LENA DELTA WATER BODY MAPPING BASED ON RAPIDEYE SATELLITE
IMAGE SERIES FOR DETERMINATION OF LAKE LEVEL HEIGHTS AND
DELTA CHANNEL INCLINATION USING ALTIMETRY AND DEM DATA
Master thesis
M.Sc. Program for Polar and Marine Sciences POMOR
Saint Petersburg State University / Hamburg University
by
Aleksandr Volynetc
Saint Petersburg / Hamburg, 2017
Supervisors
Dr. F. Günther, Alfred Wegener Institute Helmholtz Centre for Polar and Marine
Research, Potsdam, Germany
Dr. Fedorova I.V., Saint Petersburg State University, Institute of Earth Sciences, Saint
Petersburg, Russia
Content
Abstract (in English) ......................................................................................................... 3
Abstract (in Russian) ......................................................................................................... 5
1. Introduction ................................................................................................................... 7
1.1. Objectives ............................................................................................................... 9
1.2. Study area ............................................................................................................. 10
1.3. Geomorphological and Geological settings.......................................................... 12
2. Materials and Methods ................................................................................................ 15
2.1. Data ...................................................................................................................... 16
2.1.1. Satellite images .............................................................................................. 16
2.1.2. Digital Elevation Model................................................................................. 18
2.1.3. Additional GIS layers .................................................................................... 20
2.2. Methods ................................................................................................................ 21
2.2.1. Satellite images sets processing ..................................................................... 21
2.2.2. Creation of water mask .................................................................................. 26
2.2.3. DEM refinement ............................................................................................ 29
2.2.4. Computation of characteristics ...................................................................... 30
3 Results .......................................................................................................................... 32
3.1. Spatial statistics .................................................................................................... 33
3.2. Lake height distribution........................................................................................ 38
3.3. Vertical section of the Lena Delta based on the lakes level heights .................... 41
4. Discussion ................................................................................................................... 45
4.1. Applicability of remote sensing data processing and GIS methods for creation of
water objects scheme ................................................................................................... 45
4.2. Discussion of obtained results .............................................................................. 51
5. Conclusions ................................................................................................................. 56
6. Acknowledgements ..................................................................................................... 58
7. References ................................................................................................................... 59
2
LENA DELTA WATER BODY MAPPING BASED ON RAPIDEYE SATELLITE
IMAGE SERIES FOR DETERMINATION OF LAKE LEVEL HEIGHTS AND
DELTA CHANNEL INCLINATION USING ALTIMETRY AND DEM DATA
Aleksandr Volynetc
Master Program for Polar and Marine Sciences POMOR / 022000 Ecology and
environmental management
Supervisors:
Dr. F. Günther, Alfred Wegener Institute Helmholtz Centre for Polar and Marine Research, Potsdam,
Germany
Dr. Fedorova I.V., Saint Petersburg State University, Institute of Earth Sciences, Saint Petersburg, Russia.
Past and present climate changes lead to a significant impact on dynamics of
periglacial landscapes in Eastern Siberia. Thermokarst and polygonal lakes, ponds,
watercourses and swamps are the inherent part of landscapes and primarily are subject
to thermokarst and thermal erosion processes. The main objectives of master thesis are
mapping of lakes and watercourses in the Lena Delta and adjacent Bykovsky Peninsula
with subsequent determination of the height of the water's edge of these objects.
This region of interest is a notable as it is the largest arctic delta with its
complicated geomorphological structure represented with three different in origin,
structure and age main terraces. Such structure distinguishes the Lena Delta from other
subarctic rivers deltas.
The study based on complex combination of Remote Sensing methods, GIS
handling and statistical calculations. As initial data for the mapping of water bodies
were used sets of high-resolution images acquired by two satellite surveying systems
RapidEye and SENTINEL-2. To determine the height values of lakes was used
TanDEM-X (TDX) digital elevation model. Received results were checked and
compared with collected by GLAS laser altimetry data that was aboard ICESat satellite
observing mission.
Statistical and visual analysis based on obtained map of lakes, watercourses of
Lena delta with area more than 100 m² and Bykovsky Peninsula and the height of the
water's edge of these objects allows differentiating terraces by limnicity and reflects
differences in plane and altitude characteristics of lakes in different terraces. On the
other hand, some generic features of the region were highlighted. Distribution of lakes
and its areas by terraces demonstrated similarity with the results obtained in previous
3
studies. However using of high-resolution satellite imagery in this work allowed to take
into account the influence of small lakes and consequently increased the detail of the
map. Comparing the lake heights based on digital elevation model with averaged data of
laser altimetry showed a correlation close to unity.
4
ОПРЕДЕЛЕНИЕ УРОВНЯ ОЗЕР И УКЛОНА ВОДОТОКОВ В ДЕЛЬТЕ РЕКИ
ЛЕНЫ, ОСНОВАННОЕ НА МОЗАИКЕ КОСМИЧЕСКИХ СНИМКОВ, ДАННЫХ
ЛАЗЕРНОЙ АЛЬТИМЕТРИИ И ЦМР
Волынец Александр
Магистерская программа «Полярные и морские исследования» («ПОМОР») /
022000 «Экология и природопользование»
Выпускная квалификационная работа магистра
Научные руководители:
Доктор Гюнтер Ф., Институт им. Альфреда Вегенера центр полярных и морских исследований им.
Гельмгольца, Потсдам, Германия
Доцент, к.г.н. Федорова И.В., Санкт-Петербургский государственный университет, Институт Наук
о Земле, Санкт-Петербург, Россия
Климатические изменения, как в прошлом, так и в настоящем, оказывают
заметное влияние на динамику перигляциальных ландшафтов в Восточной
Сибири. Термокарстовые и полигональные озера, водотоки и болота являются
неотъемлемой частью таких ландшафтов и в первую очередь подвержены
процессам термокарста и термоэрозии. Основными целями магистерской работы
являются картографирование озер и водотоков в дельте Лены и на прилегающем к
ней полуострове Быковский и последующее определение высоты уреза воды
данных объектов.
Данная территория является примечательной в связи с тем, что
представляет собой крупнейшую в арктическом регионе дельту, которая обладает
сложной геоморфологической структурой, выраженной в трех различных по
строению и возрасту главных террасах, что отличает ее от дельт других рек в
субарктической зоне.
Проведенное исследование основано на комплексном сочетании методов
дистанционного зондирования Земли (ДЗЗ), инструментария географических
информационных систем (ГИС) и математической статистики. В качестве
исходных данных для картографирования водных объектов использовались
наборы снимков высокого разрешения спутниковых съемочных систем RapidEye,
а также SENTINEL-2. Для определения высот озер использовалась цифровая
модель рельефа (ЦМР) TanDEM-X (TDX). Для оценки точности полученных
5
высот в качестве эталона использовались данные лазерной альтиметрии GLAS,
полученные со спутниковой системы ICESat.
Статистический и визуальный анализ, проведенный на основе полученной
карты озер, водотоков дельты Лены площадью более 100 м² и полуострова
Быковский, а также высот урезов воды озер, позволяет дифференцировать
террасы по озерности и отражает различия плановых и высотных характеристик
озер, расположенных на разных террасах. С другой стороны, были выделены
некоторые общие черты распределения, соответствующие данному региону.
Распределение озер и их площадей по террасам показало сходство с результатами,
полученными в предыдущих исследованиях. Однако использование в данной
работе спутниковых снимков высокого разрешения позволило учесть влияние
небольших по площади озер и, как следствие, увеличило подробность карты.
Сравнение высот озер, полученных на основе ЦМР с усредненными данными
лазерной альтиметрии, показало корреляцию, близкую к единице.
6
1. Introduction
Current climate change is affecting arctic regions at a faster rate than the rest of
the world. Over the last century the average surface temperature in the Arctic has
increased by about 0.09 ⁰C per decade, a rate 50 % greater than that observed over the
Northern Hemisphere as a whole (AMAP, 2011). A notably significant increase of air
temperatures has been registered during the last decades and is projected to further rise.
Such considerable changes potentially may have an impact on thaw-vulnerable
permafrost landscapes, which occupy about 24% of the northern hemisphere's land mass
(Zhang et al., 2008). Frozen ground contains a great amount of organic carbon, which
can be released due to thawing of permafrost and accelerate climate warming (Grosse et
al., 2011). For example, recent quantifications for the Holocene Lena Delta River
terrace by Zubrzycki et al. (2013), which comprises the area of modern deltaic
processes, mean soil organic carbon stocks for the upper 1 m of soils were estimated at
29 ± 10 kg m−2.
Since natural systems and low-lying permafrost-dominated arctic river deltas in
particular are sensitive to climate variability, climate change in polar region resulted in
growing interest of geoscientists in Land-Ocean interactions including estuary dynamics
and in close monitoring of the Arctic river deltas dynamics to better estimate landscape
scale climate change impacts and to quantify carbon fluxes (Rachold et al., 2000; Are
and Reimnitz, 2000; Grigoriev et al., 2004).
For several months in years 2016/2017 I conducted an internship in the Alfred
Wegener Institute in the Periglacial Research Section in Potsdam (Germany). This
section aims at the observation and quantification of current periglacial processes and
environmental changes and their causes in order to assess the state of permafrost today
and its future transformation (AWI, 2017). In order to better understand permafrost
dynamics in far polar regions, remote sensing of landscape-scale changes is becoming
an increasingly important method (Jorgenson and Grosse, 2016).
Permafrost, defined as the ground that remains at or below 0 ⁰C for more than
two years and can be differentiated by its spatial extent into continuous (90–100%),
discontinuous (50–90%), sporadic (10–50%), and isolated (0–10%) permafrost, as well
as by its thickness, the amount of ground ice present, and its temperature (Grosse G. et
al., 2013). Northeastern Siberia belongs to the zone of continuous permafrost. Arctic
permafrost landscapes are among the most vulnerable and dynamic landscapes globally
(Khury et al., 2013). One of the main destructive processes affecting permafrost is
7
thermokarst (Kokelj and Jorgenson, 2013). Thermokarst refers to the process by which
characteristic landforms form following disturbance of the thermal equilibrium of the
ground resulting in thaw of ice-rich permafrost or melting of massive ice (van
Everdingen, 2005). Thermokarst is one of the most obvious types of permafrost
degradation in arctic landscapes (Morgenstern et al., 2011). This process is usually
expressed in formation, growth and vanishing of thermokarst lakes (Grosse et al., 2013)
which are defined as lakes that usually occupy closed depressions formed by the
settlement of frozen ground following thawing of ice-rich permafrost or melting of
massive ice (van Everdingen, 2005). Thermokarst lakes and basins are ubiquitous
landforms in ice-rich permafrost deposits in Siberia and typical features of the northern
permafrost ecosystems. Thermokarst in East Siberian ice-rich permafrost massively
developed at the transition from Pleistocene to Holocene, but after the Boreal period (9–
7.5 ka BP), the thermokarst landscapes appeared as they do today and have experienced
only minor changes since then (Morgenstern et al., 2013). Modern estimates of
thermokarst lakes areal coverage span a wide range and are scale dependent, but
obviously they can occupy a significant proportion of the land area in high latitude
regions (up to 40% in some areas (Antonova et al, 2016). Therefore, it is obvious that
arctic water bodies play a crucial role in land-atmosphere exchanges of greenhouses
gases and energy fluxes. That is why the study of these objects is highly important for
assessing the impacts global climate change on local (Boike et al., 2008; Muster et al.,
2013).
Due to interaction between hydrosphere, lithosphere, atmosphere, biosphere and
cryosphere arctic deltas are among the most dynamic and complex natural phenomena
on earth (Walker, 1998). The Lena River Delta, situated in Northern Siberia, is one of
key areas in these investigations because it is the largest delta in the Arctic. The
importance of this place is also confirmed by long-term monitoring efforts at the
research station on Samoylovskiy Island (Boike et al., 2013). Since natural deltas are
characterised by complex geomorphological patterns, hydrological conditions and
various types of ecosystems, precise and modern information on the distribution and
extent of the delta water objects is necessary for a spatiotemporal assessment and
accurate quantification of processes which drive the different types of changes in this
region (Schneider et al., 2009)
Due to permafrost landscapes extent and remoteness, most of their changes
remain unnoticed (Nitze and Grosse, 2016). Remote sensing techniques have the
potential to afford precise and cost-effective means for observation, mapping and
8
analysis of Earth surface. These methods in geographical research provide a lot of
opportunities and advantages especially for investigation of remote regions. Therefore
these techniques are broadly applied for a large variety of 2D and 3D geomorphodynamic monitoring applications in the Laptev Sea region and the Lena Delta (Günther
et al., 2013; Günther et al., 2015; Nitze and Grosse, 2016).
Lake level changes can be considered as one important indicator for the water
balance in subarctic regions and for frozen ground mutability. Since the 1990s, satellite
radar altimetry has effectively been used for monitoring the water surface elevation
changes (Vu Hien Phan, Roderik Lindenbergh, 2011). However, although thermokarst
lake change detection over time is a common remote sensing application (Kravstsova
and Bystrova, 2009; Nitze et al., 2017), comprehensive inventories of thermkarst lakes
with respect to lake water levels have been rarely carried out (Ulrich et al., 2017). It is
worth saying that modern geographical studies are often based on GIS methods, which
open a variety of opportunities for the processing initial data, analysis of spatial
information and maps creation.
1.1. Objectives
The major objective of this work is to comprehensively map the spatial extent of
water objects in the Lena Delta region at high resolution and to determine their water
level height using remote sensing data and GIS methods. In order to realize this
purpose, satellite images of the RapidEye and Sentinel-2 missions and the highly
precise TanDEM-X digital elevation model (DEM) are used as input data.
Water level height of thermokarst lakes in the Lena Delta is an important
indicator to assess the extent of permafrost degradation that this landscape has
experienced in the past as well as to analyze the vulnerability of this region to future
inundation against the background of current sea level rise, which has been observed to
be around 1.84 mm per year in the Laptev Sea (Proshutinskiy et al., 2004). Moreover,
thermokarst lake water levels may add to the existing knowledge of Lena Delta
geomorphology (Grigoriev, 1993; Schwamborn et al. 2002) and reveal new insights into
interesting spatial patterns of lake distribution. By this means, it is useful for a better
understanding of arctic permafrost vulnerability against the background of climate
warming influence on thermokarst processes.
9
Finally, this study is based on the theoretical and practical base in the research
field of permafrost (i.e. French H. and Romanovskii N.), studies on thermokarst and
thermokarst lakes (i.e. Fedorova I., Grosse G., Morgenstern A., Günther F., Shur Y.),
and remote sensing (Grosse G., Günther F., Muster S.).
1.2. Study area
The Lena River Delta
The Lena River, which flows into the Arctic Ocean, is one of the biggest rivers
in Russia: 4400 km long, the mean annual discharge rate is near 16 800 m³/ s, the mean
annual sediment flux is about 680 kg/s for suspended and 170 kg/s1 for bottom
sediments according Alekseevsky, 2007. The months of maximal discharge is June one
third of discharge occurs in this month (Walker, 1998). Thus the Lena river is the major
terrestrial source of water and sediment for the Laptev Sea (Are & Reimnitz, 2000), and
it forms the largest delta in the Arctic (Fedorova et al., 2015).
The delta (72.0–73.8° N, 122.0–129° E) is situated in North-Eastern Siberia and
belongs to the typical Arctic tundra zone with continuous permafrost (Morgenstern et al.
2011). It is surrounded by the Laptev Sea to the west, north, and east and the
Chekanovsky and Kharaulakh mountain ranges to the south (Nitze & Grosse, 2016).
The total area of the delta is over 2 000 km² and includes more than 1 500
islands of various size, about 60 000 lakes, and numerous branches of the Lena River
(Are & Reimnitz, 2000). If the delta’s upstream limit is set as including the Bulkurskaya
Lena River branch to Tit-Ary Island, the delta area exceeds 32 000 km² (Fedorova et al.,
2015). Schneider et al. (2009) obtained an area 29 036 km², which is 98% of territory.
Thus according such estimation a total area of the Lena Delta equals 29630 km²
(Bolshiyanov et al., 2013). According to Muster et al. (2012) 21 719 km² of this area
represent land and the remaining areas are occupied by rivers and coastal zones.
One of the large islands in the delta, known as Erge-Muora-Sisse (Arga), has an
area of 6997 km². The Lena River delta (fig. 1.1) is a complex of more than 800 arms
with a total length of about 6 500 km. These branches flow in different directions, some
of them diverging, others converging. There are four major branches in the delta. The
main branch in the delta is the Trofimovskaya branch; from this branch the
Sardakhskaya branch diverges after Sardakh Island. The second largest branch by
volume is Bykovskaya channel that turns sharply to the east after Sardakh Island and
10
flows into Buor Khaya Gulf. The secondary branches are also Olenekskaya, which
flows west into the Kuba Gulf, and the northward flowing Tumatskaya. Recently, a
decrease in discharge has been observed in the Olenekskaya and Tumatskaya branches
(Fedorova et al., 2015). Water bodies of different size, shapes, depths and types of
formation cover about 20% of the delta’s land area (Muster et al., 2012). A great
number of ponds and lakes are presented in a variety of deltaic environments including
old river channels, terrace-flank depressions, thaw depressions, inter- and intra-dune
depressions, swales in ridges and swale deposits, low-centered polygons and the troughs
between polygons (Walker, 1998). Most of these ponds are related to the permafrost.
The climate of the Lena River Delta area is characterized by extremely cold,
long winters and short, cool summers. The annual mean air temperature on Samoylov
Island from 1998–2011 was −12.5 ⁰C (Boike et al., 2013). During most of the year, the
Lena Delta is in what might be called a "dormant state", in winter most of the water
bodies are frozen to depths of 1.5 down to 3 m and ground water is immobilized
(Walker, 1998).
Bykovsky Peninsula
The Bykovsky Peninsula and Khorogor Valley are part of the recent coastal
lowland of the Laptev Sea and are situated in the Russian North-Eastern Siberia in
southeastern direction of the Lena River Delta Grosse et al., (2005). An area of the
Khorogor Valley is about 86.2 km² and an area of the BYK is about 172.5 km² (Grosse
et al., 2005).
The peninsula is surrounded by large bays of the Laptev Sea. According to
Grosse et al., (2007) the peninsula is an erosional remnant of a Late Pleistocene
accumulation plain consisting predominantly of silty to sandy ice-rich permafrost
deposits of the yedoma. Maximal elevation of this remnant on the Bykovsky Peninsula
is about 43 m a.s.l., lower elevations of the Yedoma is about 25 m a.s.l. and of the
peninsula 0 m a.s.l.
The permafrost in this region is continuous and reaches depths of 300–500 m.
About 46% of the peninsula covered by deep thermokarst depressions (Grosse et al.,
2005) appeared due to early Holocene climate warming (Grosse et al., 2008), the third
part of thermokarst affected area is occupied by polygonal ponds and thermokarst lakes.
Lakes, predominantly of early Holocene thermokarstic origin are abundant not only in
the depressions, but also are situated on the yedoma uplands (Grosse et al., 2008).
11
Climate of the Bykovsky Peninsula is similar to the Lena Delta due to their close
location.
1.3. Geomorphological and Geological settings
Three main geomorphological units (river terraces) in the Lena River Delta are
identified by Schwamborn et al. (2002) after Grigoriev (1993) (fig. 1.1). The first terrace
is characterized by ice-wedge polygonal tundra, large thermokarst lakes and active flood
plains, which are affected by modern deltaic processes. This terrace formed during the
Holocene and occupies most of the central and eastern parts of the delta. According to
Schirrmeister et al., (2011), the second and third terraces, which dominate the western
and partially southern parts of the delta, are erosional remnants of arctic
paleolandscapes. The second terrace is characterized by frozen sediments that
predominantly consist of fluvial sands, which are several tens of meters thick and that
have been formed during the Late Pleistocene from > 52 to 16 kyr BP (Schirrmeister et
al. (2011). This terrace features many large thermokarst lakes and is located in the
northwestern part of the delta. The third and oldest terrace is an erosional remnant of a
Late Pleistocene plain consisting of fine-grained, organic-rich and ice-rich sediments,
characterized by polygonal ground and thermokarst processes (Schirrmeister et al.,
2003). Continuous permafrost which is one of the most important factors in delta
underlies the area to between about 400 and 600 m below surface (Grigoriev, 1960). It
is represented here in form of ice wedges, which are expressed on the surface as icewedge polygons and pingos (Walker, 1998).
12
Figure 1.1. Geomorphological division of the Lena Delta (after Grigoriev, 1993; Schwamborn et al.,
2002). Geomorphological units: black color – first terrace; grey – second terrace; light grey – first terrace.
Circles – river bed sediments
An active role in the Lena Delta formation and evolution plays tectonism during
the Pleistocene and also the Holocene. The modern seismicity indicates current vertical
block movements. It was noted that Olenyokskaya and Bykovskaya channels flow in
accordance with the general west-east direction of and along a Cenozoic fault line.
Moreover, the Lena Delta can be divided into a structural eastern and a western part,
which are charactzerised by general uplift and subsidence tendencies, respectively. This
is believed to be caused by the continuation of the sub-longitudinal Gakkel Ridge,
which crosses the Arctic Basin and propagates onto continental crust (Drachev et al.,
1998). This suggestion proved not only through trough seismic data, but is also
supported by geomorphological evidence such as a general drainage of most delta
channels towards east. A key role in the area evolution played sedimentation of ice
deposits during the Pleistocene that covered vast territories including most shallow
shelf, and sea transgression during Holocene. The Lena River carries a huge volume of
13
suspended sediment. About 30 % of this load is thought to reach the Laptev Sea, while
the remaining part of this volume is constantly redeposited between older islands, on
sandbanks, or expands the delta towards the Laptev Sea (Fedorova et al., 2015).
Thus, the delta formed as a result of vertical tectonic movement, continuous
sedimentation of Upper Pleistocene depositional units, on Holocene relief development,
and modern hydrological processes of the Lena River. The pattern of the main channels
is therefore predetermined by structural and near-surface geology, modern
sedimentation, and hydrology
14
2. Materials and Methods
The main goals of this work is to map the spatial distribution of water bodies in
the Lena Delta and to measure their water levels by means of remote sensing data and
Geographical Information Systems (GIS) methods. To achieve these goals, the specific
objectives are to:
create a map of water objects in the delta;
preprocess and correct DEM;
combine the map of water objects with corrected DEM to retrieve lakes
levels;
process and statistically evaluate the obtained results.
The implementation of these plans demand a great amount of data and computer
resources. The methodology and execution of this study is based on a combination of
digital satellite image processing and GIS analyses.
Broad-scale processes in the Arctic, such as hydrological, vegetation or climate
dynamics, are generally monitored with remote sensing data at different temporal and
spatial resolutions, but most often with resolution of 250 m or coarser (Stow et al.,
2004; Beck & Goetz, 2011; Fensholt & Proud, 2012; Goetz et al., 2011; Urban et al.,
2014). However, this scale is usually not sufficient enough to study different natural
processes in the Arctic such as thermokarst. Therefore, a large variety of remote sensing
studies employing higher resolution satellite imagery in conjunction with data obtained
during field expeditions have already been conducted in the past (i.e. Ulrich et al., 2009,
Morgenstern et al., 2011, Muster et al., 2012, Günther et al., 2015). Due to spatial
limitations of the remote sensing data that has been used in all these studies (i.e. ALOS,
Chris Proba, and airborne photography), only selected field sites have been analyzed at
a higher level of detail. In contrast Schneider et al. (2009) and Nitze and Grosse (2016)
used Landsat data with 30 m pixel size in order to characterize land cover classes and
land surface changes. Concerning the high heterogeneity of tundra surfaces in the region
of interest (the Lena Delta), it is important to take into consideration the effect of mixed
pixels, since many river distributaries, islands, lakes, and ponds are relatively small.
Therefore, it is of great value to use higher resolution remote sensing data, in order to
more precisely discriminate distinct features, such as water bodies for further study.
15
2.1. Data
2.1.1. Satellite images
For creation of a Lena Delta water-objects database, 58 high-resolution (ground
sampling distance: 6,5 m) multi-temporal RapidEye satellite images were acquired
within the framework of the RapidEye science archive (RESA) special area project
"Remote sensing of permafrost thaw-related landscape dynamics in the Lena Delta
region: coastal erosion, river delta changes, and thermokarst" (principal investigator
Frank Günther) (fig. 2.1). The images were provided at processing level 1B and span a
time period from June 2009 to September 2015 (tab. 2.1). The RapidEye images which
cover almost entire Lena Delta provide 5 spectral bands:
1.
Blue (440-510 nm);
2.
Green (520–590 nm);
3.
Red (630–685 nm);
4.
Red Edge (690–730 nm);
5.
Near infrared (NIR) (760–850 nm).
RapidEye – is the commercial operational class Earth observation system using a
constellation of 5 satellites that provide unparalleled performance in order to achieve a
high revisit frequency and thus temporal resolution, which can be used for change
detection purposes and to capture seasonal variability (Behling et al. 2014),.
The RapidEye Basic product is radiometric and sensor corrected, providing
imagery from the spacecraft without correction for any geometric distortions inherent in
the imaging process. Prior to direct georeferencing, the spectral characteristics of all 56
images have been normalized by atmospheric correction (ATCOR module of PCI
Geomatica) to top of atmosphere reflectance at particular dates and under consideration
of varying solar zenith and solar azimuth values, and subarctic summer rural conditions.
Although the imagery is not mapped to a cartographic projection, the wide area imagery
(70 x 140 km) comes with rational polynomial coefficients (RPC) that provide all
spacecraft telemetry for geometric processing of the data into a geo-corrected form.
Based on orthorectified (terrain corrected) very high resolution (0.5 m pixel size)
GeoEye and WorldView satellite images that were available for three key regions
(Kurungnakh, Sobo-Sise, and Bykovsky Peninsula) distributed across the southern Lena
Delta, ground control points were collected for RapidEye scenes overlapping with these
sub regions. Finally, based on RPCs, all remaining RapidEye scenes have been included
16
into the bundle block adjustment procedure through a large amount of common points
(tie points) between images that were collected automatically. By this means, handling
of all images within one adjustment procedure not only provides a highly self consistent
set of satellite images relative to each other, but also in absolute geocoding accuracy.
Finally, the atmospherically corrected images were orthorectified based on DEM terrain
elevation data and resampled to 5 m spatial resolution within the UTM 52N coordinate
system.
The Provided images represent different years and different seasons, because
images for one year or one season don’t cover all parts of the Lena Delta due to frequent
cloud cover. Although this seems to be not ideal from a perspective of capturing equal
conditions, it offers the possibility to actually map the entire delta with all its strong
seasonal variations from snow melt water saturated surfaces to dry conditions in late
summer and phenological vegetation cover changes.
Figure 2.1. Imaging frequency of the RapidEye archive data between the start of operations in February
2009 to June 30, 2014 (eoportal.org), red frame outlines the extent of the Lena Delta special area
RapidEye Science Archive project
With respect to these effects, a review of obtained RapidEye data showed that only
images acquired during late spring, summer and early September can be further
considered to get a correct result. Because during the second half of autumn winter and
spring the territory is covered by snow and water bodies are covered by ice. Therefore,
17
after whole data preprocessing path only 46 georeferenced and orthorectified images,
that present late June to early September conditions were used in a further analyses (tab.
2.1).
Table 2.1. Data acquisition
year
2009
2010
2011
2014
months
Number of shots
June, July
5
July, August
8
June, July, August
5
May, June, July, August, 19
September
June,
July,
August, 9
September
September
5 (SENTINEL-2 images)
2015
2016
Although a lot of RapidEye images were acquired, still some data gaps in the
southwestern and northwestern part of the delta remained. In order to fill these gaps it
was decided to use also SENTINEL-2 images. Since five Sentinel-2 images acquired in
September 2016 covered the entire region of our interest, they have been used not only
to fill the data gaps, but also as another independent dataset for water body mapping
across the entire delta. Thus was chosen 5 images for September 2016 with channels
Resolution of 10 meters.
SENTINEL-2 is a European wide-swath, high-, pushbroom multi-spectral
imaging. The full mission specification of the twin satellites flying in the same orbit but
phased at 180°, is designed to give a high revisit frequency in 13 bands with different
spatial resolution (Sentinel-2_User_Handbook).
Within this project, 5 SENTINEL-2 images, each containing 4 Visible-Near
Infrared (VNIR) bands (490 nm (B2), 560 nm (B3), 665 nm (B4), 842 nm (B8) with a
resolution of 10 meters were used. Similar to the RapidEye data, all SENTINEL-2
images have been atmospherically corrected and radiometrically normalized. Because of
the higher spatial resolution of the RapidEye imagery, SENTINEL-2 images were then
slightly adjusted to this dataset by using only common feature tie-points. Due to near
nadir viewing geometry of SENTINEL-2 and given the generally flat topography in the
study region, no additional orthorectification was carried out.
2.1.2. Digital Elevation Model
18
As the base for elevation measurements the TanDEM-X DEM was used.
TanDEM-X DEM is global Digital Elevation Model (DEM) of the land masses of the
Earth. This DEM was generated from bistatic X-Band interferometric SAR acquisitions
from two satellites (TerraSAR-X (TSX) and a second satellite TanDEM-X (TDX),
which orbit Earth at an altitude of around 500 kilometers (EOC TanDEM-X, 2016).
TanDEM-X DEM data is provided at either 12 m or 30 m nominal ground resolution.
Here the higher elevation product is used, which has been granted to AWI based on the
DEM_GEOL_1684 proposal “Application of TanDEM-X elevation data for remote
sensing change detection of permafrost thaw in the Laptev Sea region” (Principal
investigator Frank Günther).
The elevations are defined with respect to the reflective surface of X-Band
interferometric SAR returns. Therefore, the TanDEM-X DEM products represent
predominantly a Digital Surface Model (DSM), which generally corresponds to ground
elevation in this tundra environment. The acquisition of this model took four years,
from December 2010 to January 2015. In order to reach the target accuracies all land
masses are covered at least twice in the same looking direction, but with different
baselines (EOC TanDEM, 2016).
The elevation values of the initial DEM sub-tiles represent the ellipsoidal heights
relative to the WGS84 ellipsoid in the WGS84-G1150 datum. One elevation value h
reflects a weighted height average for a given pixel, computed by the height values of
all contributing DEM scenes (EOC TanDEM-X, 2016). Extreme important parameters
which determine the quality and applicability of each DEM are different types of
accuracy:
1.
Absolute horizontal accuracy is defined as the uncertainty in the horizontal
position of a pixel with respect to a reference datum, caused by random1 and
uncorrected systematic 2 errors. The value is expressed as a circular error at 90%
confidence level.
2.
Absolute vertical accuracy is the uncertainty in the height of a pixel with respect
to a reference height caused by random and uncorrected systematic errors. The
value is expressed as a linear error at 90% confidence level.
3.
Relative vertical accuracy is specified in terms of the uncertainty in height
between two points (DEM pixels) caused by random errors. The corresponding
values are expressed as linear errors at 90% confidence level (LE90) [A1]. The
reference area for two height estimates is a 1° x 1° area, corresponding to
approximately 111 km x 111 km at the equator.
19
The spatial resolution is determined by the pixel spacing in latitude direction is
0.4 arcseconds, which corresponds to 12.37 meters at the equator and to 12.33 meters
near the poles and in longitudinal direction in zone between 70° – 80° North/South it
equals to 1.2’’ (12.69m – 6.44m) (EOC TanDEM-X, 2016)
Table 2. 2. Specification of accuracy of the TanDEM-X DEM (2016).
For the Lena Delta region one product tile has an extent of 1° x 2°. It means that
to cover the entire region it was necessary create a mosaic of nine TanDEM-X tiles that
was done using GEOMATICA software. Tandem mosaic was reprojected from the
initial Geographical Coordinate System on the ellipsoid WGS84-G1150 to the geodetic
coordinate system UTM WGS84 52N which is suitable for the Lena Delta. In order to
consider geoid height undulations, it was decided to transform ellipsoidal elevation
values to heights above mean sea level (m a.s.l), therefore the values of DEM were
recalculated from elevations above ellipsoid WGS84-G1150 to m a.s.l. heights using the
geoid EGM2008 at 2.5’ spacing.
2.1.3. Additional GIS layers
In this work polygon vector layers containing the spatial extent of terraces in the
Lena Delta were provided as ready to use dataset by Anne Morgenstern, see
Morgenstern et al. (2008, 2011).
Space-borne laser altimetry data (Release 34) from the Geoscience Laser
Altimeter System (GLAS) instrument that was aboard the NASA Ice, Cloud, and land
Elevation (ICESat) satellite has been obtained from the National Snow & Ice Data
Center (NSIDC) in Boulder (USA). The data were processed by Frank Günther for the
study region considering only measurements with laser transmit energies below 30
millijoules, in order to exclude possible cloud height measurements. A correction to
20
elevation for saturated waveforms has been applied. Since ICESat uses another ellipsoid
(Topex/Poseidon) elevation values had to be converted to WGS84 ellipsoidal heights
and finally to heights in m a.s.l. for comparison with TanDEM-X data. ICESat acquired
data during several single campaigns between February 2003 and October 2009. Point
measurements of the earth surface elevation within 35 m radius footprints had a 175 m
spacing along-track and 16 – 25 km across track in the Lena Delta (fig. 2.2).
Figure 2.2. Image map (Google Maps) over the Lena Delta with ICESat/GLAS tracks
2.2. Methods
2.2.1. Satellite images sets processing
In order to receive a consistent scheme of water objects with elevations we
applied a strategy of data refinement, fusion, examination and analysis to the study area.
Accurate mapping and heights determination using multi-temporal, multi-platform
remotely sensed data requires consideration of various distortions, including distortions
associated with the platform, the map projection, and shape of the earth’s surface
(Gunther et al., 2013). For the overview and preprocessing (fig. 2.3.) of satellite images
was used Geomatica software package provided by PCI Geomatics.
21
Optical remotely sensed data are expressed in arbitrary units such as digital
number (DN), and always affected by sensor characteristics, illumination geometry and
atmospheric conditions (Vaudoura et al., 2014). The atmospheric effects influence on
the signal registered by remote sensors should be minimized using special methods in
order to provide reliable spectral information (Bernardo et al., 2017). Atmospheric
correction allows researchers eliminate different types of clouds, hazes and convert
digital numbers into relatively similar surface reflectance taking into account various
factors.
PCI Geomatica‘s Atmospheric Correction provides a variety of atmospheric
corrections methods (PCI, 2015).
From several techniques for atmospheric correction was chosen ATCOR Ground Reflectance workflow, which calculates ground reflectance values for supported
optical imagery and optionally performs haze removal and cloud masking (fig. 2.3). The
output is in reflectance values (0 - 100%), which are based on processed digital numbers
(PCI, 2015).
Geomatica software package allows automatize an algorithm of processing by
means of model creating for batch processing of remotely sensed data. The modeller
module provides access to a number of standard operations such as data import and
export, as well as most PCI Geomatics processing algorithms (PCI, 2015). This
opportunity was used for automatization of further data pre-processing steps, such as
atmospheric correction.
For this correction, a model was created, which uses all the metadata of the
ingested imagery, terrain elevation information (DEM), which was provided by AWI
Potsdam and visibility information/settings (Atmospheric conditions – Subarctic
Summer, Aerosol type - Rural, visibility – 100 km.). All images were corrected this
way.
The whole set of RapidEye and SENTINEL-2 images covers an area wider than
the actual Lena Delta. Consequently, in order to accelerate the data processing it was
necessary to exclude areas which don’t correspond to our region of interest from all
images by creating appropriate clip masks.
22
Figure 2.3. Result of haze and cloud masking. Blue spots – cloud mask, red spots – haze mask
Before the remotely sensed information can be gathered in a manner that is
suitable for a mapping or GIS analysis, the imagery data must be prepared in a way that
corrects different distortion effects and transforms in the required projection. Transfer of
satellite image to the geometry of the map consists of georeferencing - correction for
distortions connected with the acquisition system, transformation in the required
projection (gis-lab.info, 2012); and orthorectification correction for relief-induced
displacement effects using a rigorous math model and a digital elevation model (F.
Günther et al.: The disappearing East Siberian Arctic island Muostakh). To carry out
computation of the math model and orthorectification Geomatica’s module OrthoEngine
was used (fig. 2.4).
Project creation
Data Selection
Calculation of Sensor
Model
GCPs and TPs automatic
collection
Orthorectification
using existing DEM
Figure 2.4. Satellite orbital modeling workflow diagram
Processing of RapidEye imagery was done within an OrthoEngine project with
Math Modelling Method for optical satellite data based on Rational Functions Math
23
Model. The Rational Functions Math Model is a simple Math Model that builds a
correlation between the pixels and the ground locations. The math model is computed
for each image separately using four polynomials, which are functions of latitude,
longitude, and height or elevation. To obtain polynomial coefficients it is necessary to
collect Ground Control points. A ground control point (GCP) is feature that can be
clearly identified in the raw image for which ground coordinates there are known and
determines the relationship between the raw image and the ground by associating the
pixel and line image coordinates to the x, y, and z coordinates on the ground (PCI,
2015).
Although the western part of the Lena Delta belongs to UTM 51N zone, the
project coordinates were WGS 84 UTM projection Zone 52N, because all of the
reference data for ground control point collection were located in the eastern part of the
Lena Delta. Ground control points were sampled automatically from already rectified
and georeferenced (geocoded) 4 GeoEye images with ultrahigh resolution, which were
provided by AWI Potsdam. Expanding manual ground control point collection with
automatic collection, it has been possible to exclude inaccurate points and to constantly
correct the image model during iterative refinements based also on tie points. Manual
correction mostly comprised removal of false points with unacceptable errors, often
situated on the water or clouds.
As a result 271 ground control points were obtained. It should be noted that there
was a lack of ground control points for the western part of Lena Delta, because all
rectified GeoEye images are situated in the Eastern part of the Lena Delta, which could
have an influence on the accuracy of the rectification. 30 Ground Control points with
significant residual errors were transferred to Check Points to get the independent
accuracy assessment of the math model, because check points aren’t considered during
math model calculation.
In order to extend ground control over entire area and therefore beyond the
ground control clusters, 7013 tie points were collected automatically for 58 images. A
tie point is a feature that is clearly identified in two or more images and that can be
selected as a reference point. These points identify how the images in the project relate
to each other. The result of sampling was corrected manually. Using collected Ground
Control Points and Tie Points a rigorous math model was computed for 55 images that
is often referred to as a bundle adjustment. The math-model solution calculates the
position and orientation of the sensor—the aerial camera or satellite—at the time when
24
the image was taken, GCPs, wherein tie points are automatically weighted inversely to
their estimated errors (PCI, 2015).
For project quality estimation in addition to Check Points, Root Mean Squared
(RMS) Errors for the entire project were calculated on the base of single image Residual
Errors. Residual Error is the difference between the real coordinates of the ground
control points or tie points and where those points are according to the computed math
model (PCI, 2015). As a rule of thumb, a reasonable RMS error generally should be
around half of the resolution of the image (one pixel), in case of this project, it means
that residual errors should be under 6.5 meters.
Table 2.3. Residual Summary for 55 Images
number
X RMS
GCPs
271
0.28
Check points
30
0.85
Tie points
7013
0.13
RMS (x,y,z) for worst 5% of points 0.66, 0.74
Y RMS
0.32
0.83
0.15
The final Math Model and the Digital Elevation Model were used for
orthorectification of RapidEye images. Resampling of pixels (extraction and
interpolation of the grey levels from the original pixel locations to the corrected
locations) were carried out using Cubic convolution method, which determines the gray
level from the weighted average of the 16 closest pixels to the specified input
coordinates and assigns that value to the output coordinates. The resulting image is
slightly sharper than one produced by Bilinear Interpolation, and it does not have the
disjointed appearance produced by Nearest Neighbor Interpolation (PCI, 2015).
Actually SENTINEL – 2 images are already snapped with the precision higher
than 1 pixel but because of discovered difference in spatial position between
SENTINEL – 2 and RapidEye images, SENTINEL-2 imagery was also decided to
georeference in Geomatica OrthoEngine using RapidEye dataset.
To map water objects two different methods were tested:
Firstly it was tried to create mosaic of RapidEye image for the whole Lena Delta
and then classify it using combination of Unsupervised and Supervised image
classification in PCI Geomatica. Obtained results were unsatisfactory due to several
classification problems that mostly were related to strong seasonal variability of the lake
water turbidity. While many lakes showed very clear water conditions, others nearby
exhibit very turbid sediment loaded water surfaces that were misclassified as land. This
25
couldn’t be solved with additional and refined training areas because of the limited
spectral information with five bands.
The second method, which is finally used in this study imply the extraction of
water objects from every single image using only the NIR channel. This approach
requires following preprocessing steps: NIR channels extraction – images in NIR
channel correction (clipping of regions with cloud shadows). After considering seasonal
aspects and image coverage inspections, 35 RapidEye and 5 SENTINEL-2 images were
used for further water body mapping.
2.2.2. Creation of water mask
The extraction of water objects was based on an empirical threshold that has
been determined in top of atmosphere reflectance RapidEye and SENTINEL-2 images.
Because of varying spectral band characteristics between RapidEye and SENTINEL-2,
mapping was done separately for each of both datasets, but followed a common
approach:
1.
Overlay of all images and searching for the minimum value at a particular pixel
location, where then each output cell value is a minimum value of the values
assigned to the corresponding cells in the input raster map layers using
algorithm GRASS r.series (grass.osgeo, 2017). (fig. 2.5a)
2.
As pixel value domain which has been determined as being associated with open
water, two separate ranges from 0.1 to 8.8 and from 0.1 to 2.7 for RapidEye and
SENTINEL-2, respectively were used. This separate treatment has been
necessary, because after atmospheric correction with the same settings, images
from these two systems showed different incomparable reflectance values. This
selection was also done using GRASS r.series (grass.osgeo, 2017). As a result,
two different layers portraying water bodies of the Lena Delta were obtained
(fig. 2.5 b, c).
3.
Received rasters were converted into binary form using GDAL raster calculator
(fig. 2.6 a, b).
4.
Obtained rasters contained noise (a lot of small objects and forms) that was
removed using filtering algorithms. After testing a variety of methods with
different settings, it was decided to use a Majority filter from SAGA with a
window size of 3*3 and then to filter resultant rasters once again with OTB
26
binary Morphological filtration operation Opening with Structuring Element
Type – Ball, radius 5 pixels (fig.2.7).
5.
Cleaned rasters were converted to polygons using algorithm GDAL “vectorize
raster layer”.
6.
Resulting vector layers contained the masks of all water objects (fig. 2.8 a,b).
Both masks were combined to one single water body shapefile using the
Dissovle function in GIS.
For the further analysis received files were divided on two types: the first one,
layers, which contain isolated water objects or lakes, the second one layer that contains
connecting river channels (fig.10). After splitting the whole water body dataset into
these types, joined layers of RapidEye and SENTINEL-2 water channels and lakes
(isolated water bodies) were created using QGIS and ArcGIS tools. Combination of two
data sets provided several advantages. Using of SENTINEL-2 data allowed to cover
gaps in RapidEye imagery and to get sharper water surface representation. On the other
side RapidEye dataset provides more detailed scene and was better for shallow water
classification.
Wherein in the layer with lakes were selected only objects with the size more
than 100 m² and 99 objects with an area less this threshold were deleted. The total area
of deleted features is about 0.002 km².
a
b
27
2
1
c
d
Figure 2.5. a. Lakes obtained by SENTINEL-2 in NIR channel; b. Lakes after
application a r.series algorithm on SENTINEL-2 images in NIR channel; c. Lakes
obtained after application of r.series algorithm on RapidEye images in NIR channel; d.
Comparison of r.series algorithm results, (SENTINEL-2 image (2) situated inside
RapidEye (1)
a
b
Figure 2.6. Differences in resolving water body shape: a. Binary image based on RapidEye
data; b. Binary image based on Sentinel data
Figure 2.7. Lake on SENTINEL-2 image covered by mask, filtering method (SAGA majority filter
window 3*3 pixel)*OTB binary morphological filter opening with structuring element ball with r=1
28
a
b
Figure 2.8. Comparison of obtained masks with input raster data (scale 1: 50 000). Cian masks based on
the SENTINEL-2 imagery, magenta – based on the RapidEye imagery. a. Pure masks; b. Masks
superposed on the input raster data, where small dots – pixels of water on the input raster, that were
excluded as result of the filtering
Figure 2.9. Vector scheme of main river channels of the Lena Delta. Scale 1: 2 500 000
2.2.3. DEM refinement
The TanDEM-X DEM is not hydrologically corrected and is not edited for water
surfaces. This results in very noisy water level elevations in the DEM. Therefore, the
very detailed vector mask of water bodies was used to edit heights of the lakes level in
the DEM. To improve the height values for lake surfaces the following method was
29
applied. From the lake layer only objects with an area more than 1000 m² were selected,
to edit areas of lakes location on TanDEM-X DEM using the DEM editing toolbox in
Geomatica. The target of editing was to remove all bumps and pits by leveling and
interpolating accurate heights along the shoreline over the surface of lakes. By this
approach it has been possible not just to average wrong elevations of the lake surface,
but to use the precise elevations of the lake margins and then to fill lakes surfaces with
values from the edges of the polygon (PCI, 2015). Thus was used a following sequence
of processing: “Remove Noise” – “Remove bumps” – “Remove pits” – “Fill from
edges”. The size of the filter box was set as 10 pixels.
To refine surfaces of several prominent water objects were manually rectified
shorelines polygons to avoid mistakes and applied method “Opposite ends fill”. The
filter examines the selected polygon on each short end and finds an average elevation
for each to use them in an interpolation (PCI, 2015).
2.2.4. Computation of characteristics
To extract elevation values from the DEM and connect them with appropriate
polygons in vector layer, that display water objects was used operation QGIS “Zonal
statistics”, which computes: minimum, mean, maximum values, mode, median, standard
deviation, count of cells, quantity of unique values of cells, and other parameters inside
zones of interest. Also for all lakes were derived areas and perimeters.
Worth saying that despite of refinement of water surfaces on the DEM, it wasn’t
possible to get unique height values for appropriate lake. Consequently it was necessary
to select the most plausible way for heights definition.
Highly important step of this study was estimation of elevation data extracted
from the DEM and the choice of the most realistic value of lakes level height, calculated
from the DEM. For this objective were used laser altimetry data. GLAS data provided
by AWI were stored as vector layer with 104485 polygonal footprints of measurements
for the entire Lena Delta. To avoid influence of snow cover on altitude measurements
for the further analysis were selected only data obtained in period from the end of
September until the end of November (Laser identifiers: L2A, L3A, L3D, L3G, L3I).
By means of GIS spatial query were acquired only laser altimetry footprints falling into
the lakes. The square of measurement footprint is about 3800 m², all footprints with an
area over water surface < 3000 m² were excluded in case they fall into water – land
30
area. Thus were obtained 2985 elevation values for 762 lakes. For the large lakes with
several altitude measurements were calculated mean values.
F. Günther proposed three options for heights calculations (personal
consultations) on the basis of DEM data:
1.
; 2.
where: h – assumed lakes level height,
; 3.
,
- mean value for each lake, σ – Standard Deviation (std), Me –
median.
Thus were obtained three sets of heights extracted from DEM and one set
received by laser altimetry (accepted for the standard). Pair comparison between these
series and their accuracy evaluation showed that elevation calculated as difference
between mean values and std is the most similar to laser altimetry data.
Table 2.4. Comparison of laser altimetry and DEM data
Sum of
elevations, m
Root Mean
Squared Error
Correlation
694.16
699.95 925.0908
0.9544
0.9747
0.9584
0.9746
1.1018
0.9712
Mentioned above supposed receiving of real heights of water surfaces, that
should be close to the shore line minimum elevation.
31
3. Results
Analyses of the results are based on a GIS layer (ESRI shape file), which
contains the shape of all water objects. Furthermore, a comprehensive attribute table
stores all relevant information on lake area, elevation above sea level and the
assignment to particular river terrace. As a result, 198851 features including ponds,
lakes, river channels with a size of more than 100 m² and 118947 features with an area
of more than 1000 m² were obtained for the whole Lena Delta including Bykovsky
Peninsula.
Figure 3.1 shows the complete scheme and spatial distribution of water objects.
There were found 189497 ponds and lakes with the size more than 100 m² including
118224 objects with an area more than 1000 m² and 29631 lakes of more than 1 ha.
These objects were then investigated in more detail. A total area of the watercourses in
the delta mapped in this study is about 4237 km². Moreover a total length of channels in
the Lena Delta, which is about 13132 km (12819 km belongs to Lena River
distributaries), was calculated using acquired map of water objects.
32
Figure 3.1. Vector scheme of water objects of the Lena Delta and Bykovsky peninsula.
Scale 1: 1 750 000
3.1. Spatial statistics
The total area of 189 497 lakes >100 m² found within the study area (the Lena
Delta including the Bykovskij Peninsula) amounts to 3408.06 km². Out of this, 188 968
lakes with an area of almost 3390 km² are situated within the actual delta extend,
meaning that 11.74 % of the delta are occupied by lakes. For further analysis, the Lena
Delta area according to Schneider et al. (2009) (29036 km²) was taken, which is
considered as 98% of entire area (Bolshiyanov et al., 2013), excluding the most
southern apex of the delta that is not covered by the DEM used in this study.
Distribution of lakes across the delta with respect different terrace levels is provided in
table 3.1.
33
Tab. 3.1 Spatial distribution characteristics of lakes in the study region
Total lakes
area [km²]
Number of
lakes
Percentage
of lake
number
fraction
[%]
Area of
terraces
[km²]
Areal
fraction of
lake
coverage
per
region[%]
Percentage
of lake areal
fraction[%]
Entire
Lena
Delta
3408.02
188966
100
23813**
11.7%
100
First
terrace
2239.8
132798
70.3
15840.1*
14.1%
65,7
Second
terrace
1052.19
52158
27.6
6098.6*
17.3%
30.9
Third
terrace
97.86
4010
2.1
1711.6*
5.7%
2.9
BYK***
16.41
(18.17)****
345
(526)****
172.5***
9.5%
*According Morgenstern, (2005); ** According Schneider et al., (2009); ***Bykovsky Peninsula
according Grosse et al., (2005); **** with adjacent part of Khorogor Valley
17%
20%
14%
% of Area
15%
12%
10%
6%
5%
0%
Entire Delta
1 Terrace
2 Terrace
3 Terrace
Figure 3.2. Percent of total terrace area covered by lakes
34
S km²
4000.0
3 390
2 240
3000.0
1 052
2000.0
98
1000.0
18
0.0
1 Terrace 2 Terrace 3 Terrace Entire Delta
BYK
Figure 3.3. Lakes covered area by terraces. BYK – Bykovsky peninsula
The results show that percentage of the Lena Delta territory covered by lakes is
12%. This areal fraction is considerably higher than, for example, the percent of
territory covered by lakes in Russia is near to 2.5% (http://water-rf.ru). Among the Lena
Delta terraces, values are particularly high for the first and especially for the second
terraces (tab. 3.1, fig.3.2). The percentage of open water covered area on the third
terrace is much lower in comparison with the whole Lena Delta. In absolute terms, the
greatest contribution to the lacustrine area is due to the first terrace with almost 133 000
of lakes covering a total area of 2240 km². Although 70% of the total amount of lakes is
concentrated on this terrace, their areal fraction is somewhat lower at around two-third
of the total lake area in the Lena Delta (fig. 3.3, tab. 3.1).
Tab 3.2 Statistical characteristics of the lakes elevation and area by terraces in the Lena Delta
1st terrace
2nd terrace
3rd terrace
entire Delta
area, m²
elev., m
area, m²
elev., m
area, m²
elev., m
area, m²
elev., m
mean
16866.3
4.4
20173.1
8.9
24402.9
20.3
17938.9
5.9
std
114026.4
3.3
267460.4
6.2
136179.4
11.8
171109.1
5.5
median
1550.0
3.8
1625.0
8.0
1600.0
19.0
1575.0
4.6
max
2239.8
2584500.
0
64.3
24318275
.0
64.3
min
125.0
225.0
0
125.0
0.0
1052.19
0
125.0
0
Tab 3.3 Statistical characteristics of lake elevations and surface areas on Bykovsky Peninsula
mean
area, m²*
34538.6
elevation, m*
8.6
area, m²
47592.1
elevation, m
10.5
mode
400.0
0.0
400.0
3.1
273113.7
10.8
335739.9
12.6
2550.0
3.4
2550.0
3.4
std
median
35
max
5903000.0
40.2
5903000.0
40.2
min
225.0
0
225.0
0
* including adjacent part of Khorogor Valley
As described in the methods, this study only considers lakes with an area of
more than 100 m², because of technical and methodical limitations. Average lake size in
the study region is about 1.8 ha, while maximum size is up to 25 km² (tab. 3.2). The
high standard deviation of ±17.1 ha is also reflected in a considerably lower median size
of 1.6 ha. The third terrace is characterized by the largest (24.4 ha) and the first terrace
by the smallest mean size of lakes (16.9 ha). However, the median size is relatively
similar across all terraces. Standard deviations of lakes areas are highest for the second
terrace two times higher as for other terraces moreover it is quite high when considering
lakes of the entire delta.
a
130.79
b
209.79
37.57
117623 1796
461
<0.001 km²
421
40
0.001 - 0.25
739.17
69617
1452.19
0.25 - 1
1-5
5 - 10
<0.001 km²
0.25 - 1
> 5 km²
0.001 - 0.25
1-5
838.54
> 10 km²
Figure 3.4. Lakes mirror size in the Lena Delta and Bykovsky Peninsula. a – number of lakes in each
group relatively to lakes area. b. Lakes covered area relatively to lakes area (1 <0.001 km²; 2. 0.001 –
0.25 km²; 3. 0.25 - 1 km²; 4. 1 – 5 km²; 5. 5 – 10 km²; 6. > 10 km²)
General distribution of lakes area over the entire Lena Delta is displayed in the
Fig.13. Grouping by area for analysis is based on Bolshiyanov et al. (2013) with adding
a special group of lakes with area under 1000 m². Majority of lakes in the delta 117623
have an area between 0.25 - 1 km² (fig.3.4a). Herewith only a minor part of lakes can be
considered as large, with an area more than 1 km². Fig. 3.4b shows that although
quantity of ponds and small lakes in the first group stays on the second place, an area
which is covered by this group is least in comparison with other groups. For the second
group the number of lakes correlates well with a territory occupied by this group. It
should be noted that small number of relatively major lakes plays a great role in the
36
water area distribution. For example number of lakes in the third group is about four
times more than in the fourth group, but values of filed areas are very similar.
Distribution territory covered by lakes relatively to terraces is considered
(fig.3.5) according grouping of lakes area based on the work of Boike et al. (2013),
where lakes divided on three groups according area: polygonal ponds, polygonal lakes,
thermokarst lakes. In this work one more category is added: large thermokarst lakes,
with an area of more than 1 km². The major area is occupied by thermokarst lakes (third
group) for total area as well as for the first and the third terraces. Moreover area of the
third group exceeds a total area of other groups, for the third and first terraces as well as
in general. The second terrace differs from the other terraces, because major part of the
entire lake area is occupied by large lakes with an area more than 1 km². However,
figure 14 demonstrates that the largest amount of all lakes is situated on the first terrace,
across all lake size classes.
lakes area, km²
2500
2000
1500
1000
500
0
1
2
3
4
groups
terrace 1
terrace 2
terrace 3
Figure 3.5. Distribution of territory covered by lakes relatively to 4 lakes size groups and terraces. 1. area
<0.001 km² (polygonal ponds); 2. 0.001 – 0.01 km² (polygonal lakes); 3. 0.01 - 1 km² (thermokarst lakes);
area>1 km² (large thermokarst lakes)
The most common lake size all over the delta generally as well as separately for
each terrace are polygonal lakes according to classification of Boike et al. (2013), which
have an area between 1000 m² and 1 ha (fig. 3.6). Especially on the first and particularly
on the second terrace they account for around a half of all lakes, with 45.8% and 52%,
respectively (fig. 3.6). Water bodies from the third and the fourth groups with an area of
more than 1 ha (thermokarst lakes) are relatively sparse (fig. 3.6), only 29541 from
188968 lakes in total within the delta, according to 16%. However, as it was mentioned
before they are responsible for the majority of areal lake coverage in the delta (fig.3.5),
where 3060.9 out of 3389.89 km² are covered by these lake types, corresponding to 90%
37
areal fraction. In contrast, small water bodies like ponds with an area between 100 m²
and 1 ha occupy the smallest territory despite of their number, where a huge amount of
polygonal lakes with 48% of all lakes in delta corresponds to a relatively small areal
fraction of only 9%. The shape and the size of polygonal ponds are defined by icewedge structures (Boike et al., 2013).
Terrace 1
Terrace 2
270
2196
0
Terrace 3
6406 168
1825
5
4973
4
2
3
22
1464
2729
9
6083
6
1
715
4
1
1810
2
3
4
1
2
3
4
Figure 3.6. Number of lakes by size. Groups: 1. area <0,001 km²; 2. area between 0,001 and 0,01 km²; 3.
0,01 and 1 km²; 4. area>1 km²
Figure 3.7. Distribution of lakes by area on Bykovsky Peninsula A: Total lakes area for each group; B:
Number of lakes in each group. Groups: 1. area <0,001 km²; 2. area between 0,001 and 0,01 km²; 3. 0,01
and 1 km²; area>1 km²
Lakes on Bykovsky Peninsula were considered together with Lena Delta. Figure
3.7a shows that the distribution of lakes by groups on the base of covered area as well
as distribution of groups by quantity of water bodies in each group is similar to
allocation in Lena Delta (fig.3.5; 3.6). On the other side relative quantity of polygonal
lakes is higher comparatively to Lena Delta terraces.
3.2. Lake height distribution
38
For all lakes their water level height was evaluated based on a TanDEM-X
digital elevation model with 12m spatial resolution. The results of statistical analysis
(mean, median and std values) of lakes elevation generally reflects all terrace levels of
the Lena Delta. Figure 3.8 shows that lakes on the third terrace lay above lakes from the
second terrace which lay consequently above lakes from the first terrace. For the first
time, this work determines that mean elevation of all lakes in the Lena Delta is about 6
m above sea level. For the third terrace, which is according Are and Reimnitz (2000) 20
– 60 m high coastal plain composed from ice-complex deposits, median lake elevation
is about 19 m (tab. 3.2). Mean elevation of lake water levels on the 20 - 22 m high sand
terrace (Are and Reimnitz, 2000) or the second terrace is about 9 m (tab. 3.2). Finally
mean lake water level of the first terrace is about 4.5 m.
Figure 3.8. Statistical distribution of lakes level elevations corresponding to terraces. Inside boxes – 2nd
and 3rd quartiles (25-75%), whiskers - max and min values, circles – mean values; BYK – Bykovsky
Peninsula
Statistical characteristics of lake elevations display noticeable differences
between the main geomorphological elements of the study area, including three terraces
in the delta and the Bykovsky Peninsula. Based on figure 17 it is obvious that
distribution interval of lakes elevation values is greatest for the third terrace and the
maximum elevation of lakes water level is above 64 m. Even 50% percent of central
values show notable scattering. Moreover significant distribution and heights can be
39
observed on the Bykovsky Peninsula. This is of particular notice since the Bykovsky
Peninsula geomorphologacally belongs to the third terrace (Grigoriev, 1993).
b
a
1200
area km²
60000
1000
frequency
50000
40000
30000
800
600
20000
400
10000
200
0
0
0-0.5 0.5-2 2-5 5-10 10-20 20-40 >40
elevation, m
elevation, m
Figure 3.9. Histograms lakes and elevation. a – frequency of lakes relatively to elevation; b – area
covered by lakes relatively to elevation. Red line - trend
Most often lakes in the Lena Delta and on Bykovsky Peninsula are located at
heights between 2 and 10 m (fig.3.9 a, b). Mode for elevation for the entire Lena Delta
and Bykovsky peninsula is 0 m Exception is the third terrace, where mode for elevation
is more than 18 m. Only relatively few lakes are situated at heights above 40 m. On the
other side more than 10000 ponds and lakes are located at heights under 0.5 m. The
highest lake situated on the third terrace on the elevation of 64.3 m. The highest lake on
Bykovsky Peninsula has an elevation above 40 m. General trend shows moderate
decrease of lakes number with increasing height. Decreasing of area covered by lakes
with elevation trend is slightly steeper. The pattern of lake area relative to elevation
(fig.3.9b) generally matches with the number of lakes relative to elevation (fig.3.9a).
However, the largest area is occupied by lakes with an elevation between 2 and 5 m.
40
Figure 3.10. Distribution of lakes by elevation on Bykovsky Peninsula. Boxes – 2nd and 3rd quartiles
(50% of data), whiskers – 3 max and min values, circles – mean values. Groups: 1. area <0,001 km²; 2.
area between 0,001 and 0,01 km²; 3. 0,01 and 1 km²; area>1 km²
Statistical characteristics, which exhibit the relation between lake elevation and
lake size were evaluated on the example of Bykovsky Peninsula (fig.3.10). Lake
positions according to water level heights in all groups show significant dispersion.
Roughly average height of lakes doesn’t depend on lakes size, for all groups mean
elevation is approximately equal and is about 8.5 m (tab. 3.3). Median heights of each
group are considerably lower than the mean and almost identical and close to 3.5 m
(tab. 3.3, fig. 3.10). However, at the same time polygonal lakes (second group) show
larger variety and are generally situated a bit higher, then lakes from other groups.
3.3. Vertical section of the Lena Delta based on the lakes level heights
41
A-B
3
2
B
A
3
C-D
C
D
E-F
2
E
F
1
42
G-H
H
G
A
E
F
G
H
C
D
B
Figure 3.11. Profile of lakes elevations. Top: vertical sections, turquoise triangles – lakes position: a. A –
B the entire delta; b. C-D 3rd terrace section; c. E – F 2nd terrace section; d. G – H 1st terrace section.
Red – profile line; green – second terrace border; orange – third terrace border. Bottom: plane section
In order to display the variability of the Lena Delta topography with respect to
its geomorphological structure, a section presented in figure 3.11, which crosses all
terraces in the Lena Delta in the direction from the North–West to the South-East and
43
separately each terrace from West to East shows significant scattering of lakes
elevations over the entire area. On the other side each terrace can be distinguished on
the basis of this chart. Especially prominent on this graph is Arga Island in the northwestern part of the delta, which builds up the second terrace. Sobo-Sise Island (third
terrace) in the eastern part of the delta is also well distinguishable on the profile. The
areas in between show pretty well the general elevation decrease of lakes water in the
central part of the first terrace towards the sea shoreline at the eastern margin of the
delta.
44
4. Discussion
4.1. Applicability of remote sensing data processing and GIS methods for
creation of water objects scheme
Preprocessing of satellite images, difficulties and uncertainties
Set of high-resolution satellite images provides a new opportunity for creation a
precise map of thermokarst affected landscapes over the whole Lena Delta in
comparison with coarser Landsat images, usually used for this objective (Schneider et
al., 2009), (Kravtsova and Rodionova, 2016), (Nitze and Grosse, 2016), which can be
used for the further analysis of origin and dynamics of area of interest. This can be
explained by the relation between the spatial resolution and number of landscapes
forms. The higher the spatial resolution is, the more lakes can be observed. Therefore in
our study important factors for the satellite images selection were:
1.
Full cover of the region of interest;
2.
Display of suitable situation of landscape.
In order to cover the entire Lena Delta in this study to the main set of highresolution RapidEye images was added set of SENTINEL-2 images with coarser
resolution, which was later enhanced automatically. Thus in order to get a scheme of
water objects two sets of data with different initial resolution were compiled
consequently with different detailing (fig. 4.1).
A
B
Figure 4.1. Different accuracy of surface display. A – RapidEye dataset; B – SENTINEL-2 dataset
Within this study was used overlay of near-infrared channels from RapidEye and
SENTINEL-2 sets from July, June and beginning of September from different years.
Thus, the overlay provides generally snapshot of the maximal water expansion in the
45
end-summer situation, when the discharge is low during 2009 – 2016 years, which
helped to avoid problems with water classification due to remains of ice, pointed by
Grosse et al., (2008). As in study by Schneider et al., (2009) in this study was not
correctly considered seasonal variability of water level, which is maximal in May – June
and minimum during winter months. According to Grosse et al., (2008) seasonal
hydroclimatology can have a prominent effect on lake surface area for lakes with
shallow basin topography mostly ponds. On the other hand thermokarst lakes typically
can be determined by steep banks and a more pronounced basin morphology, which
limits lateral surface changes due to seasonal vertical level variations (Grosse et al.,
2008).
The study area has a large E-W extent and covers the 51N and 52N UTM zones.
The UTM meridian is situated in the center of the Lena Delta. Despite of that all images
in this project were orthorectificated and projected in the Universal Transverse Mercator
(UTM) projection Zone 52N because main part of images and already rectified images
GeoEye with ground control points were situated in this zone. Such decision negatively
influenced on the further morphometric calculations in the western part of the Lena
River Delta. To minimize distortion effects on morphometric calculations Morgenstern,
(2012) separated the data set along the UTM meridian into a western and an eastern part
and reprojected western part to its original UTM Zone 51N for mapping. Such action
wasn’t done in this study because of huge amount of data (obtained objects).
Preprocessing of images for further mosaicking commonly includes atmospheric
correction to lead values of pixels to Digital Numbers (DN) – real reflectance value of
surface. Nitze and Grosse, (2016) used for classification top of atmosphere reflectance
values. In this study were used ground reflectance values. In order to obtain these values
there were applied several models of this correction including removing of clouds and
haze. Due to problems of algorithm to distinguish river sandy river deposits and haze
(fig.5), and problems with recognizing of other completely different objects with similar
reflection, it was decided to use corrected images without masking. As result of
atmospheric correction were obtained different DN for the same area for RapidEye and
SENTINEL-2.
The most common method for a general land cover classification of large
heterogeneous datasets is the automatic unsupervised classification based on a chain
algorithm and the consecutive labeling of land cover classes with real land cover
features (Schneider et al., 2009). But to distinguish wetlands Schneider et al., (2009)
used supervised classification approach based on a small number of classes. Grosse et
46
al., (2008) applied a simple density slice classification to distinguish water and land in
the panchromatic imagery. To extract water bodies automatically from the Landsat-7
ETM+ image mosaic of the Lena River Delta, Morgenstern (2012) applied a grey-level
thresholding on mid-infrared band with threshold values of top-of-atmosphere
reflectance of 0 to 0.1. In this study also was tried to produce different methods of
supervised and unsupervised classification over the mosaic created on the basis of
RapidEye images in order to extract water bodies. Such classification results gave
unsatisfactory results, because of misclassification of moist or slightly flooded river
deposits -sand islands or beaches on the first terrace of the Lena River Delta resulting in
lower DN due to low water reflectance. Lakes influenced by shallow water levels
(probably less than 1 m), resulting in higher DN were misclassified due to reflectance of
the lake bottom, or due to turbid water with high sediment suspension, resulting in
higher DN from the sediment load (Grosse et al., 2008).
Thus, taking into account previous study designated that in near and midinfrared wavelengths water bodies are strong absorbers, easily distinguishable from
other land cover types (Morgenstern, 2012), overlay of near-infrared bands of images
was carried out at which each output cell value was set as a minimum value of the
values assigned to the corresponding cells in the input images. To extract water bodies
was applied a grey-level thresholding with different threshold values for SENTINEL-2
and RapidEye imageries (values of 0.1 to 2.7 and of 0.1 to 8.8 correspondingly) due to
difference in DN described above. Usually, there is a strong difference in reflectance in
near-infrared band between water bodies (low DN displayed dark) and bare or vegetated
land surface (high DN exhibited bright) (Grosse et al., 2008). Applied method has weak
points, which lead to misclassified results in case of steep north-facing cliffs or slopes
and deep thermo-erosional valleys, which was also considered by Grosse et al., (2008).
More over Schneider et al., (2009) marked that vegetation either growing or floating in
the lake can pose a challenge for any water-land distinguishing methods. They
estimated the effect of unclassified water due to vegetation on the order of <2% of the
overall water body area for some lakes. But in case of shallow or mud water
classification based on minimum values in near - infrared channel fits enough.
Filtering of obtained mosaics
Filtering of binary mosaics in order to remove noises, enhance objects and get
proper results in the further automatic raster to vector conversion is necessary (fig 4.2),
when there is no opportunity to correct resulting vector layers manually due to a huge
47
amount of objects. In order to determine the best method of filtering a short study based
on the visual evaluation of obtained results was carried out.
A
B
Figure 4.2. A - initial binary image based on RapidEye data; B - image after filtering (opening)
Main target was to select methodic of filtering, which removes noises and at the
same time keeps important forms and narrow rivers channels. Morphological filtration
was used as a basic type. The field of mathematical morphology contributes a wide
range of operators to image processing, all based around a few simple mathematical
concepts. Especially important are two operators: dilation and erosion, their join creates
correspondingly opening or closing operators. Erosion shrinks an image by stripping
away a layer of pixels from both the inner and outer boundaries of regions. The holes
and gaps between different regions become larger, and small details are eliminated.
Dilation has an opposite effect to erosion. Morphological opening of an image is
defined as the dilation of the erosion of the image. The result is the reduction of small
positive regions within the image. Morphological closing of an image is defined as the
erosion of the dilation of the image, the result of closing is opposite to opening.
Important settings are size and type of structuring element, which is used like filter
window. For targets mentioned above the best operation is opening (fig. 4.3a).
When dealing with the Lena River Delta mosaic, it becomes apparent that there
are very tiny water cluster within a single large land or vice versa. To remove such
noise and to produce a more realistic scene, we decided to use the Majority Filter
(SAGA) (fig. 4.3b), which considers all the pixels in the convolution window, and
assigns the most abundant class in this window to the central pixel (Introduction to
SAGA, 2017).
48
A
B
Figure 4.3. RapidEye initial mosaic covered by RapidEye filtered binary image. A - filter opening with
structuring element ball, r=5; B – (majority filter window 3*3 pixel)*( filter opening with structuring
element ball, r=1)
As the result of this study for the filtering was chosen method: (SAGA majority
filter window 3*3 pixel)*(OTB binary morphological filter opening with structuring
element ball with r=1) (fig 4.3).
A
B
Figure 4.4. SENTINEL-2 image covered by water mask based on RapidEye image (A) and SENTINEL-2
image (B); filtering method (SAGA majority filter window 3*3 pixel)*(OTB binary morphological filter
opening with structuring element ball with r=1)
Figure 4.4 shows significant difference between filtered mask based on the
SENTINEL-2 images (fig.4.4, a) and based on the RapidEye images (fig.4.4, b).
Reasons for this difference are able to be: a. difference of threshold values used to
separate waters for SENTINEL-2 and RapidEye sets; b. Different levels of lakes due to
different time of observations, SENTINEL-2 images were created at the beginning of
September 2016, all RaidEye images were obtained in years 2009 – 2015. Different
level of lake means relative significant changes of shallow water deep for the remote
sensing, which leads to significant changes of DN and further misclassification.
49
After filtering of initial mosaic a river channels were partly divided in the
narrow places. Thus were received additional lakes, which in fact are rivers, drainage
channels, small streams or oxbows, but not lakes. Morgenstern in her study (2012)
manually removed all such water pixels. In this work we also tried to correct resulted
water scheme manually.
Water elevation extraction
For the analysis of lakes level heights in this study we needed to get value of
averaged shoreline elevation for each water body. DEM presented water as very rough
surface sometimes with significant height differences. To minimize these differences
and to smooth initial DEM, water mask was corrected using different methods including
interpolation of elevation values from polygon edges (shoreline). Obviously that such
method supposed mistakes in case when polygon of lake situated incorrect. In some
places such mistakes were manually fixed. Mean standard deviation of elevation value
for lakes with raw (initial) DEM surface is about 0.22 and for lakes with smoothed
surface is 0.15. Maximal standard deviation for smoothed lakes is =3.29 m, for nonsmoothed 4.41 m, minimum STD for both is 0.
To estimate accuracy of obtained mean elevations for each lake they were
compared with 77 elevations of significant lakes in various parts of study area, extracted
from topographic map with terrain conditions on 1981 and scale: 1: 200 000. Mean
Square Error (MSE) of difference between these two rows of data 7.6 m. MSE of
difference between elevations from topographic map and mean elevation value from
DEM minus STD of mean value for each lake is 4.5. Thus in this study was decided to
use as elevations difference between mean height value for each lake and STD. This
approach corresponds to the understanding that water level generally fits to the lower
shoreline of water body. On the other side, hence thermokarst landscapes in the Lena
River Delta are volatile, comparison of modern results with data obtained 30 years ago
can bring mistakes.
Comparison of heights data calculated from DEM with altitude obtained by laser
altimetry of GLAS/ICESat showed a correlation between two datasets with a coefficient
near to 1. Thus convergence of elevation results confirms the possibility to use
combination of edited DEM and GLAS laser altimetry data for the determination of
inland water-edge.
One more weak point is that despite of corrections a lot of water bodies, derived
elevation under 0. Majority of such lakes is situated near to the sea or in the river
50
channels valleys. Probably these values are caused by inaccuracy of used DEM water
mask, especially for fluid water.
4.2. Discussion of obtained results
Entire study area
The Lena Delta as other Arctic deltas is characterized by an abundance of ponds
and lakes. These ponds and lakes display highly varied sizes, shapes, depths, elevations
and methods of formation (Walker H., 1998). Kravtsova and Bystrova (2009) marked
that the region of the Lena River Delta is described relatively to the major part of MidEast Siberia, by a concentration of lakes network and an increase in the size of lakes.
They are present in a variety of deltaic environments including old river
channels (e.g., oxbow lakes), terrace-flank depressions, thaw depressions, inter- and
intra-dune depressions, swales in ridge and swale deposits, low-centered polygons and
the troughs between polygons.
An area of the delta covered by lakes obtained in this study is a bit less than 12%
which proved previous study. The limnicity between terraces and distribution of lakes
differs until several times, which can be explained by strong connection of hydrological
and geomorphological factors.
Characteristic of limnicity, and lakes distribution is complicated due to different
minimal reference areas of lakes, accepted for calculations. Comparison of results can’t
be direct because number of detected lakes and ponds depends on the spatial resolution
of initial data. By far the most common type of water bodies in the study region is
polygonal ponds.
According to Morgenstern (2008) the total area of lakes ≥ 20 ha is 1,861.8 km²,
or 6.4 % of the delta area, as the delta area assumed to be 29,000 km² (Schneider et al.,
2009). Schnieder et al., (2009) relying on Landsat -7 imagery classification determined
a surface covered by lakes with lakes area more than 0.36 ha – 3008 km², or 33.8% of
the Lena River Delta land area. Boike et al., (2013) found that the limnicity of Samoilov
island is about – 15%. Results of this study show that about 189,000 ponds and lakes
cover a bit higher area in comparison with previous studies – about 3,390 km² or 11.7%
of the entire Lena River Delta. In comparison with another Arctic deltas, for example
the Mackenzie River Delta, where lakes occupy about 25% of entire delta (Walker,
1998), or the Kolyma Delta tundra zone where limnicity is found to be 13.5 %
(Veremeeva and Glushkova, 2016), limnicity of the Lena River Delta isn’t high.
51
Bolshiyanov et al., (2013) relying on the map 1:100 000 found 29,483 lakes
(which is more than in 6 times less in comparison with our study), including 27165
lakes with an area under 0.25 km², 0.25 till 1 km² – 1817, from 1 km² till 5 km²– 443,
from 5 till 10 km²– 48, 10 lakes with an area more than 10 km². Our study shows that
the number of lakes with an area less than 0.25 km² is in almost in seven times higher.
But in the other size classes was received slightly smaller numbers. Significant
difference in the number of small lakes and ponds can be explained by difference in
resolution of initial data. Differences in larger lakes can be explained by descent of
lakes for the last 30 years. Obtained results also shows that the limnicity of Arga island
is highest all over the delta, which correspond with data of lakes distribution presented
with Bolshiyanov et al., (2013).
Mean lake area in the delta is about 18,000 m². This is under previous
estimations (Morgenstern et al., 2008) due to difference in spatial resolution. Relatively
great lakes, mean area is larger than 20,000 m² are situated mostly on the second terrace
in the eastern part of the delta, probably due to antiquity of this part (Bolshiyanov et al.,
2013).
According visual estimations majority of present thermokarst lakes in the
depressions all over the delta are apparently smaller than the depressions, which was
also marked by Grosse et al., (2005). Such inequality of area is most prominent on the
Bykovsky peninsula and on the third terrace and caused by constant changes of
thermokarst. Shrinking of taw lakes in the permafrost affected regions of northern
Russia caused mainly by river drainage and vegetating (Kravtsova and Rodionova,
2016).
The Lena River Delta is generally lowland because the first terrace occupies the
biggest area and this is displayed in the mean height of lakes level, which is only 5.9 m
a.s.l. But differences in structure and origin of various parts of the delta are reflected in
the broad heights distribution of water bodies elevation from 0 until 65 m a.s.l. and
significant standard deviation of elevation series, more than 5 m. In light of the above it
is difficult to compare the Lena River Delta with other arctic deltas like the Colville or
the Mackenzie River deltas in sense of lakes sizes and elevation.
1st terrace
The first terrace is the youngest from main terraces in the Lena River Delta. An
age of the central part of the terrace is about 8 thousand years and decreases towards
52
shore line until several hundreds of years (Bolshiyanov et al., 2013). Thus the structure
of this terrace displays modern and Holocene fluvial dynamics.
Presented in this study section shows the significant difference in elevation
between the central part and coastal parts of this terrace. Elevation of lakes shorelines
levels in the central part are basically between 6 and 8 m a.s.l., in comparison with 0 – 2
m a.s.l. near to the shore of Laptev Sea. This coincides with data presented by
Bolshiyanov et al., (2013): elevation of terrace in the middle part is about 10 - 12 m
a.s.l., which falls on 10 -12 m towards the coastal margins. Such inequality in elevation
of one terrace is caused by significant difference in the age of areas. The mean elevation
of lake shorelines on this terrace is closer to elevation of margins, where frequency of
water bodies id higher than in central part.
Morgenstern et al., (2008) found that there are 1,796 lakes with an area more
than 20 ha, which cover 997 km² on the 1st terrace, and limnicity of terrace is 6.3%
correspondingly. In this study was determined in 17 times more water objects, which
cover an area more than in two times larger comparing to previous studies. Such
considerable difference, relatively to this study, origins from taking into account not
only thermokarst lakes but also ponds and polygonal lakes, which are frequent in this
part of the delta.
Thus lakes on the first terrace are on average small with low elevation. This
correlates well with genetic types of lakes typical for such environment: polygonal
ponds and lakes, small circular thermokarst lakes and abandoned lakes. Abandoned
lakes resulting from channel braiding and from meandering (oxbow lakes) are
especially typical for this terrace, which is proved by the mode of elevation and
concentration of heights of 75% of lakes near to sea level.
2nd terrace
The 2nd terrace is intermediate in the Lena River Delta by area and elevation,
maximal heights of surface reach 30 m a.s.l. Elevation differences of surface is until 10
m (Bolshiyanov et al., 2013). The surface is characterized by thermokarst lakes, alases.
Currently there are some difficulties in understanding of the genesis of large water
bodies on the second terrace. It is only clear that the origin is connected with ice-wedges
thaw (Bolshiyanov et al., 2013).
Limnicity especially of Island Arga, which is the main part of terrace, is highest
(>17%) in the Lena River Delta. More than 52,000 lakes cover an area of 1,052 km².
Mean elevation of lakes is about 10 m a.s.l., which also points on abundance of
53
thermokarst basins forming since early Holocene. Important feature of the second
terrace is size of lakes. On this terrace large lakes predominate. This is caused by the
conjunction of lakes through termoerosion valleys or union of lakes, but the lakes basins
preserve the individuality through the deep central parts, which can be distinguished on
satellite images or DEM.
3rd terrace
The 3rd terrace presented as individual islands formed from sandy sequences
covered by “Ice Complex” is the smallest but the highest from the main
geomorphological units in the Lena River Delta. An elevation of the relic terrace varies
between 20 and 60 m a.s.l., and average value is about 35-40 m a.s.l. (Bolshiyanov et
al., 2013). The terrace can be subdivided into two areas due to more than 20 m
difference in altitude between the western and the eastern sector (Grigoriev, 1993). This
difference is not so large but noticeable in the lakes elevation. Average elevation of
lakes shorelines on this terrace is 20 m a.s.l. and it coincides with bottom estimation of
terrace elevation. Moreover the distribution of lakes by elevation shows larger
dispersion than other terraces. It can be caused by thermokarst lakes which have large
weight among the lakes and are situated in thermokarst basins with various depths. The
highest lakes all over the delta are situated in the centre of Kharadang Island with an
elevation more than 60 m a.s.l.
Morgenstern (2012) detected 2,327 water bodies (minimum one pixel, 900 m²)
with a total area of 88.3 km² considering 98.6% of the third terrace area (1,711.6 km²).
Due to difference in resolution in our study were obtained much more lakes (4,010 ≥
100 m²) with a total area of 97.86 km² which is 5.7% of a 3 rd terrace area. It is obvious,
that results coincide with each other considering absence of small ponds in the study of
Morgenstern (2012).
Lakes coverage of 5.7% is the smallest among terraces in the Lena River Delta
as well it is low compared to other arctic tundra regions like the western arctic coastal
plain of Alaska with about 20 % lake coverage Morgenstern (2012).
The thermokarst basins mostly significantly exceed thermokarst lakes situated
within them, which can be an evidence of recent lakes drainage on the 3rd terrace.
Moderate changes of permafrost in the Lena River Delta for the last 30 years are
connected with descend of small thermokarst lakes on the third terrace (Kravtsova and
Bystrova, 2009). Generally characteristics of the lakes on the third main terrace are
54
typical for thermokarst lakes in ice-rich permafrost and controlled by the thermokarst
process.
Bykovsky Peninsula study area
In our study was considered not only Bykovsky Peninsula but also adjacent
lower part of Khorogor Valley. Bykovsky Peninsula is a Pleistocene accumulation
plain. Therefore the relief of the peninsula is dominated by flat elevated areas up to 40
m a.s.l., maximum elevation is 43 m a.s.l. The area of the Bykovsky Peninsula occupied
by lakes is less than the area occupied by thermokarst depressions according Grosse et
al., (2005) in three times. This shows probable decreasing of lakes since the Late
Pleistocene Holocene transition. A lot of modern lakes appeared during the Holocene in
the old drained basins. The thermokarst depressions according to Grosse et al., (2007)
mostly have steep slopes and very low mean elevations of 1–8 m a.s.l. and cover 46% of
the entire area.
Grosse et al., (2008) calculated that lake cover by land area at BYK is 15.4%, or
2,053 ha are covered by 32 lakes with an area > 10 ha. In this study were detected 345
lakes with an area > 225 m², which cover 16.41 km², or 9.5% of BYK area. In the
combined region of lower of Khorogor Valley and BYK Grosse et al., (2005) mapped
569 water bodies with an area > 186 m². Such number is compatible with results
obtained in our study - 526 lakes, including only the neighbor to BYK part of Khorogor
Valley.
Mean elevation of water bodies above 10 m a.s.l. can be explained due to
combination of lakes situated on flat elevated areas, with maximal lakes height about 40
m a.s.l., and lakes situated in the deep thermokarst basins with lakes elevations less than
5 m a.s.l. up to 0 m a.s.l. For example, the largest lake on the BYK with an area about 6
km² situated in the vast thermokarst basin is elevated only on 0.5 m a.s.l. This difference
is displayed in STD (Standard Deviation) of lakes elevation, which equals to 12.6 m, the
highest dispersion among all studied areas.
55
5. Conclusions
This study is aimed at comprehensive mapping of the Lena Delta region water
body and lake level height mapping based on remote sensing data and GIS methods.
Ponds, polygonal and thermokarst lakes are a major component of vast arctic and
subarctic landscapes in Siberia. Changes in the extent, number and elevation of these
water objects, which are typical for permafrost affected territories, can be viewed as
critical indicators of landscapes response to climate changes. Spatial altitudinal analysis
of water features plays an important role in understanding thermokarst variability in
polar region and impacts on the global hydrological and chemical cycles.
The Lena Delta is the largest arctic delta and its complicated structure,
conditioned by the modern deltaic processes, presented in the first terrace, and various
types of past dynamics reflected by the second and the third terrace correspondingly.
Such diversity underlines the Lena Delta among other arctic deltas. Thus the Lena Delta
and adjacent Bykovsky Peninsula is a crucial territory for studying of subarctic
periglacial ecosystem changes, caused by permafrost variability foremost expressed in
the thermokarst and thermal erosion.
Applied methodology, based on the Remote Sensing data, GIS handling and
statistical calculations allows to map and to determine an elevation of water bodies with
high accuracy. Using of two sets of high-resolution images acquired by two satellite
surveying systems RapidEye and SENTINEL-2, was created two compositions of
orthorectified and atmospherically corrected images, which jointly covers entire study
region. Thus, taking into account previous study designated that in near and midinfrared wavelengths water bodies are easily distinguishable from other land cover
types, overlay of near-infrared bands of images was carried out at which each output
cell value was set as a minimum value of the values assigned to the corresponding cells
in the input images. For further extraction of water bodies was applied a common
method of grey-level thresholding for mosaic separately. Comprehensive scheme of
water bodies was created after filtering and refinement of overlay results. To determine
elevation values of acquired lakes as initial data were used heights extracted from
improved water mask of DEM TanDEM-X. Received results were checked and
compared with laser altimetry data collected by ICESat observing mission. The
comparison of two elevation sets showed minor discrepancies. Thus used in this study
methodology shows its applicability for mapping and further analysis.
56
In addition during the execution of this thesis were carried out several practice
investigations aimed at determination of the suitable methods of morphological filtering
and refinement of DEM water mask.
As a result, 198851 water features with a size of more than 100 m² were mapped
the first time in the region of the whole Lena Delta and adjacent Bykovsky Peninsula.
For 189497 ponds and lakes that cover about 3426 km² were determined heights of
water edge. Obtained spatial and elevation data fairly accurate display the division of
the delta on three main terraces. Estimation of results shows a significant scattering of
lakes characteristics and heights, as well as differences of limnicity between main
geomorphological terraces. A mean elevation of lakes changes from 20 m a.s.l. for lakes
situated on the 3rd terrace to 4.5 m a.s.l. for lakes on the first terrace. Profiles show
general decline trend towards the eastern, modern margin of the delta. Thus data
mentioned above correspond well with previous studies connected with analysis of
lakes spatial features and distribution as well as terraces heights values assessment.
However worth saying that using of high resolution data allowed us to newly estimate
the number of small ponds and their influence on statistical features of water bodies in
the region of interests, in comparison with previous studies.
Moreover this study opens a lot of opportunities for further continuation of
research, especially connected with the determination of river channel inclination. This
objective hasn’t been achieved in presented work. It is planned in near feature to receive
inclination of water channels, because main preparation steps, including a mapping of
river channels for the whole delta have been already done.
57
6. Acknowledgements
I would like to express my gratitude to my supervisor from the German side Dr.
Frank Günther for the continuous support of my master thesis and related research, for
his patience, perfect knowledge in fields of remote sensing and permafrost. His
guidance helped me in all the time of research and writing of my master thesis.
I would like to express my sincere gratitude to my supervisor from the Russian
side Dr. Irina Fedorova for the remarks, helpful comments and his participating in the
discussion of this master thesis.
My sincere thanks goes to The Alfred Wegener Institute (Potsdam) and PETACARB project team, especially to Prof. Dr. Guido Grosse. Without his immense
support, it would not have been possible to conduct this study.
Furthermore, I would like to thank whole staff of Master Program for Polar and
Marine Sciences (POMOR) in particular Dr. Nadezhda Kakhro, Dr. Heidemarie
Kassens and Prof. Dr. Eva-Maria Pfeiffer for their permanent and comprehensive
support.
58
7. References
Alekseevsky, N., 2007. Geoecological state of Russian Arctic coast and their safety of
nature management, GEOS Publ.,Moscow, Russia, 586 p (in Russian).
AMAP, 2011. Snow, Water, Ice and Permafrost in the Arctic (SWIPA): Climate
Change and the Cryosphere.
Antonova S., Duguay Claude R; Kääb A.; Heim B.; Langer M.; Westermann S.; Boike
J., 2016. Monitoring bedfast ice and ice phenology in lakes of the Lena River
Delta using TerraSAR-X backscatter and coherence time series. Remote Sensing,
8(11): 903.
Aret F., Reimnitz E., 2000. An Overview of the Lena River Delta Setting: Geology,
Tectonics, Geomorphology and Hydrology. Journal of Coastal Research, Vol. 16,
4: 1083-1093.
Behling, R., Roessner, S., Kaufmann, H. & Kleinschmit, B., 2014. Automated
Spatiotemporal Landslide Mapping over Large Areas Using RapidEye Time
Series Data. Remote Sensing, 6: 8026-8055.
Bernardo N., Watanabe F., Rodrigues T., Alcantara E., 2017. Atmospheric correction
issues for retrieving total suspended matter concentrations in inland waters using
OLI/Landsat-8 image Advances in Space Research Volume 59, Issue 9: 2335–
2348.
Boike, J., Wille, C., Abnizova, A., 2008. Climatology and summer energy and water
balance of polygonal tundra in the Lena River Delta, Siberia Journal of
Geophysical Research: 113.
Boike, J., Kattenstroth, B., Abramova, K., Bornemann, N., Chetverova, A., Fedorova, I.,
Fröb, K., Grigoriev, M., Grüber, M., Kutzbach, L., Langer, M., Minke, M.,
Muster, S., Piel, K., Pfeiffer, E.-M., Stoof, G., Westermann, S., Wischnewski, K.,
Wille, C., and Hubberten, H.-W., 2013. Baseline characteristics of climate,
permafrost and land cover from a new permafrost observatory in the Lena River
Delta, Siberia (1998–2011), Biogeosciences, 10: 2105–2128.
Bolshiyanov, D., Makarov, A., Schneider V. and Stof G., 2013. Evolution and
development of the Lena River delta, Publ. of AARI, St. Petersburg, Russia, 268 p
(in Russian).
Drachev, S. S., Savostin, L. A., Groshev, V. G., Bruni, I. E., 1998. Structure and
geology of the continental shelf of the Laptev Sea. Eastern Russian Arctic
Tectonophysics, 298: 357 – 393.
59
Grigoriev, N.F., 1960. The temperature of permafrost in the Lena Delta basin: Deposit
conditions and properties of the permafrost in Yakutia, 2: 97–101 (in Russian).
Grigoriev, M.N., 1993. Criomorphogenesis in the Lena Delta. Permafrost Institute
Press, Yakutsk, 176 (in Russian).
Grigoriev, M., Rachold, V., Schirrmeister, L., Hubberten, H.-W. Stein, R., Macdonald,
R., 2004. The organic carbon cycle in the Arctic Ocean: present and past,
Springer. Berlin, 37-65.
Grosse, G., Schirrmeister, L., Kunitsky, V.V., Hubberten, H.-W., 2005.The use of
CORONA images in remote sensing of periglacial geomorphology: an illustration
from the NE Siberian coast. Permafrost and Periglacial Processes 16: 163–172.
Grosse, G., Romanovsky, V., Walter, K., Morgenstern, A., Lantuit, H., Zimov, S., 2008.
Distribution of thermokarst lakes and ponds at three Yedoma sites in Siberia.
Proceedings of the 9th International Conference on Permafrost, Fairbanks, Alaska,
29 June - 3 July 2008, Kane, D. L. and Hinkel, K. M. (Eds.), Institute of Northern
Engineering, University of Alaska Fairbanks: 551-556.
Grosse, G.; Romanovsky, V.; Jorgenson, T.; Walter Anthony, K.; Brown, J. &
Overduin, P., 2011. Vulnerability and feedbacks of permafrost to climate change
Eos Trans. AGU, AGU, 92: 73-74.
Grosse G., Jones B., and Arp C., 2013. Thermokarst Lakes, Drainage, and Drained
Basins. In: John F. Shroder (Editor-inchief), Giardino, R., and Harbor, J. (Volume
Editors). Treatise on Geomorphology, Vol 8, Glacial and Periglacial
Geomorphology, San Diego: Academic Press: 325-353
Günther, F., Overduin, P. P., Sandakov, A. V., Grosse, G., Grigoriev, M. N., 2013.
Short- and long-term thermo-erosion of ice-rich permafrost coasts in the Laptev
Sea region. Biogeosciences, 10: 4297-4318.
Günther, F., Overduin, P. P., Yakshina, I. A., Opel, T., Baranskaya, A. V. & Grigoriev,
M. N., 2015. Observing Muostakh disappear: permafrost thaw subsidence and
erosion of a ground-ice-rich island in response to arctic summer warming and sea
ice reduction. The Cryosphere, 9: 151-178.
Jorgenson, M. T., and Grosse, G., 2016. Remote Sensing of Landscape Change in
Permafrost Regions. Permafrost and Periglacial Processes, 27: 324–338.
Kravtsova, V. I., Bystrova, A. G., 2009. Changes in thermokarst lake sizes in different
regions of Russia for the last 30 years. Kriosfera Zemli (Earth Cryosphere) 13: 1626. (in Russian).
60
Kravtsova, V. I., Tarasenko, T. V., 2011. The dynamics of thermokarst lakes under
climate changes since 1950, Central Yakutia. Kriosphera Zemli (Earth
Cryosphere) 15 (3): 31-42. (in Russian).
Kravtsova, V. I., Rodionova, T.V., 2016. Variations in size and number of thermokarst
lakes in different permafrost regions: spaceborne evidence. Kriosfera Zemli (Earth
Cryosphere) 20: 75-81.
Kokelj, S. V. and Jorgenson, M. T., 2013. Advances in Thermokarst Research.
Permafrost and Periglac. Process., 24: 108–119.
Kuhry, P., Grosse, G., Harden, J. W., Hugelius, G., Koven, C. D., Ping, C.-L.,
Schirrmeister, L. and Tarnocai, C., 2013. Characterisation of the Permafrost
Carbon Pool. Permafrost and Periglac. Process., 24: 146–155.
Morgenstern, A., Grosse, G., Schirrmeister, L., 2008. Genetic, morphological, and
statistical characterization of lakes in the permafrost-dominated Lena Delta, in:
Proceedings of the 9th International Conference on Permafrost, Fairbanks, Alaska,
29 June - 3 July 2008, Kane, D. L. and Hinkel, K. M. (Eds.), Institute of Northern
Engineering, University of Alaska Fairbanks: 1239-1244.
Morgenstern A.,Grosse G., Gunther F., Fedorova I., and Schirrmeister L., 2011 Spatial
analyses of thermokarst lakes and basins in Yedoma landscapes of the Lena Delta.
The Cryosphere Discuss., 5: 1495–1545.
Morgenstern, A., 2012 .Thermokarst and thermal erosion: Degradation of Siberian icerich permafrost. PhD thesis, Univ.Potsdam, 116 p.
Morgenstern A., Ulrich M., Günther F., Roessler S., Fedorova I., Rudaya N., Wetterich
S., Boike J., Schirrmeister L., 2013. Evolution of thermokarst in East Siberian icerich permafrost: A case study. Geomorphology, 201: 363–379.
Muster S., Langer M., Heim B., Westermann S., Boike J., 2012. Subpixel heterogeneity
of ice-wedge polygonal tundra: a multi-scale analysis of land cover and
evapotranspiration in the Lena River Delta, Siberia. Tellus B, 64.
Muster, S., Heim, B., Abnizova, A., Boike, J., 2013. Water Body Distributions Across
Scales: A Remote Sensing Based Comparison of Three Arctic Tundra Wetlands.
Remote Sensing, 5: 1498-1523.
Nitze I., Grosse G., 2016. Detection of landscape dynamics in the Arctic Lena Delta
with temporally dense Landsat time-series stacks Remote Sens. Environ., 181:
27–41.
PCI, 2015: ATCOR- Ground Reflectance Workflow Geomatica 2015 Tutorial, 11 p.
61
Proshutinsky, A., Ashik, I., Dvorkin, E., Häkkinen, S., Krishfield, R., Peltier, W., 2004.
Secular sea level change in the Russian sector of the Arctic Ocean. Journal of
Geophysical Research: Oceans: 109.
Rachold, V., Grigoriev, M.; Are, F.; Solomon, S.; Reimnitz, E.; Kassens, H. &
Antonow, M., 2000. Coastal erosion vs riverine sediment discharge in the Arctic
Shelf seas, International Journal of Earth Sciences, Springer-Verlag, 89: 450-460.
Schneider J., Grosse G., Wagner D., 2009. Detection of landscape dynamics in the
Arctic Lena Delta with temporally dense Landsat time-series stacks. Remote
Sensing of Environment 113: 380–391.
Schwamborn, G., Rachold, V., Grigoriev, M., 2002. Late Quaternary sedimentation
history of the Lena Delta Quaternary International, 89: 119-134.
Schirrmeister, L., Grosse, G., Schwamborn, G., Andreev, A., Meyer, H., Kunitzky, V.,
Kuznetsova, T., Dorozhkina, M., Pavlova, E., Bobrov, A., Oezen, D., 2003. Late
Quarternary history of the accumulation plain north of the Chekanovsky Ridge
(Lena Delta, Russia). A multidisciplinary approach Polar Geography, 27: 277319.
Schirrmeister, L., Grosse, G., Schnelle, M., Fuchs, M., Krbetschek, M., Ulrich, M.,
Kunitsky, V., Grigoriev, M., Andreev, A., Kienast, F., Meyer, H., Babiy, O.,
Klimova, I., Bobrov, A., Wetterich, S., Schwamborn, G., 2011. Late Quaternary
paleoenvironmental records from the western Lena Delta, Arctic Siberia
Palaeogeography, Palaeoclimatology, Palaeoecology, 299: 175-196.
SENTINEL-2 User handbook, 2015: ESA Standard Document, Issue 1 Rev 2, 64p.
TanDEM-X, 2016: Ground Segment DEM Products Specification Document, issue 3.1,
EOC – Earth Observation Center, 46 p.
Ulrich, M., Grosse, G., Chabrillat, S. and Schirrmeister, L., 2009. Spectral
characterization of periglacial surfaces and geomorphological units in the Arctic
Lena Delta using field spectrometry and remote sensing. Remote Sensing of
Environment, 113: 1220-1235.
van Everdingen, R. O. (Ed.), 2005. Multi-language glossary of permafrost and related
groundice terms, National Snow and Ice Data Center/World Data Center for
Glaciology, Boulder. Available at: http://nsidc.org/fgdc/glossary.
Vaudoura E., Gilliota J.M., Belc L., Bréchetb L., Hamiachec J., Hadjarb D.,
Lemonnierb Y., 2014. Uncertainty of soil reflectance retrieval from SPOT and
RapidEye multispectral satellite images using a per-pixel bootstrapped empirical
62
line atmospheric correction over an agricultural region. International Journal of
Applied Earth Observation and Geoinformation, 26: 217–234.
Veremeeva, A.A., Glushkova N.V., 2016. Formation of relief in the regions of ice
complex deposits distribution: remote sensing and GIS studies in the Kolyma
lowland tundra. Kriosfera Zemli (Earth Cryosphere), 20 (1): 14–24
Walker, H.J. 1998. Arctic deltas. Journal of Coastal Research 14: 718-738.
Zhang T., Barry R.G., Knowles K., Heginbottom J.A., Brown J., 2008. Statistics and
characteristics of permafrost and ground-ice distribution in the Northern
Hemisphere. Polar Geogr., 31: 47–68.
Zubrzycki, S.; Kutzbach, L.; Grosse, G.; Desyatkin, A. , Pfeiffer, E.-M., 2013. Organic
carbon and total nitrogen stocks in soils of the Lena River Delta, Biogeosciences,
10: 3507-3524.
Internet sources:
Awi.de, 2016. Alfred Wegener Institute PETA-CARB. [online] Available at:
https://www.awi.de/en/science/junior-groups/peta-carb
GRASS
GIS
7.3.svn,
2017
Reference
Manual.
[online]
Available
at:
https://grass.osgeo.org/grass73/manuals/r.series.
eoportal.org,
2017
[online]
Available
at:
https://directory.eoportal.org/web/eoportal/satellite-missions/r/rapideye.
gis-lab.info, 2017. Orthocorrection of space images using RPC. [online] Available at:
http://gis-lab.info/qa/ortho-rpc.html.
GRASS
GIS
7.3.svn,
2017.
Reference
Manual.
[online]
Available
at:
https://grass.osgeo.org/grass73/manuals/r.series.
dst-iget.in,
2017
Introduction
to
SAGA.
[online]
Available
at:
http://dst-
iget.in/tutorials/iget-rs/iget-rs-001
63
Statement on the thesis’ originality
Herewith I, Aleksandr Volynetc, declare that I wrote the thesis independently and did
not use any other resources than those named in the bibliography, and, in particular, did
not use any Internet resources except for those named in the bibliography. The master
thesis has not been used previously as part of an examination. The master thesis has not
been previously published.
64
Отзывы:
Авторизуйтесь, чтобы оставить отзыв