Does drone surveying produce the same accuracy every time? Six causes of variation explained

When introducing drone surveying, many responsible parties worry about "it worked well last time, so why is it slightly off this time?" In fact, while drone surveying is efficient and makes it easy to capture wide areas in a short time, it does not always produce exactly the same accuracy every time. Even with the same aircraft and processing software, weather, flight technique, imaging conditions, placement of ground control, processing settings, and the site environment itself can cause variations in results.

This does not mean drone surveying is inherently unstable. What matters is understanding where the sources of accuracy variation lie and standardizing procedures to improve reproducibility. Also, if you define for each use case "how much variation is acceptable," it becomes easier to make on-site decisions. This article clearly explains why drone surveying may not yield the same accuracy each time by organizing the causes into six categories, and discusses approaches to improve reproducibility in practice.

Table of Contents

• Why doesn’t drone surveying produce the same accuracy every time?

• Cause of variation 1: Effects of weather conditions

• Cause of variation 2: Differences in flight conditions

• Cause of variation 3: Differences in imaging settings

• Cause of variation 4: Placement and measurement of ground control

• Cause of variation 5: Differences in processing settings and procedures

• Cause of variation 6: Differences in the site environment itself

• You can’t make variation zero, but you can improve reproducibility

• Key points to standardize in practice

• How much variation should you tolerate?

• Summary

Why doesn’t drone surveying produce the same accuracy every time?

In short, the accuracy of drone surveying is not determined solely by the performance of the aircraft. The accuracy of deliverables is determined by a combination of factors: the quality of the images acquired from the air, the geometric relationships during capture, the ground control information, and the modeling conditions during processing. In other words, accuracy is shaped across the entire workflow—from planning to fieldwork to processing—not by a single step.

For example, at the same site, one day might have light winds, stable sunlight, a clearly visible ground surface, and sufficient ground control points. In that case, image overlap and feature extraction are stable, and results tend to be relatively reproducible. Conversely, on another day there might be strong winds, pronounced shadows on the ground, and an insufficient number or poor distribution of ground control points; even if nothing appears obviously wrong, the processing results can easily include slight distortions or positional shifts.

It is important to note that "good or bad accuracy" and "whether the same result is obtained every time" are somewhat different issues. Even if you get an excellent result once, if you cannot reproduce the same conditions next time, the operation is unstable. In practice, it is more important to establish a system that consistently meets the required accuracy than to chase the highest possible value.

Therefore, when deciding to adopt drone surveying, you should not rely solely on catalog specs or a single success story. Instead, you need to organize which conditions cause what degree of variation and which factors your operation can control. Below, we examine six factors that directly affect accuracy variation.

Cause of variation 1: Effects of weather conditions

One of the most familiar factors affecting the reproducibility of drone surveying is weather. Wind, solar radiation, cloud cover, and ground moisture in particular have direct or indirect impacts on accuracy.

First, wind. Strong winds force the aircraft to make frequent attitude corrections, disrupting the intended straight-line flight and altitude maintenance. While modern aircraft are highly stabilized, they cannot completely eliminate the effects of crosswinds or gusts. Large attitude changes introduce subtle differences in image capture angle and overlap, which can make feature matching during processing unstable. This is especially true at sites with large elevation changes, slopes, or near structures, where wind eddies can produce localized motion.

Next, sunlight and shadows. Clear skies may seem ideal for imaging, but strong direct sunlight produces deep shadows, increasing brightness contrasts even for the same ground features. Feature extraction tends to be unstable at shadow edges, which can degrade the quality of ground interpretation and point cloud generation. If shadow directions differ between morning and afternoon, the site can look different and become a source of noise for comparisons.

Cloudy conditions diffuse light and generally reduce harsh shadows, improving visual uniformity. However, if it’s too dim, shutter speeds may slow and small blurs can occur. In short, you can’t simply assume that "sunny is good" and "cloudy is bad"—you need to consider the balance of wind and light.

Also, rain the day before or on the day of capture can cause reflection issues, mud, and puddles. Water surfaces are poor for feature detection in images and reflections change over time. On construction sites, agricultural land, or unpaved areas, differences in wetness can drastically change the ground appearance and reduce the accuracy of as-built verification or time-series comparisons.

Weather is an unavoidable disturbance in practice, but it can be managed. The key is not to judge flight permission by “can we fly” alone. Check wind speed, gusts, solar intensity, shadow presence, and ground moisture, and decide whether “today’s results will be comparable to the previous run.” Even when it is safe to fly, the day may be unsuitable for highly reproducible surveying. Whether you can standardize this assessment will significantly affect operational stability.

Cause of variation 2: Differences in flight conditions

Even with the same aircraft, variations in flight altitude, speed, route planning, and overlap rates will change the accuracy of results. Unlike weather, flight conditions are relatively controllable on site; therefore, whether you standardize them has a direct impact on result differences.

First, flight altitude. Flying higher covers a larger area per image but increases ground sampling distance, making it harder to capture small shape changes. Flying lower provides finer detail but increases the number of images, flight time, and processing load. The problem occurs when altitude settings vary by feel between projects—one time flown lower, another time higher for efficiency—which changes the character of the results for the same site.

Flight speed also matters. Flying too fast can degrade image quality depending on shutter speed and aircraft attitude changes. Speeding up on a windy day, in particular, can reduce image stability. Slowing down stabilizes imaging but increases operation time and battery swaps. Again, it’s better to determine an acceptable range per site type rather than relying on intuition.

Route planning can be simple grid flight for some sites, but sites with slopes or structures may require oblique shots or additional routes. Leaving this to the operator’s discretion each time may lead to days with necessary supplementary imaging and days without it, resulting in differences in model quality.

Forward and lateral overlap rates heavily affect reproducibility. Low overlap reduces processing load but weakens feature connectivity, increasing distortion and gaps. Excessive overlap stabilizes results but dramatically increases image count. The important thing is to set adequate parameters for site conditions and follow them consistently.

A common practical mistake is changing flight conditions on the day due to deadlines or weather. Flexibility is necessary, but if you change settings, record what changed; otherwise you cannot trace the cause of result differences later. Simply recording flight altitude, speed, overlap, camera angle, and whether supplemental routes were used makes it easier to isolate causes of variation.

Cause of variation 3: Differences in imaging settings

In drone surveying, not only how you fly but what kind of images you acquire has a major impact on results. Differences in imaging settings are often overlooked but are crucial to accuracy stability.

Pay particular attention to shutter speed, ISO sensitivity, exposure, focus, and camera angle. For example, flying with a slow shutter speed can introduce slight motion blur, which may not be obvious visually but can reduce the precision of feature extraction. This affects the overall positional accuracy and geometric stability in processing. Relying entirely on automatic settings in dark conditions increases the likelihood of this kind of blur.

High ISO introduces noise and coarsens image quality, reducing the ability to interpret fine details. Conversely, underexposure eliminates information in shadowed areas and makes feature points scarce. Automatic settings are convenient but behave differently under varying light, so be cautious when aiming for reproducible results.

Focus shifts should not be ignored. When capturing large areas, a few slightly soft images can go unnoticed in the field but cause local instability during processing. Lens dirt or condensation is the same issue. These problems are common during morning operations or periods with large temperature swings, and pre-flight inspections can miss them.

Camera angle matters too. For some sites, nadir (straight-down) images are sufficient, while for slopes, retaining walls, or around structures, adding oblique imagery stabilizes models because nadir shots alone may not represent shapes well. On the other hand, inconsistent angle composition between runs reduces comparability.

In the field, don’t leave imaging settings to operator experience alone. Define baseline camera settings for each site type, a combination of flight speed and settings, and review criteria for poor light. This reduces quality variation. Especially for beginners, don’t assume “automatic capture is fine”; check whether the images are “suitable for processing.”

Cause of variation 4: Placement and measurement of ground control

If you want stable accuracy in drone surveying, you cannot neglect ground control. Here, ground control refers mainly to aerial targets, control points, check points, or connections to known points. Rather than producing deliverables purely from aerial images, providing reliable ground-based positions suppresses overall distortion and absolute positional shifts.

A common misconception is that high onboard positioning performance makes ground control almost unnecessary. While aircraft positioning has improved recently, ground control remains important depending on site conditions and required accuracy. For tasks requiring trustworthy absolute positions or heights—such as as-built verification, volume management, construction management, or alignment checks with existing structures—reliable ground references form the foundation of deliverable quality.

Variation often arises when the number, distribution, visibility, or measurement procedure of control points is inconsistent. For example, one day you might place aerial targets balanced around the perimeter and center, while another day targets are biased to one side due to site constraints. This changes the constraint conditions during processing and can cause local distortion or vertical instability.

Small or poorly contrasting targets that blend into the background are also problematic. You might think you placed them correctly, but from the air they are hard to identify, leading to input errors. Standardize marker shape, size, color, and placement strategy.

Managing measurement accuracy itself is essential. If ground control measurement varies, even careful aerial imaging and processing won’t stabilize deliverables. Standardize observation time, positioning method, fixation conditions, and verification procedures. Often the drone operator and ground surveying personnel are different people, and poor coordination can lead to subtle, hard-to-detect errors.

From an adoption standpoint, decide not “do we use only the drone?” but rather “for which uses and to what extent should ground control be included?” Quick situational awareness demands different reproducibility than quantity calculation or construction management. If you expect the same accuracy every time, you need processes that maintain the same level of ground control as well as aerial procedures.

Cause of variation 5: Differences in processing settings and procedures

Even if you fly the same way, different processing conditions produce different results. In drone surveying, the processing step—generating point clouds, orthophotos, and terrain models from images—is critical. When there are differences in how people perform this step, reproducibility suffers.

A typical example is differing criteria for image selection. One operator may discard any slightly blurred images, while another prioritizes keeping more images. This changes both the quality and quantity of input data. Additionally, settings such as feature extraction strength, matching conditions, camera calibration handling, point cloud decimation, and ground-surface extraction parameters all affect smoothness and detail in the results.

Vertical evaluations are especially sensitive to how the ground surface is extracted from the point cloud. On grassy areas, embankments, or sites with scattered materials, separating ground from non-ground is difficult, and results can vary depending on settings. Using different filtering conditions during processing can produce differing volume calculations for the same site. This is not solely a flight accuracy issue.

If coordinate systems and vertical datums are not handled consistently, comparing results itself becomes unreliable. Deliverables may look tidy, but if reference frames differ, you cannot make strict comparisons with prior results. This is an area where beginners often stumble but it is extremely important in practice.

Do not over-rely on automated processing. Automation is convenient, but automatic outputs are not guaranteed to be the same each time. When image conditions or ground appearance differ, automatic behavior also changes. To improve reproducibility, create processing templates by project type and document which settings are standard and which conditions require review.

Also standardize final validation procedures. If it is unclear what constitutes acceptance—comparison with check points, reconciliation with known points, cross-sections, overlay with previous results—different operators will make different judgments. Standardize not only processing but also evaluation rules to reduce accuracy variation.

Cause of variation 6: Differences in the site environment itself

Finally, the often-overlooked factor is differences in the site environment. Even at the same site name, timing or stage of works can change ground conditions, obstacle layout, and visibility, which causes accuracy variation.

For example, on a development site, a previously exposed graded surface may now be covered by materials or heavy equipment. On grasslands, ground visibility changes with vegetation growth. For slopes or excavation sites, shadowing and the exposure of slopes differ by season. These changes affect how easy the site is to process.

Proximity to structures, trees, power lines, and temporary facilities also matters. The more objects appear in aerial images, the harder it can be to extract the target ground or structure surfaces. In narrow sites or areas near urban zones, simple nadir imaging may not achieve sufficient results.

Ground texture is also important. Uniform pavement, sand, water, or reflective materials often produce unstable feature points. Conversely, surfaces with moderate patterns or granular texture are easier to process. In other words, even with identical flight conditions, differences in how a site looks will change results.

Understand that some variation from site environment is unavoidable. Practically, separate which types of sites are likely to produce stable results from drone-only surveying and which require ground surveying or additional checks. Treating open sites and obstacle-dense sites with the same expectations leads to poor operational decisions.

You can’t make variation zero, but you can improve reproducibility

As we’ve seen, drone surveying accuracy is determined by multiple interacting factors. Therefore, expecting zero variation is unrealistic. What’s important is understanding the range of variation and keeping it within acceptable limits.

To improve reproducibility in practice, don’t let tasks depend on individual skill. Standardize pre-flight checks, ground control placement methods, flight conditions, imaging settings, processing templates, and validation methods so that different operators can make the same decisions.

For example, before flight check not only wind speed but also shadow intensity, ground moisture, and any areas of poor visibility. For flight, set altitude, speed, overlap rates, and criteria for supplemental imaging by site type. Standardize the placement strategy and measurement procedures for ground control. For processing, establish standard settings and validation items per project type. Fixing these elements one by one reduces variation sources.

Accumulating records is also effective. If you record weather, flight conditions, imaging settings, ground control details, processing settings, and validation results for each site, you can compare conditions that produced good results with those that did not. Teams with high reproducibility maintain these records; if you only say “it was slightly off this time” you cannot improve.

Key points to standardize in practice

To improve reproducibility, focus on areas prone to inconsistent judgment. First, clarify the purpose before starting work. Required accuracy differs depending on whether the task is situational awareness, quantity calculation, or as-built management. Flying without a clear purpose tends to result in inadequate flight conditions and ground control.

Next, classify sites. Create standard flight and imaging templates for flat terrain, terrain with slopes, sites with large elevation differences, and obstacle-dense sites. Using standard patterns yields more consistent quality than rethinking plans from scratch each time.

For ground control, decide not only the number and layout of points but also in which projects check points are mandatory. Without independent verification points, objective evaluation of processing results is difficult. Although check points add effort, they are effective in preventing rework.

In processing, reduce individual operator quirks. Specify which template to use, when to perform manual checks, and what criteria constitute acceptance. For projects requiring comparison with previous results, processing philosophy and standards must be the same.

Finally, standardize report and internal sharing formats. Document not only the achieved accuracy but also the disturbance factors on the day, deviations from usual procedures, and improvements for next time—this helps align the team’s sense of accuracy.

How much variation should you tolerate?

A common operational question is “it’s slightly off—how much is acceptable?” This decision should be based on purpose, not aircraft specifications.

For example, for broad situational awareness or progress monitoring, it is more important to consistently capture general trends than to focus on small local differences. On the other hand, as-built verification, quantity calculations, or coordination with existing structures require stricter consistency. Expected accuracy levels vary by use.

Avoid the mindset that there must be zero deviation. Some variation is inevitable on site. Therefore, before adoption or at project start, decide “how much error or variation is acceptable for this use” and “what additional checks are triggered if that range is exceeded.”

Without such criteria, site operators become anxious each time and processors become overly cautious, reducing operational efficiency. With defined tolerance and additional-check procedures, unnecessary re-shooting and re-processing can be avoided.

In practice, treat drone surveying as a suitable tool within a defined scope and supplement it with ground checks where necessary. For critical heights, boundaries, or structural areas, do not rely solely on aerial results; combine them with high-precision ground positioning. This approach stabilizes operations. When considering combinations, options such as LRTK, an iPhone-mounted GNSS high-precision positioning device, can be part of the solution. Thinking about accuracy as a combination of aerial and ground roles will become increasingly important in practice.

Summary

Drone surveying does not produce exactly the same accuracy every time because six factors—weather, flight conditions, imaging settings, ground control, processing settings, and site environment—each vary slightly. This is not abnormal; since drone surveying depends on a sequence of processes, such variation naturally occurs.

What matters is not trying to eliminate variation entirely but understanding its causes and standardizing the parts your organization can control. Define required accuracy by site purpose, and standardize flight planning, imaging conditions, ground control, processing, and validation rules to steadily improve reproducibility.

Also, don’t rely on drone surveying alone for everything; combine it with high-precision ground positioning for critical points, heights, and interfaces to increase operational confidence. Efficiently capture wide areas with drones while securing key points on the ground—this balanced approach minimizes site difficulties. When considering such combinations, options like LRTK, an iPhone-mounted GNSS high-precision positioning device, are worth considering. Separating and composing the roles of air and ground will be increasingly important in future practice.