Listen to this Post
A Promising Dataset With a Significant Accuracy Problem
NASA has completed a detailed quality assessment of the Polar Geospatial Center’s (PGC) EarthDEM dataset, offering both encouraging news and an important warning for researchers who depend on high-resolution elevation data.
The assessment, released on August 8, 2026, examined two EarthDEM Digital Surface Model (DSM) products—the Strip DSM and Mosaic Tile DSM—to determine how accurately they represent the Earth’s surface. The evaluation was performed under NASA’s Commercial Satellite Data Acquisition (CSDA) program by Digital Elevation Model subject matter experts.
The results are striking.
EarthDEM performed extremely well in horizontal accuracy, with both products meeting the provider’s specifications and receiving an “Excellent” rating. But when NASA evaluated vertical accuracy, the picture changed dramatically. Instead of the provider’s stated 0.5-meter RMSEV, NASA measured errors of 5.6 meters for Strip DSM and 4.9 meters for Mosaic Tile.
That gap is not a minor technical detail.
For researchers studying terrain, glaciers, coastlines, vegetation, flooding, erosion, infrastructure or changes in Earth’s surface, vertical accuracy can determine whether a dataset is scientifically useful—or potentially misleading.
What Is EarthDEM?
EarthDEM is a high-resolution digital elevation dataset produced by the Polar Geospatial Center at The Ohio State University using imagery collected by the optical satellite fleet operated by Vantor, formerly known as Maxar.
Despite its association with
The distinction matters because EarthDEM represents an increasingly important model for modern Earth observation: commercial satellite imagery can be transformed into datasets that support scientific research at scales that would otherwise require enormous amounts of government-funded observation infrastructure.
NASA Put EarthDEM Through a Real-World Accuracy Test
NASA did not simply compare EarthDEM against theoretical specifications.
The assessment team tested the products against airborne lidar measurements, using samples from locations across the United States and Senegal.
Airborne lidar is particularly valuable for this kind of evaluation because it can provide highly detailed measurements of surface elevation and serve as a reference dataset for assessing satellite-derived terrain products.
The researchers examined EarthDEM across different surface and land-cover conditions to determine whether performance remained consistent across environments.
The outcome revealed a sharp contrast between horizontal and vertical performance.
Horizontal Accuracy Earns an “Excellent” Rating
The strongest result in the NASA assessment involved horizontal positioning.
For the EarthDEM Strip DSM, NASA measured approximately 0.5 meters RMSEH.
For the Mosaic Tile DSM, the result was even better, at approximately 0.3 meters RMSEH.
Those measurements were consistent with the specifications supplied by the Polar Geospatial Center.
NASA therefore graded the horizontal accuracy of the EarthDEM products as:
Excellent.
This is an important achievement because accurate horizontal positioning is essential for researchers attempting to align EarthDEM with other geospatial datasets.
A dataset that is horizontally misaligned can create problems when researchers compare satellite imagery, lidar, maps, infrastructure, vegetation or other geospatial layers.
EarthDEM appears to perform strongly in this area.
The Vertical Accuracy Results Tell a Different Story
The biggest concern emerged when NASA examined elevation accuracy.
The EarthDEM Strip DSM produced a 5.6-meter RMSEV.
The Mosaic Tile product produced a 4.9-meter RMSEV.
The difference between these results and the
In other words, the measured vertical error was roughly an order of magnitude greater than the stated specification.
This is not the kind of discrepancy researchers can safely ignore.
Elevation data is fundamentally different from simple imagery. A few meters of vertical error can dramatically affect scientific calculations when researchers are examining relatively subtle changes in terrain.
Cloud Cover Metadata Appears to Matter
NASA also discovered that vertical accuracy in the Strip DSM varied according to cloud-cover metadata associated with the data.
The measured vertical RMSE ranged from approximately 4.6 meters to 6.8 meters depending on the cloud-cover metadata field generated by PGC.
That finding is particularly interesting because it suggests that researchers may need to pay closer attention to metadata when selecting or filtering EarthDEM datasets.
Cloud contamination and incomplete cloud masking can introduce problems into satellite-derived elevation models.
If a researcher assumes that a
Land Cover Also Influences Performance
NASA’s evaluation found that vertical accuracy varied across different land-cover categories.
This makes intuitive sense.
Generating a digital surface model from optical satellite imagery is considerably more complicated when the sensor is observing forests, buildings, vegetation, snow, clouds, shadows or other features that obscure the actual ground.
A DSM does not necessarily represent bare-earth elevation.
Instead, a Digital Surface Model can include structures and vegetation present above the underlying terrain.
That distinction is critical.
A researcher studying glacier elevation, for example, may interpret the data differently from someone analyzing urban structures or forest canopy height.
DSM Is Not the Same as Bare-Earth Terrain
One of the most important concepts for understanding this report is the difference between a Digital Surface Model and a Digital Terrain Model.
A DSM attempts to represent the surface, potentially including buildings, trees and other objects.
A DTM, by contrast, is generally intended to represent the underlying terrain after surface objects have been removed.
This means researchers cannot automatically treat EarthDEM as a perfect representation of bare ground.
The NASA assessment specifically warns users about characteristics including poor cloud masking, missing surface features and data voids.
Those limitations may not prevent EarthDEM from being useful.
They simply mean that researchers need to understand exactly what the dataset represents before using it in scientific analysis.
NASA Still Supports Earth Science Use
Despite the vertical accuracy concerns, NASA did not conclude that EarthDEM should be rejected.
Quite the opposite.
The assessment supports the use of EarthDEM for NASA Earth science research and applications, provided researchers ensure that the dataset’s characteristics are compatible with their specific scientific objectives.
That qualification is arguably the most important conclusion in the entire report.
EarthDEM is not useless because its vertical accuracy falls short of the stated specification.
Instead, its usefulness depends on what researchers are trying to measure.
For applications where sub-meter vertical precision is essential, the dataset may require additional validation or a different elevation source.
For applications where several meters of vertical uncertainty are acceptable, EarthDEM could remain extremely valuable.
Why Horizontal Precision Still Matters
The strong horizontal results should not be overlooked.
A horizontal RMSE of 0.3 meters for Mosaic Tile is impressive and potentially valuable for a wide range of geospatial applications.
Researchers can use accurate horizontal positioning to integrate satellite-derived elevation with other spatial datasets.
This could benefit studies involving land-use changes, infrastructure, environmental monitoring, disaster assessment and regional mapping.
The challenge is making sure that horizontal precision is not mistaken for equivalent vertical precision.
A dataset can be exceptionally well positioned while still having significant elevation uncertainty.
Why This Matters for Climate Research
Elevation data plays an increasingly important role in climate science.
Scientists use terrain information to model glaciers, snow accumulation, sea-level impacts, flooding, erosion, watershed behavior and landscape changes.
In mountainous environments, even relatively small elevation errors can influence calculations involving slope, drainage and surface change.
In coastal environments, vertical errors can become even more consequential.
When researchers are attempting to determine whether a landscape sits a meter above or below a particular threshold, an elevation uncertainty of several meters could fundamentally alter the conclusion.
Flood Modeling Could Be Particularly Sensitive
Flood modeling provides another example of why vertical accuracy matters.
Flood models often depend on detailed terrain information to determine where water will accumulate and how it will move.
If elevation values are several meters higher or lower than reality, predicted flood boundaries can shift.
That does not mean EarthDEM should never be used for flood studies.
It means researchers should understand the uncertainty and, where necessary, compare EarthDEM with higher-accuracy elevation datasets.
The Dataset Could Still Be Extremely Valuable at Global Scale
There is another side to the story.
A dataset does not have to be perfect to be scientifically useful.
Coverage, accessibility and consistency can sometimes be just as important as absolute accuracy.
Commercial satellite imagery provides an opportunity to generate elevation products across enormous geographic areas.
For regions where high-quality airborne lidar is unavailable, EarthDEM could provide an important baseline dataset.
A few meters of vertical uncertainty may be preferable to having no useful elevation information at all.
NASA Also Praised the Documentation
The quality assessment did not focus solely on numerical accuracy.
NASA also reviewed documentation associated with EarthDEM, including information published on the Polar Geospatial Center website, peer-reviewed research papers from researchers at Ohio State University’s Byrd Polar and Climate Research Center, and material available through the PGC GitHub repository.
The assessment concluded that EarthDEM is “well documented” overall.
Most of the information needed to understand the product was available through these resources.
Good documentation is not a trivial advantage.
For scientific datasets, researchers need to understand how data was produced, processed, validated and distributed before they can confidently interpret the results.
The Importance of Transparent Metadata
The findings also reinforce the importance of metadata.
A numerical elevation value without information about its uncertainty, source conditions, processing history and limitations can easily be misinterpreted.
The cloud-cover-related differences observed by NASA demonstrate why metadata should be treated as part of the dataset rather than as optional background information.
For researchers building automated Earth observation pipelines, metadata can become an essential filtering mechanism.
Commercial Satellites Are Changing Earth Science
The EarthDEM assessment also illustrates a much larger transformation taking place in Earth observation.
NASA and other government organizations increasingly have access to enormous quantities of imagery collected by commercial satellite constellations.
Instead of relying exclusively on government-operated spacecraft, scientists can supplement traditional observations with commercial data.
The CSDA program exists largely because NASA sees potential in this ecosystem.
Commercial imagery can provide additional spatial coverage, frequent observations and new opportunities for scientific applications.
But greater access also creates a new responsibility: independent validation.
Why Independent Validation Matters
A data
Independent validation determines how it actually performs in practical conditions.
The difference between those two concepts is exactly why NASA’s assessment is valuable.
Without independent testing, researchers could assume that a 0.5-meter vertical accuracy specification meant their measurements would routinely achieve that level of precision.
NASA’s airborne-lidar comparisons show that this assumption would not be safe.
Independent quality assessment protects researchers from turning specifications into unsupported scientific conclusions.
What Researchers Should Do Before Using EarthDEM
Researchers considering EarthDEM should begin by defining the accuracy requirements of their project.
If horizontal positioning is the primary concern, the NASA results are encouraging.
If vertical measurements are central to the research question, considerably more caution is required.
Researchers should inspect metadata, identify land-cover conditions, look for cloud-related problems and determine whether data voids or missing surface features could influence their conclusions.
Where possible, EarthDEM should also be cross-validated against independent elevation datasets.
EarthDEM Should Be Treated as a Context-Dependent Tool
The most reasonable interpretation of the NASA report is not that EarthDEM is either “good” or “bad.”
It is more nuanced.
EarthDEM appears to be a strong horizontal geospatial product with significant vertical limitations.
That makes it a specialized scientific resource rather than a universally reliable elevation reference.
The correct question is not whether EarthDEM is accurate.
The correct question is:
Accurate enough for what?
That distinction should guide every future application.
Deep Analysis
Why RMSE Matters
Root Mean Square Error, or RMSE, measures the magnitude of differences between observed and reference measurements.
For horizontal accuracy, NASA reported:
Strip DSM:
RMSEH ≈ 0.5 m
Mosaic Tile:
RMSEH ≈ 0.3 m
For vertical accuracy:
Strip DSM:
RMSEV ≈ 5.6 m
Mosaic Tile:
RMSEV ≈ 4.9 m
The most important analytical observation is the enormous difference between horizontal and vertical performance.
A Simple Accuracy Comparison
Researchers working with geospatial data can calculate basic error statistics with tools such as GDAL.
For example:
gdalinfo earthdem.tif
This provides basic information about the raster, including its coordinate system, resolution and dimensions.
To inspect metadata:
gdalinfo -mdd all earthdem.tif
To compare an EarthDEM raster with a reference elevation dataset, researchers can first ensure that both datasets use compatible projections and grids.
A common preprocessing operation is reprojection:
gdalwarp \n-t_srs EPSG:4326 \nearthdem.tif \nearthdem_wgs84.tif
Researchers should not blindly use this command in production, however. The target coordinate system should be selected according to the geographic region and scientific application.
Calculating Elevation Differences
After aligning the datasets, raster subtraction can reveal differences between EarthDEM and a reference dataset.
For example:
gdal_calc.py \n-A earthdem.tif \n-B reference_dem.tif \n--calc="A-B" \n--outfile=elevation_difference.tif
The resulting raster can then be statistically analyzed.
A simple Python workflow might look like:
Run import rasterio import numpy as np
with rasterio.open("elevation_difference.tif") as src:
errors = src.read(1).astype(float)
nodata = src.nodata
if nodata is not None: errors = errors[errors != nodata]
errors = errors[np.isfinite(errors)]
rmse = np.sqrt(np.mean(errors 2)) mean_error = np.mean(errors) median_error = np.median(errors)
print("RMSE:", rmse)
print("Mean error:", mean_error)
print("Median error:", median_error)
This type of analysis can help researchers determine whether errors are random, systematic or concentrated in specific environments.
Bias Is Just as Important as RMSE
RMSE alone does not tell the entire story.
A dataset could have a significant systematic vertical bias in one direction.
For example, if an elevation model consistently overestimates terrain height, the mean error would reveal that tendency.
Researchers should therefore examine:
Mean Error
Median Error
RMSE
Standard Deviation
Maximum Error
Minimum Error
Percentiles
Looking only at one accuracy metric can hide important characteristics of the dataset.
Land-Cover Segmentation Is Essential
The NASA
A useful workflow would divide the study region into categories such as:
Urban
Forest
Grassland
Agriculture
Bare Ground
Snow/Ice
Wetlands
Coastal Areas
Researchers can then calculate RMSE independently for each category.
That approach provides a much more realistic picture of how EarthDEM performs in a particular research environment.
Cloud Masking Deserves Special Attention
Cloud contamination is another critical factor.
Optical satellite imagery cannot reliably observe terrain through opaque clouds.
A dataset that contains poor cloud masking may contain elevation artifacts or incomplete surface information.
Researchers should therefore inspect available cloud-related metadata before using individual scenes.
Automated processing pipelines should consider rejecting or flagging scenes with problematic metadata rather than treating every pixel as equally trustworthy.
Missing Data Can Create Hidden Problems
Data voids are another issue highlighted by NASA.
A void may appear harmless if it occupies only a small percentage of a study region.
But if that void happens to overlap a scientifically important location, it can bias the entire analysis.
Researchers should calculate the percentage of valid pixels before beginning major processing.
A simple GDAL inspection can help:
gdalinfo -stats earthdem.tif
For more advanced workflows, researchers can generate a valid-data mask and calculate coverage statistics programmatically.
DSM Versus DTM Must Remain Clear
Perhaps the most important conceptual warning is that EarthDEM is a DSM.
Researchers should not automatically interpret every elevation value as ground elevation.
Trees, buildings and other surface objects can influence the measured surface.
For urban research, that may actually be useful.
For terrain modeling, it may be problematic.
For vegetation studies, it could potentially become an advantage.
The same dataset can therefore be highly valuable in one discipline and unsuitable in another.
Reproducibility Should Be Built Into Research
Researchers using EarthDEM should document:
Dataset version
Acquisition information
Processing date
Projection
Resolution
Cloud metadata
NoData handling
Filtering criteria
Reference dataset
Validation method
Accuracy statistics
This makes the resulting research easier to reproduce and audit.
It also prevents future users from assuming that the results apply universally to every EarthDEM scene.
NASA’s Assessment Has a Broader Message
The larger lesson extends beyond EarthDEM.
As commercial satellite data becomes more important to scientific research, independent validation will become increasingly necessary.
A satellite dataset can have impressive resolution and enormous geographic coverage while still containing errors that matter for specific scientific applications.
High resolution does not automatically mean high accuracy.
More pixels do not automatically mean better measurements.
And commercial data does not automatically mean commercial-grade precision.
The Future of Earth Observation Will Depend on Data Fusion
The most promising approach may not be choosing between commercial and government datasets.
Instead, researchers can combine them.
EarthDEM could provide broad spatial coverage while airborne lidar, government elevation products or other reference datasets provide localized high-accuracy validation.
This creates a layered Earth observation ecosystem in which different datasets serve different roles.
That approach could be especially powerful in regions where expensive lidar surveys are only available for limited areas.
EarthDEM Could Become More Useful With Better Quality Controls
The NASA assessment also provides a roadmap for improvement.
Better cloud masking, improved handling of surface features, stronger vertical correction methods and clearer uncertainty information could make future versions significantly more useful.
The current limitations should therefore be viewed not only as weaknesses but also as opportunities for refinement.
Independent assessments like this one can help data providers understand where their products need additional attention.
What Undercode Say:
- The Report Is More Important Than a Simple Accuracy Score
NASA’s assessment should not be reduced to the headline number.
The real story is the enormous gap between horizontal and vertical performance.
2. Horizontal Accuracy Is a Major Strength
A Mosaic Tile RMSEH of approximately 0.3 meters is a strong result.
It suggests that EarthDEM can align effectively with other geospatial datasets.
3. Vertical Accuracy Is the Main Warning
The approximately 4.9-meter and 5.6-meter RMSEV measurements are considerably higher than the stated 0.5-meter specification.
That deserves serious attention.
4. Researchers Should Not Panic
The report does not say EarthDEM has no scientific value.
NASA explicitly supports its use when its limitations are compatible with the intended application.
5. Context Determines Data Quality
A four-meter error can be irrelevant for one study and catastrophic for another.
Scientific usefulness is therefore application-dependent.
- Cloud Metadata Could Become a Quality Signal
The observed variation from approximately 4.6 to 6.8 meters suggests cloud-related metadata may provide valuable clues about expected performance.
7. Surface Type Matters
Forests, buildings, snow and other surface characteristics can complicate satellite-derived elevation measurements.
- DSM Users Need to Understand What They Are Measuring
Researchers should remember that surface models are not automatically bare-earth terrain models.
- The Strong Horizontal Results Should Not Be Ignored
The excellent horizontal performance makes EarthDEM potentially valuable for geospatial alignment and spatial analysis.
10. Vertical Analysis Requires Independent Validation
Projects dependent on elevation precision should compare EarthDEM against an independent reference dataset whenever possible.
11. Airborne Lidar Remains Extremely Valuable
Lidar provides an important benchmark for testing satellite-derived elevation products.
- Commercial Satellite Imagery Is Becoming Scientific Infrastructure
The EarthDEM case demonstrates how commercial satellite imagery can feed large-scale scientific datasets.
13.
The program allows NASA to investigate whether commercial data can complement traditional Earth observation resources.
- More Data Does Not Automatically Mean Better Data
Researchers must distinguish between coverage, resolution, accuracy and uncertainty.
15. Documentation Matters
NASA’s positive assessment of EarthDEM documentation is an important strength.
16. Open Technical Resources Help Researchers
The availability of product information, publications and GitHub resources improves transparency.
- Metadata Should Be Treated as Scientific Data
Metadata can determine whether an individual scene is suitable for a particular analysis.
18. Data Voids Need Explicit Handling
Missing pixels should never be silently treated as valid measurements.
19. Cloud Contamination Can Create Hidden Errors
Poor cloud masking can affect downstream analysis even when the problem is not visually obvious.
20. Researchers Should Avoid Blind Automation
Automated pipelines should include quality-control checks before processing EarthDEM at scale.
- The Product May Be Excellent for Some Applications
Regional mapping, spatial alignment and broad Earth observation applications may benefit considerably from EarthDEM.
- Other Applications May Require Better Vertical Precision
Engineering, detailed flood modeling and precision terrain analysis may demand stronger elevation accuracy.
23. Scientific Conclusions Must Reflect Uncertainty
Researchers should report the limitations of their elevation source rather than presenting measurements as absolute truth.
24. Accuracy Specifications Need Independent Verification
Provider specifications are useful, but independent validation is what builds scientific confidence.
- NASA Has Provided a Valuable Reality Check
The report demonstrates the importance of testing geospatial products under real-world conditions.
- EarthDEM Should Not Be Judged by One Number
A single RMSE figure cannot capture all characteristics of a complex elevation dataset.
27. Error Distribution Matters
Researchers should investigate whether errors are random, systematic or concentrated in specific environments.
- Bias Can Be More Dangerous Than Random Noise
A systematic elevation offset can produce consistently misleading scientific conclusions.
- Data Fusion May Offer the Best Solution
Combining EarthDEM with higher-accuracy datasets could provide both coverage and precision.
30. Local Validation Can Strengthen Global Research
Researchers can use high-quality reference areas to understand how EarthDEM behaves elsewhere.
31. Future EarthDEM Versions Could Improve
The identified weaknesses provide clear targets for future processing improvements.
- Better Cloud Handling Could Have a Major Impact
Improving cloud detection and masking could reduce some of the observed problems.
33. Clearer Uncertainty Information Would Help
Researchers need to know not just the elevation value but how confident they should be in it.
34. Commercial Data Requires Scientific Oversight
Commercial availability should never eliminate independent quality control.
35.
Government agencies can evaluate private-sector-derived datasets before encouraging broader scientific use.
36. Earth Science Is Becoming Increasingly Data-Driven
Modern research increasingly depends on combining satellite imagery, lidar, AI, GIS and large-scale computational analysis.
- Quality Control Must Scale With Data Volume
As datasets become larger, automated validation becomes increasingly important.
- Researchers Should Match Data to the Question
The right dataset is the one whose uncertainty fits the scientific objective.
- EarthDEM Is Neither a Failure nor a Perfect Product
It is a useful dataset with measurable limitations.
40. The Biggest Lesson Is Simple
Know your error before trusting your elevation.
✅ NASA Assessment Date Is Consistent
The supplied report states that the quality assessment was issued on August 8, 2026.
The article correctly identifies
✅ Horizontal Accuracy Results Are Correct
The assessment reports approximately 0.5 meters RMSEH for Strip DSM and 0.3 meters RMSEH for Mosaic Tile.
NASA graded the horizontal accuracy “Excellent” because the results agreed with PGC specifications.
✅ Vertical Accuracy Results Are Correct
The reported vertical results are approximately 5.6 meters RMSEV for Strip DSM and 4.9 meters RMSEV for Mosaic Tile.
The assessment found that these results exceeded the provider’s stated 0.5-meter RMSEV specification, leading to a “Basic” vertical-accuracy compliance grade.
✅ Independent Lidar Validation Was Used
NASA compared EarthDEM products with airborne lidar samples from the United States and Senegal.
This provides an independent reference for assessing horizontal and vertical accuracy.
✅ NASA Still Supports Appropriate Scientific Use
The assessment does not declare EarthDEM unusable.
Instead, it supports its use for NASA Earth science research when the dataset’s characteristics and limitations are compatible with the intended application.
⚠️ EarthDEM Should Not Be Called a Conventional Commercial Product
The report specifically notes that EarthDEM is not technically a commercial product.
It is derived from imagery collected by
⚠️ Vertical Accuracy Should Not Be Interpreted as Uniform
The NASA assessment found variations associated with land cover and cloud-cover metadata.
Researchers should therefore avoid assuming that every EarthDEM pixel has identical vertical uncertainty.
Prediction
(+1) EarthDEM Will Likely Remain Valuable as a Large-Scale Scientific Dataset
The most likely outcome is that EarthDEM will continue to be used for Earth science research, particularly in applications where broad coverage and strong horizontal accuracy are more important than sub-meter vertical precision.
The NASA assessment provides researchers with something even more valuable than a simple endorsement: a clearer understanding of where the product works well and where caution is necessary.
Future versions could improve vertical accuracy through better processing, stronger cloud masking, improved surface reconstruction and more sophisticated quality-control mechanisms.
The broader trend is also difficult to ignore. As commercial satellite constellations produce increasingly large volumes of imagery, datasets derived from those observations will become more important to climate science, environmental monitoring, disaster response and geographic research.
The real opportunity is therefore not to replace EarthDEM with another dataset.
It is to combine EarthDEM with lidar, government elevation models and other satellite observations to create increasingly reliable and comprehensive representations of Earth’s changing surface.
For scientists, the message from NASA is ultimately constructive:
Use the data—but understand the uncertainty.
🕵️📝Let’s dive deep and fact‑check.
🎓 Live Courses & Certifications:
Join Undercode Academy for Verified Certifications
🚀 Request a Custom Project:
Secure, high-velocity infrastructure and disruptive technological engineering. Contact our engineering team for high-tier development and proprietary systems:
[email protected]
💎 Smart Architecture | 🛡️ Secure by Design | ⭐ Trusted by Thousands
References:
Reported By: science.nasa.gov
Extra Source Hub (Possible Sources for article):
https://www.twitter.com
Wikipedia
OpenAi & Undercode AI
Image Source:
Unsplash
Undercode AI DI v2
🔐JOIN OUR CYBER WORLD [ CVE News • HackMonitor • UndercodeNews ]
📢 Follow UndercodeNews & Stay Tuned:
𝕏 formerly Twitter 🐦 | @ Threads | 🔗 Linkedin | 🦋BlueSky | 🐘Mastodon | 📺Youtube
