CoChat
Sign Up
Cited by 79Satellite Image Processing and PhotogrammetryRemote Sensing and LiDAR ApplicationsSynthetic Aperture Radar (SAR) Applications and Techniques

The DOI matches its Crossref record; 61 of the 62 references it cites were found in a registry or an index; 21 of the 41 sources its claims cite were checked for support.

ANADEM: A Digital Terrain Model for South America

Open access
Laipelt, Leonardo, Comini de Andrade, Bruno, Collischonn, Walter, de Amorim Teixeira, Alexandre, Paiva, Rodrigo Cauduro Dias de, Ruhoff, Anderson·2024·Remote Sensing
DOI 10.3390/rs16132321
Open this paper in CoChat — free account

Abstract

Digital elevation models (DEMs) have a wide range of applications and play a crucial role in many studies. Numerous public DEMs, frequently acquired using radar and optical satellite imagery, are currently available; however, DEM datasets tend to exhibit elevation values influenced by vegetation height and coverage, compromising the accuracy of models in representing terrain elevation. In this study, we developed a digital terrain model for South America using a novel methodology to remove vegetation bias in the Copernicus DEM GLO-30 (COPDEM) model using machine learning, Global Ecosystem Dynamics Investigation (GEDI) elevation data, and multispectral remote sensing products. Our results indicate considerable improvements compared to COPDEM in representing terrain elevation, reducing average errors (BIAS) from 9.6 m to 1.5 m. Furthermore, we evaluated our product (ANADEM) by comparison with other global DEMs, obtaining more accurate results for different conditions of vegetation fraction cover and land use. As a publicly available and open-source dataset, ANADEM will play a crucial role in advancing studies that demand accurate terrain elevation representations at large scales.

Overview

AI snapshot

Written from the paper’s full text.

TL;DR

ANADEM is a new open-source digital terrain model for South America that removes vegetation bias from Copernicus GLO-30 using machine learning, GEDI LiDAR, and multispectral data. It reduced average BIAS from 9.6 m to 1.5 m and performed competitively against global DEMs.

Key findings

  • ANADEM reduced vegetation bias by 85.6% relative to COPDEM, with average BIAS decreasing from 9.6 m to 1.5 m.
  • For areas with >80% tree cover, COPDEM had RMSE 17.20 m and BIAS 15.38 m, while ANADEM had RMSE 6.94 m and BIAS 0.25 m; for 0–20% tree cover, RMSE/BIAS improved from 6.40/2.2 m to 5.83/1.67 m.
  • Against 600,000 ICESat-2 points, ANADEM had BIAS 1.50 m and RMSE 6.99 m, compared with FABDEM (1.76 m, 6.81 m), MERIT (3.88 m, 8.94 m), COPDEM (9.56 m, 12.40 m), and SRTM (10.33 m, 16.86 m); ANADEM had the lowest median error (0.75 m).
  • By MapBiomas land cover, ANADEM had the lowest bias in pastures (1.42 m), grasslands (0.45 m), croplands (0.36 m), and forests (0.4 m), while FABDEM performed better in savannas (2.17 m) and urban areas (0.17 m).

Methods

  • The study produced ANADEM for South America (17.8 million km²) by correcting COPDEM GLO-30 (30 m) with a gradient tree boost classifier in Google Earth Engine, using GEDI L2A elevation data and Landsat/Sentinel multispectral indices (NDMI, MSAVI, EVI selected via Boruta). The model used 93 tiles, 50,000 GEDI samples per tile, 200 trees, shrinkage 0.1, sampling rate 0.6, and Huber loss, followed by fractional vegetation cover, waterbody, and 3×3 median post-processing.
  • Validation used ICESat-2 ATL03 `hf_best_fit` ground elevations: 20,000 random points per tile for ANADEM and COPDEM, and a comparative assessment with 600,000 random points across South America against SRTM, COPDEM, MERIT, and FABDEM, using RMSE and BIAS.
  • Derived-product evaluation used TerraHidro with D8 flow routing and a 5 km² contribution threshold for watershed and drainage-network delineation, compared against the Ottocoded Hydrographic Base version 5.

Limitations

  • The paper states that ANADEM accuracy depends on the terrain elevation values used for training and the multispectral data representing vegetation; high cloud cover or inconsistent ICESat-2 terrain heights may introduce uncertainties.
  • The paper states that continental-scale parameters and temporal data periods may cause inconsistencies due to temporal differences in data acquisition, and that no bias correction was applied for urban buildings because of a lack of publicly available training datasets representing buildings.
  • The paper also notes potential overcorrection (increased negative differences), residual vegetation features from underestimation, seasonal vegetation variation due to differing acquisition periods, and slope/shadow effects on spectral response; an evident limitation is that land-cover validation uses MapBiomas for Brazil, so class-specific results may not represent all of South America.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/rs16132321

Cites

61 of 61 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

21 of 41 cited sources checked · 9 supported · 0 flagged · accuracy not yet measured

  • Premisecitation

    COPDEM is the most precise freely available global DEM.

    “Recent studies have demonstrated that COPDEM is the most precise freely available global DEM [31][32][33][34], motivating its use in this study.”
    • Also stated in: § 1. Introduction · boosted
      “we selected the COPDEM as the reference DEM due to recent studies identifying it as the most precise freely available global DEM [31][32][33][34].”
    Quote in the paperSubstantial claim3 of 4 sources checked
    • cites [31] Bielski, 2024, Novel Approach for Ranking DEMs: Copernicus DEM Improves One Arc Second Open Global Topography, IEEE Trans. Geosci. Remote Sens. source

      The Results and Conclusion report that CopDEM ranked best among the six evaluated global 1-arc-second DEMs when compared with a reference digital surface model. The unqualified claim that it is the most precise freely available global DEM overstates the finding: FABDEM ranked best against a digital terrain model, and rankings depend on the reference and criteria used. (LLM, conf=0.90)

      Experimental — accuracy not yet measured
    • cites [32] Li, 2020, Digital Elevation Models for Topographic Characterisation and Flood Flow Modelling along Low-Gradient, Terminal Dryland Rivers: A Comparison of Spaceborne Datasets for the Río Colorado, Bolivia, J. Hydrol. source

      The available indexed passages contain the paper’s title, keywords and citation metadata, but no findings comparing DEM precision. Searches for COPDEM, Copernicus DEM and comparative accuracy did not retrieve a passage establishing that COPDEM is the most precise freely available global DEM. (LLM, conf=0.73)

      Experimental — accuracy not yet measured
    • cites [33] Guth, 2021, LiDAR Point Cloud and ICESat-2 Evaluation of 1 Second Global Digital Elevation Models: Copernicus Wins, Trans. GIS source
    • cites [34] Purinton, B., and Bookhagen, B. (2021). Beyond Vertical Point Accuracy: Assessing Inter-Pixel Consistency in 30 m Global DEMs for the Arid Central Andes. Front. Earth Sci., 9. source

      Section 7.5 (page not provided) finds that Copernicus DEM has the highest inter-pixel consistency and most realistic topography among the DEMs compared in the arid Central Andes. The Conclusion likewise reports low pixel-to-pixel noise. These findings support its precision in that setting, but not the unqualified claim that it is the most precise freely available global DEM everywhere or by every accuracy measure. (LLM, conf=0.88)

      Experimental — accuracy not yet measured
  • Premisecitation

    DEM datasets tend to exhibit elevation values influenced by vegetation height and coverage, compromising the accuracy of models in representing terrain elevation.

    “DEM datasets tend to exhibit elevation values influenced by vegetation height and coverage, compromising the accuracy of models in representing terrain elevation.”
    • Also stated in: § 1. Introduction · boosted
      “The accuracy of DEMs is significantly influenced by the presence of vegetation [8,15,16], primarily because the most commonly used remote sensing acquisition techniques, such as interferometry (e.g., radar sensors, using X-radar instruments) and stereometry (e.g., panchromatic sensors), are unable to capture elevation information for the bare earth surface [17,18].”
    Quote in the paperSubstantial claim2 of 5 sources checked
    • cites [8] Okolie, 2022, A Systematic Review and Meta-Analysis of Digital Elevation Model (DEM) Fusion: Pre-Processing, Methods and Applications, ISPRS J. Photogramm. Remote Sens. source
    • cites [15] Thomas, 2014, Sensitivity of Digital Elevation Models: The Scenario from Two Tropical Mountain River Basins of the Western Ghats, India, Geosci. Front. source
    • cites [16] Kenward, 2000, Effects of Digital Elevation Model Accuracy on Hydrologic Predictions, Remote Sens. Environ. source
    • cites [17] Zhu, 2023, Digital Terrain, Surface, and Canopy Height Model Generation with Dual-Baseline Low-Frequency InSAR over Forest Areas, J. Geod. source
    • cites [18] Schlund, 2019, Canopy Penetration Depth Estimation with TanDEM-X and Its Compensation in Temperate Forests, ISPRS J. Photogramm. Remote Sens. source
  • Premisecitation

    Previous studies demonstrate a loss of drainage delineation accuracy in areas with high vegetation cover.

    “This coincides with previous studies demonstrating a loss of drainage delineation accuracy in areas with high vegetation cover [60,62].”
    Quote in the paperSubstantial claim1 of 2 sources checked
    • cites [60] Thommeret, 2010, Extraction of Thalweg Networks from DTMs: Application to Badlands, Hydrol. Earth Syst. Sci. source
    • cites [62] Yan, 2020, The Accuracy of Drainage Network Delineation as a Function of Environmental Factors: A Case Study in Central and Northern Sweden, Hydrol. Process. source
  • Interpretationown data

    Slope has a greater influence than aspect on the vertical accuracy of DEMs.

    “These results suggest that slope has a greater influence than aspect on the vertical accuracy of DEMs, supporting the conclusions of previous studies [61,63].”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites [61] Uuemaa, E., Ahi, S., Montibeller, B., Muru, M., and Kmoch, A. (2020). Vertical Accuracy of Freely Available Global Digital Elevation Models (ASTER, AW3D30, MERIT, TanDEM-X, SRTM, and NASADEM). Remote Sens., 12. source
    • cites [63] Wessel, 2018, Accuracy Assessment of the Global TanDEM-X Digital Elevation Model with GPS Data, ISPRS J. Photogramm. Remote Sens. source

      Section 3.4 (page unavailable) reports higher RMSE, NMAD, and LE90 for TanDEM-X terrain with slopes above 10°, supporting an effect of slope on vertical accuracy. The paper does not compare slope with aspect, so it does not establish that slope has the greater influence. (LLM, conf=0.84)

      Experimental — accuracy not yet measured
  • Premisecitation

    Other geomorphological characteristics can be altered due to changes in watershed delineation, such as the length of the main river, the number of segments, aspect, and slope.

    “Other geomorphological characteristics can also be altered due to changes in watershed delineation, such as the length of the main river, the number of segments, aspect, and slope [60,61].”
    Quote in the paperNot a substantial claim2 of 2 sources checked
    • cites [60] Thommeret, 2010, Extraction of Thalweg Networks from DTMs: Application to Badlands, Hydrol. Earth Syst. Sci. source

      The Discussion and conclusions table reports different total thalweg-network lengths for three extraction methods in the Moulin catchment. That supports the narrower point that how a network is delineated can affect measured length, but it does not establish that changes in watershed delineation alter the main river’s length, segment count, aspect, or slope. (LLM, conf=0.78)

      Experimental — accuracy not yet measured
    • cites [61] Uuemaa, E., Ahi, S., Montibeller, B., Muru, M., and Kmoch, A. (2020). Vertical Accuracy of Freely Available Global Digital Elevation Models (ASTER, AW3D30, MERIT, TanDEM-X, SRTM, and NASADEM). Remote Sens., 12. source

      The Introduction identifies slope and aspect as terrain variables derived from DEMs. It does not say that changes in watershed delineation alter those variables, main-river length, or the number of segments; the slope-and-aspect results concern DEM vertical accuracy instead. No page number is available in our copy. (LLM, conf=0.80)

      Experimental — accuracy not yet measured
  • Probably not a substantial claim. 21
  • It may state two separate claims as one. 32
  • The quote only partly states this claim. 2

  • Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.

    Premisecitation

    DEMs can be categorized into digital surface models (DSMs), which represent features above the surface, and digital terrain models (DTMs), which represent the elevation of the ground surface.

    “DEMs can be categorized into two types [19]: digital surface models (DSMs), which represent features above the surface, such as vegetation and buildings, and digital terrain models (DTMs), which represent the elevation of the ground surface.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [19] Guth, P.L., Van Niekerk, A., Grohmann, C.H., Muller, J.-P., Hawker, L., Florinsky, I.V., Gesch, D., Reuter, H.I., Herrera-Cruz, V., and Riazanoff, S. (2021). Digital Elevation Models: Terminology and Definitions. Remote Sens., 13. source

      Section 3.3 defines DSMs as DEMs recording the surface that includes vegetation or built features, and DTMs as bare-earth DEMs. But the Conclusions say DEMs can also represent non-vegetated surfaces, sensor surfaces, or combinations, so the claim's division into only two types is too restrictive. (LLM, conf=0.88)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.

    Premisecitation

    The main objective of the GEDI mission is to provide high-accuracy and high global coverage data on canopy structure, which can be used to estimate biomass and canopy height.

    “The main objective of the GEDI mission is to provide high-accuracy and high global coverage data on canopy structure, which can be used to estimate biomass [39][40][41] and canopy height [30,38,42].”
    Quote in the paperNot a substantial claim4 of 6 sources checked
    • cites [39] Qin, 2021, Carbon Loss from Forest Degradation Exceeds That from Deforestation in the Brazilian Amazon, Nat. Clim. Change source

      In “AGB losses from both deforestation and degradation” (page unavailable), the paper identifies GEDI as a spaceborne platform that measures vegetation canopy structure and is expected to aid forest monitoring. It does not state that GEDI’s main objective is high-accuracy, high-global-coverage data, or explicitly say that GEDI data are used to estimate biomass. The cited paper therefore supports the canopy-structure premise but not the full claim attached to the biomass citations. (LLM, conf=0.84)

      Experimental — accuracy not yet measured
    • cites [40] Duncanson, 2022, Aboveground Biomass Density Models for NASA’s Global Ecosystem Dynamics Investigation (GEDI) Lidar Mission, Remote Sens. Environ. source
    • cites [41] Santoro, 2021, The Global Forest Above-Ground Biomass Pool for 2010 Estimated from high-Resolution Satellite Observations, Earth Syst. Sci. Data source

      Section 5 (Discussion and conclusions; page unavailable) says future spaceborne lidar is expected to provide more detail on forest vertical structure and help improve biomass estimates. Section 2.1 (Satellite data; page unavailable) describes using ICESat/GLAS, not GEDI, to estimate forest height and support biomass retrieval. Our copy does not state GEDI’s mission objective or claim high accuracy and global coverage for its canopy-structure data. (LLM, conf=0.78)

      Experimental — accuracy not yet measured
    • cites [30] Liu, 2022, Neural Network Guided Interpolation for Mapping Canopy Height of China’s Forests by Integrating GEDI and ICESat-2 Data, Remote Sens. Environ. source
    • cites [38] Dubayah, 2020, The Global Ecosystem Dynamics Investigation: High-Resolution Laser Ranging of the Earth’s Forests and Topography, Sci. Remote Sens. source
    • cites [42] Lang, 2022, Global Canopy Height Regression and Uncertainty Estimation from GEDI LIDAR Waveforms with Deep Ensembles, Remote Sens. Environ. source

      The Introduction says GEDI is designed to measure ecosystem structure through forest-canopy profiles, and describes this paper’s use of GEDI data to estimate canopy height. But it also says the mission was expected to sample only 4% of land between 51.6° north and south on a sparse grid, so the claim of high global coverage needs qualification. The biomass citations are separate from the canopy-height citations attached to this source. (LLM, conf=0.82)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim.

    Premisecitation

    To correct the effect of vegetation on most pixels, the adopted solution is to develop an empirical model that relates vegetation height, or the magnitude of the error vegetation causes in DEMs, to the correction.

    “To correct the effect of vegetation on most pixels, the solution that has been adopted is to develop an empirical model that relates the height of the vegetation, or more precisely, the magnitude of the error that the vegetation causes in DEMs [22][23][24].”
    Quote in the paperSubstantial claim3 of 3 sources checked
    • cites [22] Paiva, 2016, A Multi-Sensor Approach towards a Global Vegetation Corrected SRTM DEM Product, Remote Sens. Environ. source
    • cites [23] Yamazaki, 2017, A High-Accuracy Map of Global Terrain Elevations, Geophys. Res. Lett. source
    • cites [24] Magruder, 2021, Digital Terrain Model Elevation Corrections Using Space-Based Imagery and ICESat-2 Laser Altimetry, Remote Sens. Environ. source

      Section 2.1 describes a regression model that predicts SRTM height differences from canopy cover and terrain slope, supporting the claim that an empirical model is used to correct elevation errors. It does not use vegetation height as a model input. The Discussion reports improved agreement at 70% of SRTM grid locations in the study site, rather than establishing a solution for most pixels across DEMs generally; page numbers are unavailable in our copy. (LLM, conf=0.86)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim. The paper’s conclusions may not depend on it.

    Premisecitation

    Previous studies have identified a significant error discrepancy in global DEMs within densely forested areas, primarily due to sensor limitations when accurately capturing terrain elevation.

    “Previous studies have identified a significant error discrepancy in global DEMs within densely forested areas [49][50][51][52], primarily due to sensor limitations when accurately capturing terrain elevation.”
    Quote in the paperNot a substantial claim1 of 4 sources checked
    • cites [49] Li, 2022, Global DEMs Vary from One to Another: An Evaluation of Newly Released Copernicus, NASA and AW3D30 DEM on Selected Terrains of China Using ICESat-2 Altimetry Data, Int. J. Digit. Earth source
    • cites [50] Ludwig, 2006, Validation of Digital Elevation Models from SRTM X-SAR for Applications in Hydrologic Modeling, ISPRS. J. Photogramm. Remote Sens. source
    • cites [51] 2020, Accuracy Assessment of the Global TanDEM-X Digital Elevation Model in a Mountain Environment, Remote Sens. Environ. source
    • cites [52] Hawker, 2019, Accuracy Assessment of the TanDEM-X 90 Digital Elevation Model for Selected Floodplain Sites, Remote Sens. Environ. source

      Section 5.2 reports substantially higher error for TanDEM-X 90 than MERIT in tree-covered floodplain areas, and the Conclusions attribute this partly to X-band radar struggling to penetrate tree canopies. The study assesses selected floodplain sites, however, and notes that errors vary among tree-cover classes; it does not establish the same discrepancy for global DEMs generally across densely forested areas. (LLM, conf=0.84)

      Experimental — accuracy not yet measured

  • It may state two separate claims as one.

    Findingown data

    ANADEM significantly reduced the influence of vegetation on elevation values, resulting in a smoother surface than COPDEM.

    “Overall, ANADEM significantly reduced the influence of vegetation on the elevation values, resulting in a smoother surface than COPDEM.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    COPDEM results showed an average RMSE of 12.40 m and an average BIAS of 9.56 m.

    “COPDEM results showed an average RMSE of 12.40 m and an average BIAS of 9.56 m.”
    Quote in the paperNot a substantial claim

  • It may state two separate claims as one.

    Findingown data

    ANADEM was found to be capable of decreasing error differences related to vegetation bias, resulting in a more centered normal distribution.

    “On the other hand, ANADEM was found to be capable of decreasing error differences related to vegetation bias, resulting in a more centered normal distribution.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    Differences between the reference length of the main river and the models were 7% and 8%, while the number of segments increased by 10% and 12% for ANADEM and COPDEM, respectively.

    “In our example, the differences between the reference length of the main river and the models were 7% and 8%, while the number of segments increased by 10% and 12% for ANADEM and COPDEM, respectively.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: The reference drainage network and basin (REF) is accurate.

  • It may state two separate claims as one.

    Findingown data

    At approximately 80% vegetation cover, COPDEM results are concentrated around a bias close to 20 m, whereas ANADEM results concentrate at values much closer to zero.

    “At approximately 80% vegetation cover, COPDEM results are concentrated around a bias close to 20 m, whereas ANADEM results concentrate at values much closer to zero.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    ANADEM produced watershed delineation results closest to the reference, with estimated basin area 92,292 km2 versus REF 92,424 km2 and COPDEM 93,498 km2.

    “The quantitative analysis of the derived products indicates that ANADEM produced results closest to the REF, as the estimated basin area was 92,424 km 2 , whereas for ANADEM and COPDEM the delimitated area resulted in 92,292 and 93,498 km 2 , respectively.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The reference drainage network and basin (REF) is accurate.

  • It may state two separate claims as one.

    Findingreasoning

    Visual inspection demonstrates that ANADEM-delineated streams are more realistic and smoother than those delineated using COPDEM, reflecting lower vegetation bias in ANADEM.

    “Visual inspection also demonstrates that ANADEM-delineated streams are more realistic and smoother than those delineated using COPDEM; this reflected the lower vegetation bias in ANADEM.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    MERIT DEM shows a clear improvement in removing vegetation bias from the SRTM model, reducing RMSE and BIAS to 8.94 m and 3.88 m, respectively.

    “MERIT DEM shows a clear improvement in removing vegetation bias from the SRTM model, reducing RMSE and BIAS to 8.94 m and 3.88 m, respectively.”
    Quote in the paperNot a substantial claim

  • It may state two separate claims as one.

    Findingown data

    In an area of the Amazon with extensive land cover changes, effective removal of vegetation in ANADEM is evident, leading to improvements in representation of the terrain surface, although some vegetation features are still present.

    “The effective removal of vegetation in ANADEM is once again evident, leading to improvements in the representation of the terrain surface, although some vegetation features are still present.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    SRTM has an RMSE of 16.86 m and a bias of 10.33 m.

    “showing an RMSE of 16.86 m and a bias of 10.33 m.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    NDMI, MSAVI, and EVI demonstrated superior discrimination of vegetation and effectively aggregated information into the model.

    “Based on these results, we selected three specific indices, namely the Normalized Difference Moisture Index (NDMI), the Modified Soil Adjusted Vegetation Index (MSAVI), and the Enhanced Vegetation Index (EVI), which demonstrated superior discrimination of vegetation and effectively aggregated information into the model.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Spectral indices aggregate spatial information related to vegetation and generally improve the final results of the classifier.

  • It may state two separate claims as one.

    Findingown data

    Although certain locations in ANADEM exhibited residual vegetation features due to underestimation of vegetation bias, the residual biases were considerably lower than those observed in COPDEM.

    “Although certain locations in ANADEM exhibited residual vegetation features due to an underestimation of the vegetation bias, the residual biases were considerably lower than those observed in COPDEM.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    The vertical error in slope, compared with ICESat-2 data, increased proportionally with increasing slope for both COPDEM and ANADEM, although the improvements for ANADEM were more significant.

    “The vertical error in slope, when compared with ICESat-2 data, showed a proportional increase with increasing slope for both models, although the improvements for ANADEM were more significant.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingnone given

    MERIT decreased vegetation bias considerably (5.60 m) despite having a higher value than ANADEM and FABDEM.

    “MERIT decreased vegetation bias considerably (5.60 m) despite having a higher value than ANADEM and FABDEM.”
    Quote in the paperNot a substantial claim

  • It may state two separate claims as one.

    Findingown data

    ANADEM showed the lowest bias in pastures (1.42 m), grasslands (0.45 m), croplands (0.36 m), and forests.

    “ANADEM showed the lowest bias in pastures (1.42 m), grasslands (0.45 m), croplands (0.36 m), and forests”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    Vegetation bias was lowest for ANADEM (1.50 m), whereas FABDEM showed an overestimation of 1.76 m.

    “The vegetation bias was lowest for ANADEM (1.50 m), whereas FABDEM showed an overestimation of 1.76 m.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    For areas with greater than 50% tree cover, SRTM featured the highest errors (15.78 m), followed by COPDEM (14.85 m).

    “SRTM featured the highest errors (15.78 m), followed by COPDEM (14.85 m).”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    The percentage RMSE for ANADEM varied from 2.26 m (croplands) to 7.09 m (forests), whereas those for COPDEM varied between 2.39 (croplands) and 16.44 m (forests).

    “The percentage RMSE for ANADEM varied from 2.26 m (croplands) to 7.09 m (forests), whereas those for COPDEM varied between 2.39 (croplands) and 16.44 m (forests).”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    FABDEM had an RMSE of 6.81 m, whereas ANADEM had a slightly higher RMSE of 6.99 m.

    “Our results determined an RMSE of 6.81 m for FABDEM, whereas ANADEM showed a slightly higher RMSE of 6.99 m.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    FABDEM showed better results for savannas (2.17 m) and urban areas (0.17 m).

    “whereas FABDEM showed better results for savannas (2.17 m) and urban areas (0.17 m).”
    Quote in the paperNot a substantial claim

  • The quote only partly states this claim.

    Findingown data

    The most accurate results for areas with greater than 50% tree cover were obtained by ANADEM and FABDEM, with an average BIAS of 1.49 m and 2.16 m, respectively.

    “The most accurate results were obtained by ANADEM and FABDEM, with an average BIAS of 1.49 m and 2.16 m, respectively.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim. The paper’s conclusions may not depend on it.

    Premisereasoning

    Machine learning has emerged as a powerful tool for generating high-accuracy DTMs comprising large amounts of remote sensing data, including LiDAR altimetry from ICESat and GEDI.

    “Consequently, machine learning has emerged as a powerful tool for generating high-accuracy DTMs comprising large amounts of remote sensing data, including altimetry measurements acquired using LiDAR instruments, such as ICESat and GEDI.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.

    Interpretationcitation

    Terrain characteristics can influence the model's accuracy, as slope effects may lead to shadowed regions, influencing the pixel spectral response.

    “Additionally, terrain characteristics can influence the model's accuracy, as slope effects may lead to shadowed regions [57], influencing the pixel spectral response.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [57] Isikdogan, 2017, Surface Water Mapping by Deep Learning, IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. source

  • It may state two separate claims as one.

    Interpretationnone given

    ANADEM was developed for applications by the Brazilian water agency and its calibration is specifically designed to improve its accuracy over South America.

    “ANADEM was developed for applications by the Brazilian water agency and offers several advantages, notably its calibration, which is specifically designed to improve its accuracy over South America.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim. It may state two separate claims as one.

    Study design and conductnone given

    Results from the classifier were post-processed to correct exaggerated vegetation removal and minimize noise.

    “Results from the classifier were post-processed to correct exaggerated vegetation removal and minimize noise.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim. The paper’s conclusions may not depend on it.

    Premisereasoning

    Landsat 8 was used due to its compatibility with the period of elevation data acquisition from COPDEM.

    “The Landsat 8 satellite (launched in 2013) was used due to its compatibility with the period of elevation data acquisition from COPDEM.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim.

    Study design and conductown data

    Fractional vegetation cover mosaic was used to ensure locations with sparse vegetation underwent lower vegetation removal than dense vegetation.

    “We utilized the fractional vegetation cover (fvc) mosaic generated with Sentinel-2 multispectral data to ensure that locations with sparse vegetation underwent lower vegetation removal than dense vegetation.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim.

    Interpretationnone given

    The model workflow requires only terrain elevation and multispectral data to correct vegetation bias from the selected DEM.

    “The model workflow in this study was developed to facilitate applicability to different study areas, requiring only terrain elevation and multispectral data to correct vegetation bias from the selected DEM (COPDEM).”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Premisecitation

    The gradient tree boost classifier has several advantages over the traditional random forest model due to boosting aimed at minimizing errors, and demonstrated greater robustness with low sensitivity to outliers and reduced processing time.

    “The gradient tree boost classifier [48] is based on tree decision learning methods and has several advantages over the traditional random forest model due to its optimization techniques (boosting) aimed at minimizing errors in the prediction process.”
    • Also stated in: § 2.8. Vegetation Bias Removal Algorithm · boosted
      “Additionally, the gradient tree boost method demonstrated greater robustness than the traditional random forest model in generating accurate results with low sensitivity to outliers and a reduced processing time.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [48] Friedman, 2001, Greedy Function Approximation: A Gradient Boosting Machine, Ann. Stat. source

  • Probably not a substantial claim. The paper’s conclusions may not depend on it. The quote only partly states this claim.

    Study design and conductreasoning

    To ensure data quality, only multispectral images with less than 20% cloud cover were selected.

    “To ensure data quality, a filtering process was implemented to select only images with less than 20% cloud cover.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim. The paper’s conclusions may not depend on it.

    Study design and conductnone given

    A smoothing focal median filter with a 3 × 3 pixel was applied to reduce noise after subtracting vegetation bias.

    “After subtracting the vegetation bias estimated from the COPDEM, a smoothing focal median filter with a 3 × 3 pixel was applied to reduce noise.”
    Quote in the paperNot a substantial claim

  • It may state two separate claims as one.

    Interpretationreasoning

    Vegetation removal led to a reduction in errors for ANADEM without clearly indicating changes in dependency on slope and aspect.

    “Hence, vegetation removal led to a reduction in errors for ANADEM without clearly indicating changes in dependency on slope and aspect.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim. The paper’s conclusions may not depend on it.

    Premisecitation

    Machine learning advancements have significantly enhanced the utilization of extensive data to create new topographic datasets based on existing remote sensing sources.

    “In this context, machine learning advancements have significantly enhanced the utilization of extensive data [25] to create new topographic datasets based on existing remote sensing sources [21,[26][27][28][29].”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [25] Li, 2021, A Deep Learning Semantic Template Matching Framework for Remote Sensing Image Registration, ISPRS J. Photogramm. Remote Sens. source
    • cites [21,[26][27][28][29]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.

    Study design and conductreasoning

    The study area was divided into 93 tiles to avoid processing errors and improve the geographical representativeness of vegetation characteristics.

    “Initially, we divided the study area into 93 (5 • × 5 • ) tiles to avoid processing errors and improve the geographical representativeness of vegetation characteristics.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim.

    Study design and conductown data

    The performance statistics for each model were based on 600,000 randomly selected points.

    “Table 3 shows the performance statistics for each model based on 600,000 randomly selected points.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim. The paper’s conclusions may not depend on it.

    Study design and conductnone given

    The highest quality globally available Landsat data from collection 2, which incorporates atmospherically corrected land surface reflectance, were selected.

    “We also selected the highest quality data globally available, from collection 2, which incorporates atmospherically corrected land surface reflectance.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim. The paper’s conclusions may not depend on it.

    Interpretationreasoning

    Uncertainties in vegetation removal in ANADEM may occur due to seasonal variations in vegetation, influenced by differences in time periods of acquisition between COPDEM and multispectral data.

    “Uncertainties in vegetation removal in ANADEM may also occur due to seasonal variations in vegetation, which are influenced by the differences in the time periods of acquisition between COPDEM and multispectral data from satellites used to represent vegetation.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim. The paper’s conclusions may not depend on it.

    Study design and conductreasoning

    To ensure data quality, lower-quality GEDI data points were excluded by a filtering process.

    “To ensure data quality, we applied a filtering process to exclude lower-quality GEDI data points.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.

    Study design and conductreasoning

    Waterbody elevation values remained equal to COPDEM, and an additional step was applied for locations adjacent to water bodies to avoid lowered values and ensure hydrological consistency.

    “Waterbody elevation values remained equal to those used by COPDEM, and an additional step was applied for locations adjacent to water bodies in order to avoid lowered values and ensure hydrological consistency.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim. The paper’s conclusions may not depend on it.

    Study design and conductnone given

    Machine learning classifier parameters were defined by testing different optimization settings manually and selecting those with optimal applicability, highest accuracy, and shortest processing times.

    “We defined the machine learning classifier parameters by testing different optimization settings manually and selecting those with the optimal applicability, highest accuracy, and shortest processing times.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim. The paper’s conclusions may not depend on it.

    Interpretationreasoning

    The slightly better performance of ANADEM in urban areas may be attributed to the post-processing adopted in ANADEM to reduce noise generated by the machine learning model.

    “This may be attributed to the post-processing adopted in ANADEM to reduce noise generated by the machine learning model.”
    Quote in the paperNot a substantial claim

  • It may state two separate claims as one.

    Interpretationreasoning

    Errors may arise in areas with high cloud cover or inconsistent terrain elevation values from ICESat-2, which may introduce uncertainties in vegetation removal.

    “Although we utilized publicly available high-quality data in our study, errors may arise in areas with high cloud cover or inconsistent terrain elevation values from ICESat-2; these may introduce uncertainties in vegetation removal.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Interpretationreasoning

    The parameters and temporal data period used to generate ANADEM were defined to facilitate vegetation removal on a continental scale, which could result in inconsistencies due to temporal differences in data acquisition.

    “Secondly, the parameters and temporal data period used to generate ANADEM were defined to facilitate vegetation removal on a continental scale, which could result in inconsistencies due to temporal differences in data acquisition.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Interpretationnone given

    The accuracy of ANADEM results is directly impacted by the terrain elevation values used for model training as well as by the multispectral data representing vegetation.

    “Firstly, the accuracy of ANADEM results is directly impacted by the terrain elevation values used for model training as well as by the multispectral data representing vegetation.”
    Quote in the paperSubstantial claim
48 of 48

  • The paper relies on this without support.

    LiDAR data are not affected by vegetation.

    “Light detection and ranging data (LiDAR) are not affected by vegetation.”

  • The paper relies on this without support.

    Most studies still rely on SRTM.

    “most studies still rely on SRTM”

  • The paper relies on this without support.

    The potential of machine learning methods to remove vegetation bias is yet to be fully investigated.

    “the potential of machine learning methods to remove vegetation bias is yet to be fully investigated”

  • The paper relies on this without support.

    Many machine learning methods to remove vegetation bias are not yet openly available.

    “many such methods are not yet openly available”

  • The paper relies on this without support.

    Global DEMs are usually closer to DSMs than to DTMs.

    “Since global DEMs are usually closer to DSMs”

  • The paper relies on this without support.

    Spectral indices aggregate spatial information related to vegetation and generally improve the final results of the classifier.

    “Spectral indices aggregate spatial information related to vegetation, which generally improves the final results of the classifier.”

  • The paper relies on this without support.

    The reference drainage network and basin (REF) is accurate.

    “We utilized the Ottocoded Hydrographic Base (version 5), produced by the Brazilian Water Agency repository (https://metadados.snirh.gov.br(accessed on 25 May 2024)), as the reference drainage network and basin (REF).”

  • The paper relies on this without support.

    ICESat-2 ATL03 'hf_best_fit' provides accurate ground elevation estimates.

    “Information from ICESat-2 is available in the 'hf_best_fit' band from the ATL03 product, which provides the best-fit ground elevation estimative.”

  • The paper relies on this without support.

    SRTM and MERIT have the lowest spatial resolutions (90 m) among the models examined, and this influences their statistical performance.

    “Among the models examined, SRTM and MERIT have the lowest spatial resolutions (90 m), which influences their statistical performance.”

  • The paper relies on this without support.

    Vegetation influences derived products.

    “Because vegetation influences derived products”

  • The paper relies on this without support.

    There is a lack of publicly available datasets for model training that accurately represent buildings.

    “due to a perceived lack of publicly available datasets for model training that accurately represent buildings.”
11 of 11

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Forest %RMSE is confused with RMSE

    Minor

    The text calls 7.09 m the percentage RMSE for ANADEM in forests, but Table 4 places 7.09 under RMSE and gives the corresponding %RMSE as 3.42.

  • Positive overestimation conflicts with the stated bias sign

    Minor

    The stated BIAS formula uses observed minus model, so an overestimate of terrain elevation would have a negative bias. The paper instead describes FABDEM's positive 1.76 m bias as an overestimation.

  • Reported standard deviation exceeds RMSE for the same tree-cover group

    Minor

    For ANADEM at 81-100% tree cover, the reported standard deviation is 7.10 m while the RMSE is 6.94 m. The standard deviation of errors cannot exceed their root mean square by this amount for the large validation samples described.

3 of 3

Citations

Cites

Flagged first, then by citations

1 of the 62 references couldn’t be matched to a known paper.

Research works printed but not linked: 1

  1. Abrams, 2002, ASTER User Handbook, Version 2, Jet Propuls. Lab., 4800, 135

Cited by

Sorted by citations

Related

Metadata

Authors
Laipelt, L., Comini de Andrade, B., Collischonn, W., de Amorim Teixeira, A., Paiva, R. C. D. D., & Ruhoff, A.
Affiliations
  • Universidade Federal do Rio Grande do Sul (BR)
  • Instituto de Pesquisas Hidráulicas, Universidade Federal do Rio Grande do Sul, Porto Alegre 91501-970, RS, Brazil
  • Agência Nacional de Águas e Saneamento Básico (ANA), Brasília 70610-200, DF, Brazil
Corresponding author
Laipelt, Leonardo
ORCID
Venue
Remote Sensing
Publisher
MDPI AG
Published
2024-06-25
Volume
16
Issue
13
Pages
2321
ISSN
2072-4292 (ISSN-L)
DOI
10.3390/rs16132321
OpenAlex
W4399999092
Cited by
79
References
62
Citations by year
2026: 64 · 2025: 13 · 2024: 2
FWCI
8.57
Type
journal article
Language
en
Open access
Licence
Locations
  • Remote Sensing (publishedVersion, cc-by)
  • Preprints.org (acceptedVersion, cc-by)
  • LA Referencia (Red Federada de Repositorios Institucionales de Publicaciones Científicas) (submittedVersion, cc-by-nc-sa)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
Brazilian Water Agency (ANA) (TED-05/2019-ANA)
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Satellite Image Processing and Photogrammetry
  • Life Sciences › Agricultural and Biological Sciences › Forestry › Remote Sensing and LiDAR Applications
  • Physical Sciences › Earth and Planetary Sciences › Geophysics › Synthetic Aperture Radar (SAR) Applications and Techniques
Keywords
digital terrain model, digital elevation model, machine learning, land use, South America, terrain elevation

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-11-08, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-11-08 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review