Why are LAS point clouds so large? 5 quick ways to reduce their size

In workplaces that handle LAS point clouds, even if the accuracy and amount of information seem satisfactory immediately after data acquisition, problems often arise when sharing the data internally or using it for drafting or cross-section checks, such as “too large to open,” “it takes a long time to display,” and “the scope each person can handle differs.” In particular, for projects with a wide survey area or those that capture not only terrain but also structures and surrounding objects together, the amount of data can grow unnecessarily large and become a major cause of reduced work efficiency. Many practitioners searching for LAS point clouds are not just interested in the mechanics of point clouds themselves, but also want to know how to make bloated data easier to handle on-site. Therefore, in this article we organize the causes that make the file size of LAS point clouds grow and provide a clear explanation of measures to reduce size that can be easily tried in practice right away. We delve not only into simply reducing file size but also into approaches that keep the necessary accuracy while enabling smooth viewing, sharing, editing, and deliverable creation.

Table of Contents

• Reasons why LAS point cloud file sizes increase

• Problems that occur on-site when LAS point clouds are heavy

• Immediate weight-reduction measure 1: Cut out unnecessary areas first

• Reassess the densities of 2 readily implementable weight-reduction measures by application

• Quick weight-reduction measure 3: Clean up unnecessary attribute information

• Quick measures to reduce size 4: Manage data by splitting it

• Quick Weight-Reduction Measure 5: Separate Storage and Work Sets

• Checklist to Avoid Failure in Weight Reduction

• Summary

Reasons LAS point cloud file sizes increase

The main reason LAS point clouds become large in size is that they contain a large number of points. This may seem obvious, but in practice this obvious fact is often not fully taken into account. Each point in a point cloud has positional information, and when attributes such as elevation, return intensity, classification, and color information are added, the data become not merely a collection of coordinates but highly information-rich. As the survey area expands, or as objects are captured with higher density, file sizes grow rapidly.

Additionally, capturing not only the required area but also surrounding unnecessary items is a typical cause of increased data volume. For example, even in projects where you only need to check part of the terrain or a structure, if the dataset includes large amounts of scenery along the travel route, work vehicles, temporary installations, pedestrian traffic, surrounding vegetation, and so on, it may appear information-rich but in practice becomes data that functions only as noise. When such unnecessary points accumulate, processing load increases while contributing little to improving the quality of the deliverables.

Another often overlooked issue is excessive point density. It’s easy to assume that acquiring data at a high density is safer, but not every task requires the highest density. Managing areas that need detailed shape checks and areas where a general understanding of the terrain is sufficient at the same density will unnecessarily increase the overall data size. In particular, for wide-area terrain surveys, long corridors, and comparisons before and after earthworks, point density is often excessive for the intended purpose, making the data difficult to handle as is.

Moreover, when there are many attributes such as color information and classification information, the data volume also increases. While attributes are effective in downstream processes, always retaining attributes you do not use can make the data too heavy for viewing or internal sharing. In other words, the size of a LAS point cloud expands not only with the number of points but also with the acquisition area, point density, types of attributes, the inclusion of unwanted objects, and mismatches with the intended use.

Problems That Occur on Site When LAS Point Clouds Are Heavy

If LAS point clouds are left large in size, the first thing that becomes apparent is a drop in display speed. It takes a long time to open files, and if rendering lags every time you move the viewpoint or zoom, the inspection work itself becomes stressful. When field personnel only want to take a quick look at a cross-section, long wait times for display alone will reduce how often the data is used. Point clouds that should be helpful end up becoming unused data because of their size.

The next issue that arises is the difficulty of sharing. When file sizes are large, even passing them around within the company takes time, and it becomes hard to set up a system where multiple people can handle them simultaneously. People who only want to view the data, those who want to use it to create drawings, and those who want to use it for as-built verification require different amounts of information, yet an operation in which everyone receives the same massive file is inefficient. As a result, information does not reach the right people in the form they need, and the benefits of introducing point clouds are diminished.

Furthermore, the impact on editing and deliverable creation processes is significant. If you cut cross-sections while many unnecessary points remain, the desired section becomes harder to see. If you perform diagramming with many temporary structures or vegetation present, the effort required to select the necessary objects increases. The amount of storage required is not merely a matter of storage space; it is a problem that directly affects the speed of decision-making and the accuracy of work in downstream processes.

In some sites, the performance differences of the devices used cannot be ignored. Even if a high-performance environment can somehow handle them, they may run slowly on typical office workstations or portable devices, making it impossible to use them the same way in the field and in office tasks. When such a gap exists, people who can use point clouds and those who cannot become divided, leading to a reliance on specific individuals. The problem with heavy LAS point clouds should be viewed not simply as an issue of large file size, but as a practical challenge that slows down viewing, sharing, editing, explanation, and decision-making.

Quick weight-reduction measure 1: Cut out unnecessary areas first

The most effective and easiest-to-start measure is cropping out unnecessary areas. When an LAS point cloud feels too heavy, many people first think of thinning the points, but simply leaving only the truly necessary area beforehand can greatly reduce the file size. In practice, the target area often includes a wide swath of extraneous background around its perimeter, and that alone can account for a large number of wasted points.

For example, it's not uncommon for data collected for slope inspection to also include the airspace above or the area behind, or for data intended for structure inspection to still contain surrounding roads or distant scenery. If you continue to preserve everything in this state, the dataset will become hard for anyone reviewing it later to discern what the primary subject is. First, it's important to clearly define and isolate the target area based on the project's main objective.

As a way of thinking about extraction, it is effective to consider separately the scope required for the deliverables, the scope required for verification tasks, and the scope that should be retained for future reuse. Rather than cramming everything into a single file, creating separate working data focused on the primary subject greatly improves everyday processing speed. In particular, for tasks where the viewing location is clearly defined—such as cross-section creation, as-built verification, and quantity estimation—simply narrowing the scope makes them considerably easier to handle.

The advantage of this measure is that it has relatively little impact on accuracy. If you only remove unnecessary areas, you can make the data lighter while preserving the main subject's point density and attributes. In other words, it's easy to reduce file size without substantially degrading quality. As a first step in reducing data size, it can be said to be the safest and most practical method.

Two Quick Weight-Reduction Measures: Review Density by Application

Next, one of the most effective measures is to reassess point density based on the intended use. LAS point clouds give a greater sense of reliability the denser they are, but you don't always need to use the maximum density. What matters is what the point cloud will be used for. The required density differs between applications that need detailed shape reproduction and those aimed at grasping overall trends or checking rough cross-sections.

For example, high density is useful when you need to inspect fine surface irregularities or the ends of structural members, but if you are only checking broad terrain trends or comparing changes across multiple time points, such a high density may not be necessary. Rather than performing all tasks at the same density as the original, it is important to adopt the idea of creating a lightweight version tailored to the intended use.

One point to be careful of here is that it's not enough to simply thin out the entire dataset uniformly. If you reduce density across the board without considering the characteristics of the target object and the purpose of the task, you may lose shapes or details that are required. The important thing is to first clarify which stage requires what level of detail. Even just separating densities for viewing, sharing, analysis, and storage will significantly improve operations.

Also, reviewing point density can make it easier not only to speed up processing but also for staff to make decisions. Point clouds with too much information may appear abundant at first glance, but visibility is reduced and necessary features can become obscured. Moderately organized lightweight data can actually be more usable in the field. If you regard data reduction not as a compromise but as the process of optimizing the amount of information for the intended use, you’re less likely to fail.

Quick weight-reduction tip 3: Clean up unnecessary attribute information

LAS point clouds can include various attribute information in addition to coordinates. These can be very useful depending on the purpose, but they are not necessarily required for every workflow. If viewing or simple sharing is the main objective, keeping unused attributes can simply bloat the file.

When organizing attribute information, the important thing is to clarify what is truly needed at each stage of the process. For example, if the sole purpose is position verification, you do not always need to retain all supplementary information. Conversely, in processes that use classified features or that require making judgments by observing color differences, cutting attributes too much can actually make the work more difficult. In short, organizing attributes is not aimed at reduction itself, but at arranging information to suit the people who will use it.

A common practice on site is to distribute the complete dataset as acquired to everyone. While this is convenient for analysts, it results in files that are too heavy for those who only need to view them. If you separately prepare a lightweight version with attributes streamlined as needed, routine verification tasks become much easier. The value of lightweight viewing data is especially high on projects with many stakeholders or those that include personnel unfamiliar with point clouds.

However, attribute organization should always be performed on a separate copy rather than by overwriting the original. The original may contain information that will be needed for future reanalysis or other uses. If archival data and operational data are mixed up, you may not be able to restore the necessary information later. The more you pursue streamlining, the more important this approach to original data management becomes.

Quick weight reduction measure 4: Split and manage data

On projects with large LAS point clouds, maintaining the entire dataset as a single large file becomes an operational burden. Therefore, dividing and managing the data by area and by purpose is highly effective. In particular, for linear structures, wide-area terrain, or projects spanning multiple construction sections, planning for divided management from the start makes it easier to proceed stably through subsequent processes.

The advantage of partitioned management is that you only need to open the parts you require. Since you don't have to load the whole thing each time, display speed and usability improve, and the time to start work is reduced. In addition, because you can give each person only the scope relevant to them, sharing becomes easier. Viewers no longer need to take on areas unrelated to them, which reduces the burden of daily work.

Additionally, splitting is also effective for organization. It clarifies the scope, making it easier to see which file corresponds to which construction section or which purpose. If you store a single large point cloud, management can become confused as updated or edited versions increase, but if you establish naming and storage rules from the start and split the data, it becomes easier to reuse.

On the other hand, be careful not to over-split. If you split things too finely, it becomes difficult to combine them or perform cross-checks. What’s important is to divide in a way that is meaningful at the work-unit level. Dividing by construction section, by object, by process step, or into units for viewing and for analysis — making these divisions with downstream processes in mind — makes it easier to achieve both weight reduction and operability.

Quick weight-saving tip 5: Separate storage and working sets

The most important aspect of streamlining is the idea of making operations lighter while preserving the original. Essential to that is an operational practice of separating archival copies from working copies. A common problem in many workplaces is turning the lightweight data into the single official dataset simply because it is convenient. That makes it impossible to respond when higher precision is required later or when you need to create a different deliverable.

For archival purposes, we retain complete information that is as close as possible to the point of acquisition. Because this has value for future reanalysis and as an audit trail, we do not lightly trim it even if it takes up a large amount of storage. By contrast, for operational use we prepare data that narrows the target scope, adjusts density to what is necessary, and is organized with only the minimal required attributes. If routine viewing, sharing, and quick checks are carried out using this operational dataset, on-site operational speed will improve significantly.

The advantage of this separation is that it makes it easier to decide what to slim down. Because you have the reassurance of keeping the original, you can boldly prioritize ease of use for the working copy. As a result, it becomes easier to avoid situations where data go unused because they are too heavy. Point cloud data do not realize their value simply by being acquired; their value emerges only when they are used on site. Separating archival and working copies is fundamental to ensuring that value is integrated into practical work without being diminished.

Furthermore, roles among stakeholders also become clearer. The data required differs between the person responsible for managing the original documents, the person who uses them to create deliverables, and the person who only views them. Rather than giving everyone the same materials, establishing a system that distributes data according to role reduces unnecessary burden and leads to fewer problems. It is important to consider streamlining not only as a technical process but as an aspect of operational design itself.

Points to Check to Avoid Failure in Weight Reduction

Reducing the size of an LAS point cloud is not something you should do indiscriminately. The first thing to confirm is what the point cloud will be used for. Deliverable map production, cross-section checks, as-built management, progress sharing, long-term archiving, etc.—the appropriate reduction method varies depending on the use. If you proceed with reduction while leaving this unclear, you risk discarding necessary information and having to recreate it later.

Next, what you should verify is how much accuracy needs to be preserved. Reducing file size is convenient, but sacrificing too much accuracy defeats the purpose. It is important to make that judgment while checking on the actual work screen—confirming that the object's contours haven't been degraded, that cross-section readings are not impeded, and that the details needed for comparison remain. It's safer not to judge success solely by the numeric file size.

Also, it is necessary to confirm who can easily use the reduced-size data. Even if it works fine in high-performance environments, it is meaningless if it remains heavy on on-site devices or typical office workstations. After creating a lightweight version, you should check in the actual users’ environments whether it opens more easily, is easier to view, and is easier to share. Point clouds deliver greater implementation benefits when they are tailored to the environments of the people who use them rather than those who create them.

Finally, it is important to establish data management rules before and after downsizing. If it is unclear which is the original, which is the lightweight version, or what each is intended for, confusion will arise later. By organizing file names, storage locations, version control, and records of the applicable scope, the effects of downsizing will not be temporary but will lead to continuous operational improvements. Do not stop at reducing storage; only when you design workflows that are easy to handle can you truly call the downsizing a success.

Summary

The reason LAS point clouds become large is not simply because there are many points. The inclusion of irrelevant areas, excessive point density, unused attribute information, and workflows that keep everything as a single gigantic dataset are multiple factors that combine to create data that is hard to use in the field. That's why countermeasures beyond simple downsampling are insufficient. By clipping the target area, reviewing density by use case, cleaning up attributes, managing data in partitions, and separating archival and working datasets, you can move toward a more manageable point cloud workflow without significantly degrading accuracy.

In practical work, efforts to keep point clouds from becoming heavy begin not only with post-processing but already at the planning stage before acquisition. If you clearly define the necessary scope and organize what to retain at what level of accuracy, you are less likely to create unnecessarily large datasets in the first place. If confirmation of reference positions on site or understanding of the target area is vague, you can easily capture extra areas later, which often results in point clouds that are heavy and difficult to handle.

What should be reconsidered, then, is the accuracy of preliminary location checks and simple surveying carried out before point cloud acquisition. For example, in situations where you need to quickly grasp positions on site while organizing the required area, a high-precision positioning device such as an LRTK that can be used with a smartphone mount is useful. If you establish the target area, control points, and verification points while confirming positions at the centimeter level (half-inch level), you are less likely to end up with point clouds containing many unnecessary points in later processes, and data organization becomes easier. To make LAS point clouds lighter and easier to use, it's important not only to reduce weight after acquisition but to streamline the entire workflow before and after acquisition. To turn point clouds into information assets that can be repeatedly used on site rather than merely stored data, incorporating efficient on-site reconnaissance using LRTK together with data slimming will make a significant difference in future practice.