CoChat
Sign Up
Cited by 70Satellite Image Processing and PhotogrammetryGNSS positioning and interferenceGeodetic Measurements and Engineering Structures

Vertical accuracy evaluation of freely available latest high-resolution (30 m) global digital elevation models over Cameroon (Central Africa) with GPS/leveling ground control points.

Open access
Yap, Loudi, Kandé, Ludovic Houetchak, Nouayou, Robert, Kamguia, Joseph, Ngouh, Nasser Abdou, Makuate, Marie Brigitte·2019·International Journal of Digital Earth
DOI 10.1080/17538947.2018.1458163
Open this paper in CoChat — free accountOpen PDF

Abstract

Digital Elevation Models (DEMs) contain topographic relief data that are vital for many geoscience applications. This study relies on the vertical accuracy of publicly available latest high-resolution (30 m) global DEMs over Cameroon. These models are (1) the ALOS World 3D-30 m (AW3D30), (2) the Shuttle Radar Topography Mission 1 Arc-Second C-Band Global DEM (SRTM 1) and (3) the Advanced Spaceborne Thermal Emission and Reflection Global DEM Version 2 (ASTER GDEM 2). After matching their coordinate systems and datums, the horizontal positional accuracy evaluation was carried out and it shows that geolocation errors significantly influence the vertical accuracy of global DEMs. After this, the three models are compared among them, in order to access random and systematic effects in the elevation data each of them contains. Further, heights from 555 GPS/leveling points distributed all over Cameroon are compared to each DEM, for their vertical accuracy determination. Traditional and robust statistical measures, normality test, outlier detection and removal were used to describe the vertical quality of the DEMs. The test of the normality rejected the hypothesis of normal distribution for all tested global DEMs. Overall vertical accuracies obtained for the three models after georeferencing and gross error removal in terms of Root Mean Square (RMS) and Normalized Median Absolute Deviation (NMAD) are: AW3D30 (13.06 m and 7.75 m), SRTM 1 (13.25 m and 7.41 m) and ASTER GDEM 2 (18.87 m and 13.30 m). Other accuracy measures (MED, 68.3% quantile, 95% quantile) supply some evidence of the good quality of AW3D30 over Cameroon. Further, the effect of land cover and slope on DEM vertical accuracy was also analyzed. All models have proved to be worse in the areas dominated by forests and shrubs areas. SRTM 1 and AW3D30 are more resilient to the effects of the scattering objects respectively in forests and cultivated areas. The dependency of DEMs accuracy on the terrain roughness is evident. In all slope intervals, AW3D30 is performing better than SRTM 1 and ASTER GDEM 2 over Cameroon. AW3D30 is more representative of the external topography over Cameroon in comparison with two others datasets and SRTM 1 can be a serious alternative to AW3D30 for a range of DEM applications in Cameroon.

Integrity signals

Registry

Crossref · 10.1080/17538947.2018.1458163

Cites

79 of 80 references verified0 citation errors · 1 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

13 of 43 cited sources checked · 10 supported · 0 flagged · accuracy not yet measured

  • Findingcitation

    Geolocation errors significantly influence the vertical accuracy of global DEMs, especially in areas of steep slopes.

    “After matching their coordinate systems and datums, the horizontal positional accuracy evaluation was carried out and it shows that geolocation errors significantly influence the vertical accuracy of global DEMs.”
    • Also stated in: § 4.2. Horizontal georeferencing estimation · hedged
      “Even if the scope of this study is not the assessment of geolocation errors, it is important to note that errors in horizontal positional accuracy may result the vertical errors in the DEMs, especially in areas of steep slopes.”
    Quote in the paperSubstantial claim1 of 3 sources checked
    • cites Hirt, Filmer, and Featherstone 2010 source
    • cites Tachikawa et al. 2011
    • cites Hu et al. 2017

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

  • Premisecitation

    The negative mean error for AW3D30 obtained in this study was confirmed by the validation of AW3D30 DSM datasets conducted in Japan and other foreign countries, showing a mean error of -0.44 m in these areas.

    “The negative mean error for AW3D30 obtained in this study was confirmed by the validation of AW3D30 DSM datasets conducting in Japan area and another foreign countries (Tadono et al. 2016) showing a mean error of -0.44 m in these areas.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites Tadono et al. 2016 source
  • Premisecitation

    Signal omission was found by Hirt, Filmer, and Featherstone (2010) and Rexer and Hirt (2014) to be the main cause for the considerable differences observed among the DEMs.

    “The latter was found by Hirt, Filmer, and Featherstone (2010); Rexer and Hirt (2014) to be the main cause for the considerable differences observed among the DEMs.”
    Quote in the paperNot a substantial claim1 of 2 sources checked
    • cites Hirt, Filmer, and Featherstone (2010) source

      In “Comparison among the models” (page not specified), the paper identifies signal omission as the cause of considerable differences involving DEM-9S in rugged terrain. The claim omits those limits and presents it as the main cause of differences among the DEMs generally; the same section also identifies stripe patterns associated with ASTER. (LLM, conf=0.87)

      Experimental — accuracy not yet measured
    • cites Rexer and Hirt (2014) source
  • Findingown data

    The distribution of vertical errors for all tested global DEMs is not normal.

    “The test of the normality rejected the hypothesis of normal distribution for all tested global DEMs.”
    Quote in the paperNot a substantial claim
  • Findingreasoning

    Significant discrepancies are not visible around the northern parts of Cameroon where the topography is relatively flat.

    “However, significant discrepancies are not visible around the northern parts of Cameroon where the topography is relatively flat.”
    Quote in the paperNot a substantial claim
  • May combine two claims. 42
  • Classified as probably not a substantial claim. 16
  • The quote only partly states this claim. 2

  • May combine two claims.

    Premisecitation

    The AW3D30, SRTM 1, and ASTER GDEM 2 datasets used in this study are Digital Surface Models (DSMs) that include heights of surface features such as canopy.

    “In this study, the space-based AW3D30, SRTM 1 and ASTER GDEM 2 data sets used are Digital Surface Models (DSMs), as they provide heights of surface features, including the canopy.”
    Quote in the paperSubstantial claim1 of 2 sources checked
    • cites Tadono et al. 2016 source

      Section 3.4 identifies AW3D30 as a DSM. The Introduction describes SRTM and ASTER GDEM as digital elevation models, including the 1-arcsecond SRTM release and ASTER GDEM’s second version; it does not identify those datasets as DSMs or say that any of the three provides canopy heights. Page numbers are unavailable in our copy. (LLM, conf=0.87)

      Experimental — accuracy not yet measured
    • cites Takaku et al. 2016 source

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation

    The Normalized Median Absolute Deviation (NMAD) is proportional to the median of the absolute differences between vertical errors and the median error; it can be considered an estimate for the Standard Deviation (STD) more resilient to outliers in the dataset.

    “The Normalized Median Absolute Deviation (NMAD) is proportional to the median of the absolute differences between vertical errors and the median error.It can be considered as an estimate for the Standard Deviation (STD) more resilient to outliers in the dataset (Höhle and Höhle 2009;Alidoost and Samadzadegan 2013).”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites Höhle and Höhle 2009 source
    • cites Alidoost and Samadzadegan 2013 source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    The 95% sample quantile of vertical errors means that 95% of the errors have a magnitude within the interval [0, Q(0.95)]; the remaining 5% can be of any value, making the measure robust against up to 5% blunders.

    “The 95% sample quantile of vertical errors literally means that 95% of the errors have a magnitude within the interval [0, Q (0.95)].The remaining 5% of the errors can be of any value making the measure robust against up to 5% blunders ( (Höhle and Höhle 2009).”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites Höhle and Höhle 2009 source

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation

    The median (50% quantile) is a robust estimator for a systematic shift of the DEM; it is less sensitive to outliers than the mean error and provides a better distributional summary for skew distributions.

    “The median (50% quantile) is a robust estimator for a systematic shift of the DEM.It is less sensitive to outliers in the data than the mean error and provides a better distributional summary for skew distributions (Höhle and Höhle 2009).”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites Höhle and Höhle 2009 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Previous studies reported an overall negative bias for ASTER GDEM 2 with respect to GCPs.

    “Previous studies reported an overall negative bias for ASTER GDEM 2 with respect to GCPs (Hirt, Filmer, and Featherstone 2010;Gesch et al. 2012;Alidoost and Samadzadegan 2013;Li et al. 2013Li et al. , 2016;;Athmania and Achour 2014;Jain et al. 2017).”
    Quote in the paperNot a substantial claim3 of 3 sources checked
    • cites Hirt, Filmer, and Featherstone 2010 source

      In “Model validation with ground truth data,” the paper reports that ASTER heights were about 9 m too low at 6,392 GCPs; Table 5 also gives a negative mean difference for ASTER against 911 GPS GCPs. But the paper explicitly studies ASTER GDEM ver1, not GDEM 2, so it does not establish the claimed bias for version 2. (LLM, conf=0.91)

      Experimental — accuracy not yet measured
    • cites Gesch et al. 2012 source
    • cites Alidoost and Samadzadegan 2013 source

  • May combine two claims.

    Premisecitation

    Not all outliers can be detected by the 3-sigma rule and the DEM accuracy measures (ME and RMS) will therefore be inaccurate.

    “Not all of the outliers can be detected in this way and the DEM accuracy measures (ME and RMS) will therefore be inaccurate (Höhle and Höhle 2009;Alidoost and Samadzadegan 2013).”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites Höhle and Höhle 2009 source
    • cites Alidoost and Samadzadegan 2013 source

  • May combine two claims.

    Findingown data

    Overall vertical accuracies after georeferencing and gross error removal in terms of RMS and NMAD over Cameroon are AW3D30 (13.06 m and 7.75 m), SRTM 1 (13.25 m and 7.41 m), and ASTER GDEM 2 (18.87 m and 13.30 m).

    “Overall vertical accuracies obtained for the three models after georeferencing and gross error removal in terms of RMS and NMAD over Cameroon are: AW3D30 (13.06 m and 7.75 m), SRTM 1 (13.25 m and 7.41 m) and ASTER GDEM 2 (18.87 m and 13.30 m).”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Differences between DEM elevation and GCPs larger than 100 m are returns from cloud tops.
    • Rests on an assumption: The GCP elevations are accurate enough to serve as ground truth for DEM vertical accuracy evaluation.

  • May combine two claims.

    Findingnone given

    SRTM 1 and AW3D30 are more resilient to the effects of scattering objects in forests and cultivated areas, respectively.

    “SRTM 1 and AW3D30 are more resilient to the effects of the scattering objects respectively in forests and cultivated areas.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The AW3D30-SRTM comparison over Cameroon has a positive bias of 0.10 m and an RMS of 5.30 m relative to AW3D30.

    “Table 6 shows that, the statistics of AW3D30-SRTM are very similar with positive bias of 0.10 m and 5.30 m as RMS relative to AW3D30 over Cameroon.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    AW3D30 has a slight negative bias of -0.70 m before and -0.10 m after removing outliers.

    “Also, slight negative bias is observed before (-0.70 m) and after removing outliers (-0.10 m) for AW3D30.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Differences between DEM elevation and GCPs larger than 100 m are returns from cloud tops.
    • Rests on an assumption: The GCP elevations are accurate enough to serve as ground truth for DEM vertical accuracy evaluation.

  • May combine two claims.

    Findingown data

    Vertical errors in ASTER GDEM 2 show a negative mean error of -6.27 m before and -6.16 m after removing outliers.

    “Vertical errors in ASTER GDEM 2 show a negative mean error of -6.27 m and -6.16 m before and after removing outliers respectively.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Differences between DEM elevation and GCPs larger than 100 m are returns from cloud tops.
    • Rests on an assumption: The GCP elevations are accurate enough to serve as ground truth for DEM vertical accuracy evaluation.

  • May combine two claims.

    Findingcitation

    SRTM 1 exhibits the best values of RMS and NMAD in two land cover classes compared to the other DEM despite its slightly higher RMS (15.87 m) in comparison with AW3D30 (15.67 m) in the last cover class.

    “SRTM 1 exhibits the best values of RMS and NMAD in the two land cover classes compared to the other DEM despite its slight higher value of RMS (15.87 m) in comparison with AW3D30 (15.67 m) in the last cover class.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The GCP elevations are accurate enough to serve as ground truth for DEM vertical accuracy evaluation.

  • May combine two claims.

    Findingown data

    In forests and woodlands areas of Cameroon, SRTM 1 has RMS vertical error 14.87 m and NMAD 9.13 m.

    “In the forests and woodlands areas, the vertical DEM accuracy indicates RMS and NMAD of (15.07 m and 10.38 m), (14.87 m and 9.13 m) and (20.03 m and 14.43 m) respectively for AW3D30, SRTM 1 and ASTER GDEM 2.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: More than 50% of the study area in Cameroon is occupied by forests and woodlands and shrub and herbaceous lands, estimated at 69.15% and 11.23% respectively.

  • May combine two claims.

    Findingown data

    The NMAD and 68.3% quantile in SRTM 1-GPS are 7.41 m and 5.00 m, and in AW3D30-GPS are 7.75 m and 4.13 m; these are about the same in both datasets.

    “The results of comparison between NMAD and 68.3% quantile in SRTM 1-GPS (7.41 m and 5.00 m) and AW3D30-GPS (7.75 m and 4.13 m) are about the same in both datasets.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The GCP elevations are accurate enough to serve as ground truth for DEM vertical accuracy evaluation.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data

    Land cover composition over Cameroon is 17.69% cultivated areas, 69.15% forests and woodland, 11.23% shrub and herbaceous lands, and 2.88% water bodies.

    “It is to be noted that 17.69% area is under cultivated areas (agriculture and cropland), 69.15% under forests and woodland, 11.23% under shrub and herbaceous lands and 2.88% under water bodies.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    In the SRTM 1-GPS dataset, the mean error after removing outliers is 0.54 m and the median is 2.00 m, differing by 1.46 m; the STD is 13.25 m and the NMAD is 7.41 m, differing by 5.84 m.

    “In the SRTM 1-GPS dataset, the mean error (0.54 m) obtained after removing outliers and median (2.00 m) differ by 1.46 m while the STD (13.25 m) and NMAD (7.41 m) differ by 5.84 m.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Differences between DEM elevation and GCPs larger than 100 m are returns from cloud tops.
    • Rests on an assumption: The GCP elevations are accurate enough to serve as ground truth for DEM vertical accuracy evaluation.

  • May combine two claims.

    Findingown data

    In cultivated areas of Cameroon, ASTER GDEM 2 has RMS vertical error 15.54 m and NMAD 11.30 m.

    “ASTER-GPS | Cultivated areas | 280 | -49.73 | 41.68 | -4.07 | -3.58 15.03 11.30 15.54”
    Quote in the paperSubstantial claim
    • Rests on an assumption: More than 50% of the study area in Cameroon is occupied by forests and woodlands and shrub and herbaceous lands, estimated at 69.15% and 11.23% respectively.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingnone given

    Cultivated areas and built up represent respectively 17.69% and 0.05% of the area under study.

    “The cultivated areas (agriculture and cropland) and built up represent respectively 17.69% and 0.05% area under study.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The 95% errors of the AW3D30-GPS dataset have a magnitude within [0, 17.33] m, SRTM 1-GPS within [0, 20.00] m, and ASTER-GPS within [0, 23.27] m.

    “Also, the 95% errors of AW3D30-GPS dataset have a magnitude in meters within the interval [0, 17.33], [0, 20.00] for SRTM 1-GPS, while it is [0, 23.27] for ASTER-GPS.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The GCP elevations are accurate enough to serve as ground truth for DEM vertical accuracy evaluation.

  • May combine two claims.

    Findingown data

    With increasing terrain slope, RMS and NMAD values of all tested global DEMs generally increase, except for the 0° to 2° slope class in ASTER GDEM 2.

    “With the increase of slope, RMS and NMAD values of all global DEMs generally increase except slope class 0°to 2°in ASTER GDEM 2.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    In cultivated areas of Cameroon, AW3D30 has RMS vertical error 10.91 m and NMAD 6.22 m.

    “AW3D30- | Cultivated areas | 280 | -39.26 | 42.87 | 1.16 | 1.66 10.87 | 6.22 10.91”
    Quote in the paperSubstantial claim
    • Rests on an assumption: More than 50% of the study area in Cameroon is occupied by forests and woodlands and shrub and herbaceous lands, estimated at 69.15% and 11.23% respectively.

  • May combine two claims.

    Findingown data

    In cultivated areas of Cameroon, SRTM 1 has RMS vertical error 11.25 m and NMAD 6.23 m.

    “SRTM 1-GPS Cultivated areas | 280 | -43.26 | 44.13 | 1.72 | 1.66 11.14 | 6.23 11.25”
    Quote in the paperSubstantial claim
    • Rests on an assumption: More than 50% of the study area in Cameroon is occupied by forests and woodlands and shrub and herbaceous lands, estimated at 69.15% and 11.23% respectively.

  • May combine two claims.

    Findingown data

    RMS errors for SRTM 1-GPS and AW3D30-GPS before removing outliers are 19.39 m and 19.58 m, and after removing outliers are 13.25 m and 13.07 m.

    “Relative to the RMS errors, the same correlations are observed for the both datasets (SRTM 1-GPS and AW3D30-GPS) before removing outliers (19.39 m and 19.58 m) and after removing outliers (13.25 m and 13.07 m).”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Differences between DEM elevation and GCPs larger than 100 m are returns from cloud tops.
    • Rests on an assumption: The GCP elevations are accurate enough to serve as ground truth for DEM vertical accuracy evaluation.

  • May combine two claims.

    Findingown data

    In forests and woodlands areas of Cameroon, ASTER GDEM 2 has RMS vertical error 20.03 m and NMAD 14.43 m.

    “In the forests and woodlands areas, the vertical DEM accuracy indicates RMS and NMAD of (15.07 m and 10.38 m), (14.87 m and 9.13 m) and (20.03 m and 14.43 m) respectively for AW3D30, SRTM 1 and ASTER GDEM 2.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: More than 50% of the study area in Cameroon is occupied by forests and woodlands and shrub and herbaceous lands, estimated at 69.15% and 11.23% respectively.

  • May combine two claims.

    Findingown data

    AW3D30 shows a positive median of 1.64 m and a negative bias after removing outliers of -0.10 m, while the STD (13.07 m) and 68.3% quantile (4.13 m) differ by 8.94 m.

    “Likewise, AW3D30 shows positive median (1.64 m) and negative bias after removing outliers (-0.10 m) while the STD (13.07 m) and 68.3% quantile (4.13 m) differ by 8.94 m.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Differences between DEM elevation and GCPs larger than 100 m are returns from cloud tops.
    • Rests on an assumption: The GCP elevations are accurate enough to serve as ground truth for DEM vertical accuracy evaluation.

  • May combine two claims.

    Findingown data

    For SRTM 1-GPS, the robust accuracy measures are MED 2.00 m and NMAD 7.41 m.

    “SRTM 1-GPS | 2.00 | 7.41 | 5.00 | 20.00”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The model-to-model comparison between AW3D30 and ASTER has mean difference 3.30 m and RMS 18.50 m.

    “AW3D30-ASTER | -276 | 187 | 3.30 | 18.20 | 18.50”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    For ASTER-GPS, the robust accuracy measures are MED -5.05 m and NMAD 13.30 m.

    “ASTER-GPS | -5.05 | 13.30 | 0.23 | 23.27”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The model-to-model comparison between SRTM 1 and ASTER has mean difference 3.10 m and RMS 18.20 m.

    “SRTM 1-ASTER | -277 | 191 | 3.10 | 17.90 | 18.20”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    For AW3D30-GPS, the robust accuracy measures are MED 1.64 m and NMAD 7.75 m.

    “AW3D30-GPS | 1.64 | 7.75 | 4.13 | 17.33”
    Quote in the paperSubstantial claim

  • May combine two claims. The quote only partly states this claim.

    Interpretationown data

    ASTER heights are below AW3D30 and SRTM heights over Cameroon.

    “The positive bias of 3.30 m and 3.10 m respectively for AW3D30-ASTER and SRTM-ASTER combinations show that ASTER heights are below to AW3D30 and SRTM respectively.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    Good quality GPS/leveling ground data are available in Cameroon and can be used for validation.

    “In addition, good quality ground data derived from GPS/leveling locally in the country, are available and can be introduced in such validation.”
    • Also stated in: § 1. Introduction · plain
      “GPS/Leveling data are also available.They were collected during the construction of the Cameroun Geodetic Network (Carme 2012).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Carme 2012

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

  • May combine two claims.

    Study design and conductown data

    The GCPs are not regularly distributed over Cameroon, with gaps especially in difficult-to-access areas.

    “The GCPs are not regularly distributed all over the country, and there remain many gaps especial in areas difficult to access (Figure 1).”
    Quote in the paperNot a substantial claim

  • The quote only partly states this claim.

    Interpretationown data

    Other accuracy measures (MED, 68.3% quantile, 95% quantile) supply evidence of the good quality of AW3D30 over Cameroon.

    “Other accuracy measures (MED, 68.3% quantile, 95% quantile) supply some evidence of the good quality of AW3D30 over Cameroon.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Study design and conductcitation

    The maximum height among the GCPs is 2101.11 m, which is much lower than Mount Cameroon's 4038 m, despite many GCPs near Mount Cameroon.

    “From this table, one can note that the maximal height is 2101.11m, too small compared to the highest summit of the area, Mount Cameroon at 4038 m (Kamguia, Tadjou, and Ngouanet 2015) though there are many GCPs in the vicinity of the Mount Cameroon.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Kamguia, Tadjou, and Ngouanet 2015

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    The distance between GCPs varies from less than 10 km in the west to more than 100 km in the eastern part of Cameroon.

    “The distance between them varies from less than 10 km in the west, to more than 100 km in the eastern part of the territory.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductcitation

    The 555 GPS/leveled GCPs provide horizontal accuracy at decimeter level and vertical accuracy at centimeter level, making them ideal reference data for assessing GDEMs.

    “These GCPs provide up to decimeter level and centimeter-level respectively for horizontal and vertical accuracies and are an ideal reference data to assess GDEMs.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Carme 2012

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

  • May combine two claims.

    Study design and conductcitation

    The EGM96 surface is a very close approximation to mean sea level, so matching vertical datum is not necessary in this study.

    “As the EGM96 surface is very close approximation to mean sea level (Pavlis, Rapp, and Olson 1998;Sun et al. 2003;Mukherjee et al. 2013;Jain et al. 2017), matching vertical datum is not necessary in this study.”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Pavlis, Rapp, and Olson 1998

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

    • cites Sun et al. 2003 source
    • cites Jain et al. 2017

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

  • Probably not a substantial claim: may not pass S1 (asserted), S3 (load-bearing). May combine two claims.

    Premisecitation

    Different mean values signs in various SRTM studies can be related to regional differences in terrain morphology or to differences in land cover (e.g. presence of forests).

    “Kolecka and Kozak (2013) believed that the different mean values signs in various SRTM studies can be related to regional differences in terrain morphology or to differences in land cover (e.g.presence of forests).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Kolecka and Kozak (2013)

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

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The underestimation of ASTER GDEM 2 for cultivated areas and open land has also been observed in prior studies.

    “This underestimation of ASTER GDEM 2 for cultivated areas and open land was also observed byJain et al. (2017)andMukherjee et al. (2013).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Jain et al. (2017)

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

    • cites Mukherjee et al. (2013) source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    AW3D30 is very closest to SRTM 1.

    “Hu et al. (2017) also found that AW3D30 is very closest to SRTM 1.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Hu et al. (2017)

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

  • May combine two claims.

    Interpretationown data

    SRTM 1 and AW3D30 are more resilient to the effects of scattering objects respectively in forests and cultivated areas.

    “However, SRTM 1 and AW3D30 are more resilient to the effects of the scattering objects respectively in forests and cultivated areas.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    The SRTM C-band has significantly better penetrability than SRTM X-band, but all rely on knowledge of vegetation type, density and structure to a large extent.

    “The SRTM Cband has significantly better penetrability than SRTM X-band, but all of them rely on knowledge of vegetation type, density and structure to a large extent (Carabajal and Harding 2005;Hofton et al. 2006;Farr et al. 2007;Li et al. 2016).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Carabajal and Harding 2005 source
    • cites Hofton et al. 2006 source
    • cites Farr et al. 2007 source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    The GlobCover 2009 map is based on ENVISAT-MERIS observations, with 300 m ground resolution and 22 land cover global classes.

    “The latter is based on ENVISAT-MERIS observations (Defourny et al. 2009), with 300 m ground resolution and with 22 land cover global classes (Arino et al. 2007;Li et al. 2016).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Defourny et al. 2009

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

    • cites Arino et al. 2007 source
    • cites Li et al. 2016

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

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation

    Removal of outliers before use of all DEMs in general and SRTM data in particular is absolutely essential to get the SRTM datasets close to the SRTM RMS goal of ~10 m as well as for statistically meaningful results.

    “Mukul, Srivastava, and Mukul (2016) stated that, removal of outliers before use of all DEMs in general and SRTM data in particular is absolutely essential to get the SRTM datasets close to the SRTM RMS goal of ∼10 m as well as for statistically meaningful results.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Mukul, Srivastava, and Mukul (2016) source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationnone given

    The limitations could also be related to additional effects of systematic errors, such as striping, that are a result of the pitch and yaw of the spacecraft during the data collection phase.

    “These limitations could also be related to additional effects of systematic errors, such as 'striping,' that are a result of the pitch and yaw of the spacecraft during the data collection phase.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.

    Premisecitation

    Slope is a terrain attribute that represents the magnitude of the terrain inclination and is a measure of the change of elevation over a horizontal distance.

    “Slope is a terrain attribute that represents the magnitude of the terrain inclination and is a measure of the change of elevation over a horizontal distance (Varga and Bašić 2015).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Varga and Bašić 2015 source

  • Classified as probably not a substantial claim. May combine two claims.

    Premisereasoning

    The best methods based on outlier removal do not achieve the performance of robust methods, because the latter are able to apply a smoother transition between accepting and rejecting an observation.

    “They mentioned that the best methods based on outlier removal do not achieve the performance of robust methods, because the latter are able to apply a more smooth transition between accepting and rejecting an observation.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Prior studies have also observed that RMS and NMAD values of global DEMs generally increase with terrain slope.

    “These observations were also done byJain et al. (2017)andVarga and Bašić (2015).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Jain et al. (2017)

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

    • cites Varga and Bašić (2015) source
49 of 49

  • The paper relies on this without support.

    No comprehensive study about the vertical accuracy of the newest high-resolution (30 m) global DEMs (AW3D30, SRTM 1, ASTER GDEM Version 2) has yet been reported over Cameroon.

    “This paper is justified by the fact that since the newest high-resolution (30 m) global DEMs (AW3D30, SRTM 1 and ASTER GDEM Version 2) are available, no comprehensive study about their vertical accuracy has yet been reported over Cameroon.”

  • The paper relies on this without support.

    The number of 555 GCPs is sufficient for statistical analyses to determine the vertical accuracy of each GDEM.

    “However, their number is enough for pertinent statistical analyses, for the determination of the vertical accuracy of each of the above GDEMs.”

  • The authors state this as an assumption.

    The vertical errors are normally distributed.

    “These errors assumed to be normally distributed are quantified in terms of Mean Error (ME) and the RMS widely adopted in DEMs vertical accuracy evaluation (Du et al. 2015;Varga and Bašić 2015;Mukul, Srivastava, and Mukul 2016).”

  • The authors state this as an assumption.

    Differences between DEM elevation and GCPs larger than 100 m are returns from cloud tops.

    “Also, the differences between DEM elevation and GCPs larger than 100 m were removed.These were assumed to be returns from cloud tops (Carabajal and Harding 2006;Du et al. 2015).”

  • The paper relies on this without support.

    The GCP elevations are accurate enough to serve as ground truth for DEM vertical accuracy evaluation.

    “the precise latitude/longitude true GCPs are used to estimate the geolocation errors.”

  • The paper relies on this without support.

    There is no reference DEM over Cameroon.

    “Due to the lack of the reference DEM over Cameroon, the precise latitude/longitude true GCPs are used to estimate the geolocation errors.”

  • The paper relies on this without support.

    More than 50% of the study area in Cameroon is occupied by forests and woodlands and shrub and herbaceous lands, estimated at 69.15% and 11.23% respectively.

    “More than 50% of the territory under study is occupied by forests and woodlands, shrub and herbaceous lands which have been estimated to be 69.15% and 11.23% respectively.”

  • The authors state this as an assumption.

    The elevation differences between global DEMs and GCPs are normally distributed.

    “with the assumption of normal distribution in function of outliers”
8 of 8

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.

  • AW3D30 does not have the lowest NMAD in every slope group

    Major

    The stated ranking fails for the slope group above 10°: Table 10 gives AW3D30 an NMAD of 15.00 m and SRTM an NMAD of 6.42 m.

  • Forest and shrub errors are not all overpredictions

    Major

    The claim of overprediction for all three DEMs conflicts with negative mean errors in these land-cover groups. For example, ASTER's forest mean is −4.23 m, which denotes underprediction under the paper's stated error convention.

  • SRTM built-up bias is negative, not positive

    Minor

    The prose attributes a positive bias to SRTM in built-up areas, but Table 9 reports a mean error of −0.46 m for that group.

  • Q-Q plots are identified as two different figures

    Minor

    The results text identifies Figure 5 as the Q-Q plots, while the Figure 5 caption identifies it as land-cover error histograms. The Q-Q plots are captioned as Figure 3.

  • ASTER georeferencing result differs between text and table

    Minor

    The text says ASTER's RMS after georeferencing was 24.2 m; Table 5 reports 24.47 m for the same result.

  • ASTER water-group maximum exceeds its reported total maximum

    Minor

    Table 9 reports an ASTER water-group maximum of 95.96 m, but its total row reports a maximum of only 62.70 m. A group's maximum cannot exceed the maximum of a total containing that group.

  • Two forest-coverage percentages are given

    Minor

    The prose gives forest and woodland coverage as 69.15%, whereas the land-cover table gives 68.15%.

7 of 7

Cites

Flagged first, then by citations

Research works printed but not linked: 8

  1. Amans O. C., 2013, International Journal of Scientific and Engineering Research, 4, 523
  2. Carme J. L., 2012, Revue, 37
  3. Croneborg L., 2015, Digital Elevation Models. A Guidance Note on how Digital Elevation Models are Created and Used – Includes key Definitions, Sample Terms of Reference and how Best to Plan a DEM-Mission
  4. Defourny P., 2009, Proceedings of the 33rd International Symposium on Remote Sensing of Environment (ISRSE)
  5. Forkuo E. K., 2010, The International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences, 37, 1023
  6. Kamguia J., 2007, Nordic Journal of Surveying and Real Estate Research, 4, 7
  7. Kamguia J., 2015, Ghana Journal of Science, 55, 37
  8. Shingare P. P., 2013, International Journal of Modern Engineering Research, 3, 2412

Cited by

Sorted by citations

Related

Authors
Yap, L., Kandé, L. H., Nouayou, R., Kamguia, J., Ngouh, N. A., & Makuate, M. B.
Affiliations
  • Université de Yaoundé I (CM)
  • Research Laboratory in Geodesy, National Institute of Cartography, Yaoundé, Cameroon
Corresponding author
Kamguia, Joseph
ORCID
Venue
International Journal of Digital Earth
Publisher
Informa UK Limited
Published
2018-04-04
Volume
12
Issue
5
Pages
500-524
ISSN
  • 1753-8947 (ISSN-L)
  • 1753-8955
DOI
10.1080/17538947.2018.1458163
OpenAlex
W2795932572
Cited by
70
References
70
Citations by year
2026: 8 · 2025: 6 · 2024: 7 · 2023: 9 · 2022: 7 · 2021: 12 · 2020: 13 · 2019: 8
FWCI
6.60
Type
journal article
Language
en
Open access
Licence
other-oa
Locations
  • International Journal of Digital Earth (publishedVersion, other-oa)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
Topics
  • Physical Sciences › Engineering › Ocean Engineering › Satellite Image Processing and Photogrammetry
  • Physical Sciences › Engineering › Aerospace Engineering › GNSS positioning and interference
  • Physical Sciences › Engineering › Civil and Structural Engineering › Geodetic Measurements and Engineering Structures
Keywords
global digital elevation models, vertical accuracy, georeferencing errors, terrain slope effects, terrain roughness, outlier detection, normality testing

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.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-11.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-11.

Cited sources are checked by 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