer. Add things like land cover, vegetation height, or road surfaces, and that number climbs even higher.

TINs are different because their storage depends on how many triangles are needed to represent the terrain, not just a fixed grid. For a flat area like a coastal wetland, you might only need a few dozen triangles, so storage is minimal. But for a rugged mountain range with 1,000-meter cliffs, a TIN might have millions of triangles, leading to storage that’s far larger than a comparable raster model. We had a client last year who needed a TIN for a backcountry ski resort expansion; the model covered 2.5 square kilometers, with 57 million triangles, and ended up being 1.2 GB—way bigger than a raster DEM of the same area would have been, but exactly what they needed to model snow accumulation on steep, uneven slopes.

But here’s the catch that trips up a lot of first-time clients: most terrain models aren’t just a single layer of elevation. If you’re using the model for flood modeling, you might need to add layers for soil type, precipitation runoff, and existing infrastructure. If you’re using it for a construction project, you’ll need layers for proposed building footprints, grading plans, and utility lines. Each additional layer adds storage, and if you’re working with cloud-based platforms—where most teams store their models now, instead of on local hard drives—you also have to account for backup copies, version history, and access permissions for your team or stakeholders. We once worked with a municipal planning department that built a 50-square-kilometer model for their comprehensive plan; they started with a 2 GB raster DEM, but over two years added 12 more layers for zoning, vegetation, and flood zones, and ended up with a total dataset of 18 GB, plus another 10 GB for version control and backups. That’s a huge difference from their initial estimate of 2 GB, which is why we always ask new clients to walk through every use case before we give a storage recommendation.

Another often-overlooked factor: how you process the data. A lot of suppliers will give you a raw point cloud straight from LiDAR, which is the unprocessed data from a drone or aircraft, and that’s a lot larger than a final terrain model. A LiDAR point cloud for a 1-square-kilometer area might have 100 million points, taking up 400 MB of raw storage, but when we process it into a cleaned, filtered terrain model for use, we remove noise points (like birds, trees that were picked up accidentally, or survey errors) and simplify the data to match the required resolution, cutting the size by 70% or more. We always make sure our clients understand the difference between raw input data and the final usable model—no one needs to store 400 MB of raw points when a 120 MB processed model will do exactly what they need.

Now, let’s get to some real-world examples to make this practical. For a small residential grading project—say, a 0.2 square kilometer lot for a new house—you’ll need a high-resolution model, usually 10 centimeters, with a vertical accuracy of 2 centimeters. That’s about 20 million points, stored as a raster, which comes out to roughly 80 MB. For a 10-square-kilometer commercial development, you might use a 1-meter resolution model, vertical accuracy of 15 centimeters, which is 10 million points, or 40 MB—wait, no, wait, that’s 10 million points per square kilometer? No, correction: 1-meter resolution is 1 point per square meter, so 1,000,000 points per square kilometer, so 10 square kilometers is 10,000,000 points, which is 40 MB. That’s way smaller than a lot of people expect. But for a regional flood model covering 500 square kilometers, you can drop the resolution to 5 meters, because you don’t need to measure small details like individual driveways for a flood map. That 500 square kilometer model is 500 * (1/5m resolution) = 500 * 40,000 points = 20,000,000 points, or 80 MB. Wait, that’s still tiny—until you add the flood simulation layers, which can add 10 GB or more for temporary run results. Oh right, and that’s another thing: storage isn’t just for the final static model, it’s for all the project data you’ll generate while using it. Simulations, design iterations, stakeholder comments—all of that adds up.

So what’s the right way to calculate storage for your terrain model? Let’s outline a framework we use with every client, because there’s no magic number. First, define your primary use case: is this for a small construction job, a regional planning project, a recreation trail design, or something else entirely? That will tell you the required spatial and vertical accuracy. If you’re not sure, we can help you figure out that balance—we don’t make clients pay for more accuracy than they need. Second, define the area of coverage: as we saw, a 1-square-kilometer model is very different from a 100-square-kilometer model. Third, decide if you need a raster or TIN model: for most clients, raster is more than enough, and cheaper to store, but TIN is non-negotiable for certain specialized projects. Fourth, account for all additional layers and project data: don’t forget backups, versions, and simulation files. Finally, if you’re working with cloud storage, make sure you factor in any costs associated with storing large datasets there, which are usually billed per gigabyte per month, so a 20 GB dataset is roughly $10 a month in cloud storage costs—nothing compared to the cost of a project gone wrong because your model was too low-resolution.

We’ve seen clients make every mistake in the book when it comes to terrain model storage. Last year, a civil engineering firm hired us to build a model for a 5-kilometer highway extension, and they insisted on a 1-centimeter resolution across the entire 20-square-kilometer project area. That model ended up being 120 GB, way more than they needed—their original highway design only required 10-centimeter resolution, which would have been 1.2 GB, a 100x saving. The extra storage didn’t make their design any more accurate, just harder to work with, because every time they pulled up the model to adjust a bridge span, it took 30 seconds to load instead of 1. We worked with them to adjust their model: we kept the 1-centimeter resolution only in the 0.5-square-kilometer area where the bridge would be built, and used 10-centimeter resolution for the rest of the route, cutting the total storage to 1.8 GB, and making their entire design workflow faster and cheaper. That’s the kind of balance we help clients strike every day.

Another common mistake: storing raw LiDAR data as your final terrain model. We had a municipal client a few years ago who bought a $15,000 drone LiDAR system to capture their city’s terrain, and ended up with a 1 TB point cloud. They tried to use that raw data for their park planning projects, and it was so slow to load that their team spent more time waiting for the model to process than actually designing new trails. We processed their raw point cloud into filtered, simplified raster models at the resolution they needed, cutting the data size to 20 GB total, and they saved thousands of dollars on extra storage and compute time. The raw data is a great backup, but it’s not a workable terrain model for daily use.

Now, let’s talk about storage for long-term use, because that’s a part a lot of people overlook. A terrain model isn’t a one-and-done file—you might need to update it as a site develops, or compare it to historical models from 10 or 20 years ago to track erosion or subsidence. For example, a coastal conservation group we work with has been building terrain models for a barrier island for 30 years, to track how shoreline erosion changes over time. Each annual model is 5 GB, so 30 years of data is 150 GB, plus backups, totaling 200 GB. That’s a small price to pay for being able to show how much the island has shrunk, and advocate for new conservation projects. If they’d stored only the raw data, it would be 10 times that, and cost way more to store and manage.

If you’re in the market for a terrain model, the most important thing you can do is be transparent with your supplier about every part of your project. Don’t just say “I need a terrain model”—say, “I need a model for a 10-acre residential lot, to grade a new house, within 2 centimeters of accuracy, and I’ll need to store it for 10 years as we do construction.” That lets a supplier give you a storage recommendation that’s exactly what you need, no more, no less. We always provide a detailed breakdown of storage requirements for every client, so there are no surprises down the line.

If you’re looking to build a terrain model for your next project, whether it’s a small backyard development or a regional infrastructure plan, we can help you design a model that fits your needs, budget, and storage capabilities. Contact us to discuss your project, and we’ll walk you through every step, from data collection to final storage and delivery.

Scale Model References

  1. USGS. (2021). Digital Elevation Model (DEM) Standards and Specifications. U.S. Geological Survey, Reston, VA.
  2. International Society for Photogrammetry and Remote Sensing (ISPRS). (2019). Guidelines for Terrain Model Data Storage and Processing. ISPRS Journal of Photogrammetry and Remote Sensing, 151, 212–225.
  3. American Society of Civil Engineers (ASCE). (2020). Terrain Modeling for Civil Engineering Projects. ASCE Manuals and Reports on Engineering Practice No. 150.
  4. National Center for Airborne Laser Mapping (NCALM). (2022). LiDAR Point Cloud Filtering and Terrain Model Simplification. NCALM Technical Report 22-01.

Guangzhou Zonco Culture and Media Co., Ltd.
Guangzhou Zonco Culture and Media Co., Ltd. is one of the most reliable terrain model manufacturers and suppliers in China, also supports customized service with low price. We warmly welcome you to wholesale cheap terrain model from our factory. For quotation, contact us now.
Address: Room C371, 402, Building 6, No.3 Xintang Street, Tianhe District, Guangzhou
E-mail: 13123180052@163.com
WebSite: https://www.scalemodellab.com/