Content uploaded by Arif Oguz Altunel
Author content
All content in this area was uploaded by Arif Oguz Altunel on Oct 11, 2023
Content may be subject to copyright.
5th Intercontinental Geoinformation Days (IGD) – 14-15 December 2022 – NIGMT Foundation, New Delhi, India
* Corresponding Author
Cite this study
*(aoaltunel@kastamonu.edu.tr) ORCID ID 000 0-0003-2597-5587
(oesakici@kastamonu.edu.tr) ORCID ID 0000-0003-4961-2991
Altunel, A. O., & Sakici, O. E. (2022). The ultimate vertical accuracy assessment of the third
generation Turkish 1:25000 quad maps; under canopy vs. no canopy. 5th Intercontinental
Geoinformation Days, 136-140, Netra, India
5th Intercontinental Geoinformation Days
igd.mersin.edu.tr
The ultimate vertical accuracy assessment of the third generation Turkish 1:25000 quad
maps; under canopy vs. no canopy
Arif Oguz Altunel*1 , Oytun Emre Sakici 1
1Kastamonu University, Forest Engineering Department, Faculty of Forestry, Kastamonu, Türkiye
Keywords
Abstract
Photogrammetry
Quad Maps
DEM
Raster Resolution
Elevation, vertical accuracy of any topographic Earth representation, e. g. stereo surface
models, topo maps, DEMs, etc., is important if such data will be the base of further projects or
development plans. The main form of these types of data in Türkiye is “1:25000” scaled quad
maps. The third generation such maps were produced via digital stereo air-photo capture and
photogrammetry capabilities as opposed to the previous two analogue based releases.
Through this long-adapted scale, land cover types, hydrological formations, surface features,
down to house rooftops, can be depicted in these maps. Elevation integration are also provided
through the contour lines drawn in 10 m elevation difference showing intervals. They are the
most frequently addressed topographic data type in forestry education as well as in
profession. With the establishment of county-wide active GNSS network, very high precision
elevation verification has become available for multitude of purposes. In this study, four dam
reservoirs intensively surveyed using CORS-GPS were used to assess the vertical accuracies of
the corresponding quad-map based DEMs produced in different resolutions. RMSEs ranged
from 5.49 m to 14.22 m when the entire quad sheets were used while they ranged from 2.58
m to 8.95 m when the quads were purposely cut. Canopy closure apparently worsened the
results.
1. Introduction
Although there were earlier attempts in site-specific
(Sahin et al. 2022) or country-wide (Dagdas and Bilge
2015) scales, depicting target land cover type, forests,
topography related map production, undertaken by
Turkish Mapping Command in Türkiye started revealing
the first country-wide coverage around 1959-1960.
Relatively similar to 7.5 minute, 1:24000 US
quadrangle maps, 1:25000 Turkish quads can be
considered as the country’s main elevation integrated,
forest cover and type prioritizing topographic maps.
Repeated with a second coverage around 1992-1993,
they were produced via stereo air-photo capture and
photogrammetric analyses and interpretation
capabilities. These two coverages were produced with
analogue means.
Since elevation has been embedded into such maps in
contour line fashion, they have been accepted as the first
set of reliable datasets to produce surface models, DEMs,
in cartographic studies using GIS software(s), starting
from the 1980s onward (Taud et al. 1999; Ardiansyah
and Yokoyama 2002; Guth 1999).
Although everybody has started using them in all
sorts of projects, reports, theses, etc. to generate surface
models and their inherent derivatives e.g., slope, aspect,
hillshade, roughness, hydrology, etc. not very many
studies looking into the actual vertical accuracy of those
maps surfaced until the late 2000s. Based on a
verification established over 1:16000 stereo air photo
driven models, Ozturk and Kocak (2007) found out that
1:25000 Turkish quad maps had RMSEs in the range of
±2 m. Then, Bildirici et al. (2009) used them to assess the
practicality of the newly released 3 arc-second SRTM and
showed that SRTM’s absolute height error was actually
better than the mission stated 16 m.
Although there have been even more subsequent
coverages in places where frequent land-cover/land-use
changes occur, finally, the third and current country-
wide coverage produced with digital means, was started
to be released around 2009-2010. Around the same in
2009, Türkiye also established and started effectively
using the indigenous active GNSS system, TUSAGA-Active
(TA) network (Yildirim et al. 2011). Using this new
achievement, it’s been possible to acquire positional
coordinates, x, y, z, in millimeter accuracy in much of the
country as well as in Northern Cyprus.
5th Intercontinental Geoinformation Days (IGD) – 14-15 December 2022 – New Delhi, India
137
In the scope of this study, four dam reservoirs, two to
be transformed from agricultural fields, one to be
transformed from >%90 forest cover and one to be
transformed from <%30 forest cover were surveyed
utilizing CORS GPS constantly communicating with TA
network, and the results were compared to those
extracted from quad map(s) generated different
resolution DEMs; purpose-cut, entire sheet, 10 m
resampled, 30 m resampled.
2. Study Area and Methodology
2.1. Study Area
Four dam reservoirs chosen previously by the State
Hydraulic Works (DSI) for hydro-electric and irrigation
purposes in Kastamonu province, were meticulously
surveyed in 2014-2015 period by independent surveyors
for engineering and hydrologic calculations, yet to come
in the following years (Figure 1). Two of the dam
reservoirs (Incebogaz and Hasanli) were planned over
agricultural areas with occasional single-story dwellings
for living and livestock storage. One of them (Arac) had
<%30 forest cover with human habitation signs and the
last one (Obrucak) had >90% forest cover with no
habitations.
Figure 1. Locations of the studied dam reservoirs
2.2. Methodology
GCPs were recorded using ITRF-96 coordinate
system meaning 30 Transverse Mercator projection. An
illustration of GCP set acquired in Obrucak Reservoir can
be seen in Figure 2. Elevation measurements were based
upon GRS80 vertical datum. Elevation readings were
subsequently transformed into orthometric heights by
subtracting the respective geoid heights from the
recorded ellipsoidal heights (Simav et al. 2015). After the
elevation correction, all GCP records were transformed
into 60 Universal Transvers Mercator projection over
WGS84 datum for easier comparison with the elevations
to be extracted from different versions of the quad map
generated DEMs. Tested quad coverage was also
produced adopting the same projection and datum.
ArcGIS 10.8 was used in the analyses.
Figure 2. Random GCPs within Obrucak dam reservoir
Vectorized quad sheets were used. Sheets were first
purposely cut using the polygons housing the GCPs,
Second, TIN surfaces were generated for each site within
the designated polygons. Then, a TIN to raster
conversion was performed for each site preserving the
ArcGIS recommended default cell sizes. Thus, four site-
specific DEM datasets, “Purpose-Cut Quad Sheet” were
generated.
The same sequence was repeated to create four more
DEM datasets utilizing the quad sheets as whole, “Entire
Quad Sheet”. 10 m and 30 m cell size preferences were
dictated during entire sheet TIN to raster conversion
phases of the third and fourth DEM dataset creations, “10
and 30 m Resampled Entire Quad Sheet” (Table 1).
Table 1. Summary statistics
Raster Resolution
Arac Dam
Reservoir
Obrucak Dam
Reservoir
Hasanli Dam
Reservoir
Incebogaz Dam
Reservoir
From Purpose-cut Quad Sheet (m)
11.2
7.5
12.3
7.5
From Entire Quad Sheet (m)
84.4
112.4
56.3
111.8
From 10 m Resampled Entire Quad Sheet (m)
10
10
10
10
From 30 m Resampled Entire Quad Sheet (m)
30
30
30
30
Number of GCPs
41181
26716
14894
11226
Acreage (ha)
339.7
160.5
394.9
64.1
Number of quad sheets per site (tying)
2
4
1
2
Quad sheed line-up sequence
West-East
All around
None
North-South
A total of 16 DEM datasets was generated to test how
quad sheet generated DEM resolution would differ in
elevation accuracy against precisely measured GPS GCPs.
To do this, each random GCP dataset was placed on the
generated DEM(s) and the respective elevation record of
each GCP was extracted from four different DEM
datasets. Root mean square error (RMSE), mean error
(ME), mean absolute error (MAE) and standard deviation
5th Intercontinental Geoinformation Days (IGD) – 14-15 December 2022 – New Delhi, India
138
(STD) were calculated (Satge et al. 2016). They were then
placed as input into Poudel and Cao (2013) approach to
get a collective comparison result. The respective
equations are as followed;
(1)
(2)
(3)
(4)
where n is the number of GCPs, x is the measured
elevation value (m) of the GCPs, while y is the elevation
value extracted from DEM datasets.
(5)
where Ri is the relative rank of the DEM datasets
(i=purpose-cut, entire sheet, 10 m resampled and 30 m
resampled), Si is the basis of error values produced by
each DEM dataset, Smin is the minimum value of Si and Smax
is the maximum value Si, m is the number of questioned
DEM datasets. The equation produced a ranking score
ranging from 1 to m. The remaining ranks were produced
in real numbers between 1 and m.
3. Results and Discussion
As apparent from many studies based on both active
and passive-sensor produced surface models, DEMs, the
vertical accuracy performance of the end product is
highly correlated with the land cover type and
topographic uncertainties within the target during the
actual image acquisition (Shortridge 2006; Wechsler and
Kroll 2006; Hebeler and Purves 2009; Altunel 2018;
Gonzalez and Rizzoli 2018;). Although smaller in caliber
in terms of investment, coverage and know-how, aerial
stereo image capture, today, is not entirely different from
those of the satellite-based ones. Tested third generation
quad coverage was produced from stereo captured color
infrared air-photos, better defined as air-imagery.
Imagery-wise, Yilmaz and Erdogan (2018) showed that
RMSE of DEMs produced from new stereo air-photos
captured at 45 cm ground sampling distance were ±2.51
m, ±1.38 m and ±1.3 m within Uşak, Aksaray and Dogu
Beyazit designated quad sheets. They said a 5 m GRID
spaced DEM could very well be produced for the entire
country, utilizing the new generation air-photos.
While the building blocks of the third and later
version(s) country-wide quad sheets have been this
strong, it is perfectly logical to think that elevation
accuracy of the quad sheets must also be close to above
mentioned figures.
Arac, Obrucak, Hasanli and Incegogaz reservoir areas
extended across 435-1710 m, 580-1750 m, 555-1280 m
and 680-1375 m elevations, respectively. In Arac, Hasanli
and Incebogaz reservoirs, bulk of the slope facades was
on sloping to very steep slopes, 5%< - <150%, whereas
in Obrucak they were on moderately steep to very steep
slopes, 15%< - <200% (FAO, 2006). These figures
amounted to 93% of Arac reservoir, 99.6% of Hasanli
reservoir, 95% of Incebogaz reservoir and 91% of
Obrucak reservoir land area being on steep topography.
This could be understood when water storage was
intended. Random, but rather tightly, recorded GCPs
allowed us to reach the results presented in this study.
RMSE-wise, the results were as followed: in Arac
reservoir, purpose-cut and 10 m resampled DEMs were
clearly similar and better than 30 m resampled DEM,
entire sheet-based DEM produced the least favorable
elevation performance; in Hasanli reservoir, the situation
was the same, but the gain was marginal; Incebogaz
reservoir, same results with better gain were observable
and in Obrucak reservoir, same results again with more
than 2 times better gain was obvious (Table 2).
Table 2. Ranking results based on individually calculated errors
Reservoir
Canopy
Closure
Raster Making Method
RMSE
ME
MAE
STD
Total
Rank
Overall
Rank
Arac
Partial
canopy
(<%30)
Purpose-cut Quad Sheet
4.73 (1.00)
-0.77 (1.26)
3.42 (1.00)
4.67 (1.00)
4.26
1.07
Entire Quadrangle Sheet
7.02 (4.00)
-1.08 (4.00)
5.13 (4.00)
6.94 (4.00)
16.00
4.00
10 m Resampled Entire Quad Sheet
4.73 (1.00)
-0.74 (1.00)
3.42 (1.00)
4.67 (1.00)
4.00
1.00
30 m Resampled Entire Quad Sheet
5.08 (1.46)
-0.77 (1.26)
3.70 (1.49)
5.02 (1.46)
5.68
1.42
Obrucak
Full canopy
(>%90)
Purpose-cut Quad Sheet
6.77 (1.03)
0.07 (1.28)
4.95 (1.00)
6.67 (1.00)
4.31
1.00
Entire Quadrangle Sheet
14.22 (4.00)
-0.56 (4.00)
10.6 (4.00)
14.21 (4.00)
16.00
4.00
10 m Resampled Entire Quad Sheet
6.70 (1.00)
0.07 (1.28)
4.98 (1.02)
6.70 (1.01)
4.31
1.00
30 m Resampled Entire Quad Sheet
7.54 (1.34)
0.02 (1.00)
5.54 (1.31)
7.54 (1.35)
4.99
1.18
Hasanli
Agriculture
(no canopy)
Purpose-cut Quad Sheet
8.95 (1.32)
-0.14 (1.00)
4.25 (1.29)
8.91 (1.29)
4.78
1.00
Entire Quadrangle Sheet
9.12 (4.00)
-0.19 (4.00)
4.53 (4.00)
9.07 (4.00)
16.00
4.00
10 m Resampled Entire Quad Sheet
8.93 (1.00)
-0.18 (3.40)
4.22 (1.00)
8.90 (1.00)
6.40
1.43
30 m Resampled Entire Quad Sheet
8.97 (1.63)
-0.18 (3.40)
4.30 (1.77)
8.92 (1.77)
8.34
1.95
Incebogaz
Agriculture
(no canopy)
Purpose-cut Quad Sheet
2.58 (1.00)
0.81(4.00)
2.05 (1.00)
2.45 (1.00)
7.00
1.22
Entire Quadrangle Sheet
5.49 (4.00)
0.08 (1.00)
4.04 (4.00)
5.49 (4.00)
13.00
4.00
10 m Resampled Entire Quad Sheet
2.59 (1.01)
0.68 (3.47)
2.05 (1.00)
2.49 (1.04)
6.52
1.00
30 m Resampled Entire Quad Sheet
2.94 (1.37)
0.61 (3.18)
2.32 (1.41)
2.88 (1.42)
7.38
1.40
Although the tendency in terms of elevation
performance was towards purpose-cut and 10 m
resampled DEMs in all reservoirs, the results were not
the same one another despite the fact that they were all
5th Intercontinental Geoinformation Days (IGD) – 14-15 December 2022 – New Delhi, India
139
located within same geographical region. Location-wise,
Incebogaz reservoir produced nearly the same results of
Ozturk and Kocak (2007) and Yilmaz and Erdogan
(2018), however the rest was worse. Even though their
results were basing upon direct air photography
photogrammetric calculations, it was nice to see that a
secondary product fabricated using the same
photography would match their original precision.
Closed canopy in Obrucak reservoir must have been the
reason that entire sheet-based DEM produced the overall
worse RMSE, 14.2 m. Besides, this high RMSE was also
triggered by the DEM acquired, combining four quad
sheets together. Partial canopy closure in Arac reservoir
did not tarnish the RMSE as much as that of Obrucak
reservoir. Additionally, MAE values calculated over all
questioned DEMs were the overall highest just like those
of the RMSEs in the same reservoir.
In three out of four reservoirs, Arac, Obrucak,
Incebogaz, purpose cutting the quad map clearly
improved the DEM making performance of the quad
maps compared to that of the entire sheet-based DEM.
The gains were close to more than two times. However,
the fact that no such improvement was observed in
Hasanli reservoir convinced us that it would be
impossible to get the same elevation precision from all
quad maps. Nevertheless, it is possible to say that a less
than 10 m quad specified contour interval precision can
be achieved in the third generation 1:25000 Turkish
quad maps.
Resampling clearly improved the DEM making
performance of the quad maps. Both of the tested GRID
spacing, 10 m and 30 m, were better in Arac, Obrucak and
Incebogaz reservoirs, 10 m DEM being the better one in
each, than the entire sheet-based DEM, but no noticeable
difference was observed in Hasanli reservoir. Sorensen
and Seibert (2007) showed that high resolution DEM
provided better TWI distribution while Tan et al. (2015)
said the most sensitive SWAT model DEM parameter was
DEM resolution so higher the resolution better the
outcome. Quad line-up sequence did not have any
detectable effect over the calculated error values,
however, although not certainly conclusive, it was
obvious that quad tying worsened the error values.
The remaining error calculations, ME, MAE and STD
behaved the same so detailed explanations were deemed
unnecessary to elaborate, however they were nice
additions to achieve the overall ranking results for each
reservoir area.
4. Conclusion
Quad, topographic, maps are important geographical
assets of a country. Türkiye has long had a meticulous
tradition of producing them systematically. Three
nation-wide coverages have been released as of 2022,
and they have been produced through photogrammetric
calculations performed over tens of thousands of stereo-
captured air photos. Topography is depicted via
elevation embedded contour lines, which yielded the
above results for the third and current coverage in four
dam reservoirs in Kastamonu province. This study
showed that a less than 10 m vertical accuracy can be
attained directly from 1:25000 Turkish national quads,
and the results can be further improved if secondary
products such as resampled DEMs, are produced.
References
Altunel, A. O. (2018). Suitability of open-access elevation
models for micro-scale watershed planning.
Environmental Monitoring and Assessment, 190(512).
Ardiansyah, P. O. D., & Yokoyama, R. (2002). DEM
generation method from contour lines based on the
steepest slope segment chain and monotone
interpolation function. ISPRS Journal of
Photogrammetry and Remote Sensing, 57(1-2), 86-
101.
Bildirici, O. I., Ustun, A., Selvi, Z. H., Abbak, A. R., &
Bugdayci, I. (2009). Assessment of shuttle radar
topography mission elevation data based on
topographic maps in Turkey. Cartography and
Geographic Information Science, 36(1), 95-104.
Dagdas, S., & Bilge, S. (2015). Türkiye Cumhuriyetinin
orman alanlarını gösteren ilk haritasi ve orman
varlığımız üzerine-(1926). Orman Mühendisliği
Dergisi, 52, 28-36. (in Turkish)
FAO. (2006). Guidelines for soil description (Fourth
edition). Food and Agriculture Organization of the
United Nations, Rome.
Gonzalez, C., & Rizzoli, P. (2018). Landcover-dependent
assessment of the relative height accuracy in
TanDEM-X DEM products. IEEE Geoscience and
Remote Sensing Letters, 15(12), 1892-1896.
Guth, P. L. (1999). Contour line “Ghosts” in USGS Level 2
DEMs. Photometric Engineering and Remote Sensing,
65(3), 289-296.
Hebeler, F., & Purves, R. S. (2009). The influence of
elevation uncertainty on derivation of topographic
indices. Geomorphology, 111, 4-16.
Ozturk, E., & Kocak, E. (2007). Farklı kaynalardan değişik
yöntem ve ölçeklerde üretilen sayısal yükseklik
modellerinin doğruluk araştırması. Harita Dergisi,
73(137), 25-41. (in Turkish)
Poudel, K. P., & Cao Q. V. (2013). Evaluation of methods
to predict Weibull parameters for characterizing
diameter distribution. Forest Science, 59(2), 243-252.
Sahin, A., Caglayan, I., Buyuk, H., Karademir, H., Aksu, A.,
& Sahin, H. (2022). Türkiye’nin ilk orman planlama
ünitesindeki teknik ve yapısal değişimlerin 100 yıllık
değerlendirilmesi. Turkish Journal of Forestry
Research, 9(1), 12-34. (in Turkish)
Satge, F., Denezine, M., Pillco, R., Timouk, F., Pinel, S.,
Molina, J., Garnier J., Seyer, F., & Bonnet, M. (2016).
Absolute and relative height-pixel accuracy of SRTM-
GL1 over South American Andean Plateau. ISPRS
Journal of Photogrammetry and Remote Sensing, 121,
157-166.
Shortridge, A. (2006). Shuttle radar topography mission
elevation data and its relationship to land cover.
Cartography and Geographic Information Science,
33(1), 65-75.
Simav, M., Yıldız, H., Cingöz, A., Sezen, E., Demirsoy, N. S.,
Akpınar, İ., …, & Doğan, U. (2015, March 25-28).
Türkiye Yükseklik Sisteminin Modernizasyonu ve
Gravite Altyapısının Iyileştirme Projesi. 15th Türkiye
Harita Bilimsel ve Teknik Kurultayı, Ankara, Türkiye.
5th Intercontinental Geoinformation Days (IGD) – 14-15 December 2022 – New Delhi, India
140
Sorensen, R., & Seibert, J. (2007). Effect of DEM resolution
on the calculation of topographical indices: TWI and
its components. Journal of Hydrology, 347(1-2), 79-89.
Tan, M. L., Ficklin, D. L., Dixob, B., Yusop, Z., & Chaplot, V.
(2015). Impacts of DEM resolution, source and
resampling technique on SWAT-simulated
streamflow. Applied Geography, 63, 357-368.
Taud, H., Parrot, J., & Alvarez, R. (1999). DEM generation
by contour line dilation. Computers and Geosciences,
25(7), 775-783.
Wechsler, S. P., & Kroll, C. N. (2006). Quantifying DEM
uncertainty and its effect on topographic parameters.
Photogrammetric Engineering and Remote Sensing,
72(9), 1081-1090.
Yildirim, O., Mekik, C., & Bakici, S. (2011). TUSAGA-Aktif
CORS-TR sisteminin Tapu ve Kadastro Genel
Müdürlüğüne katkıları. Jeodezi ve Jeoinformasyon
Dergisi, 104(2), 134-139.
Yilmaz, A., & Erdoğan, M. (2018). Designing high
resolution countrywide DEM for
Turkey. International Journal of Engineering and
Geosciences, 3(3), 98-107.