https://doi.org/10.1007/s12665-020-08996-3 ORIGINAL ARTICLE
Detrimental environmental impact of large scale land use
through deforestation and deterioration of carbon balance in Istanbul
Northern Forest Area
Ahmet Ozgur Dogru1 · Cigdem Goksel1 · Ruusa Magano David1 · Doganay Tolunay2 · Seval Sözen3,4 ·
Derin Orhon4,5
Received: 3 November 2019 / Accepted: 10 May 2020 / Published online: 30 May 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract
This study explored the environmental impact of the large-scale projects, the 3rd Bridge across the Bosphorus and 3rd Airport, carried out in the last decade leading to the massive deterioration of the northern forest area in Istanbul. Destroyed forest area was assessed through relevant changes in land classification detected by multi-temporal Landsat data of Istanbul between 2009 and 2016. The magnitude of destroyed carbon stocks and related CO2 emission together with the reduction
in the CO2 absorption potential inflicted by massive land-use change were also calculated. Observed results indicated that approximately 15,000 ha of forest area was destroyed in 7 years, corresponding to a 7% ultimate loss in the total forest area. The total land cover change for the same period was determined as 11.5% of the study area. The extent of land cover changes indicated that more than 4.4 million tons of CO2 were additionally emitted to the atmosphere, due to observed reduction
of carbon stocks between 2009 and 2016. More than 70% of the total C/CO2 emission associated with land cover changes could be attributed to the loss of forest land. In addition, destroyed forestland corresponded to a CO2 absorption loss of 0.3 million tons CO2/year equivalent to the emission of 830,000 people in Istanbul.
Keywords Environmental impact · Sustainable ecosystem · Deforestation · Carbon emission · Carbon balance · Land use-land cover change
Electronic supplementary material The online version of this
article (https ://doi.org/10.1007/s1266 5-020-08996 -3) contains
supplementary material, which is available to authorized users. * Derin Orhon
orhon@itu.edu.tr Ahmet Ozgur Dogru ozgur.dogru@itu.edu.tr Cigdem Goksel goksel@itu.edu.tr Ruusa Magano David maganodavid2@gmail.com Doganay Tolunay dtolunay@istanbul.edu.tr Seval Sözen
sozens@itu.edu.tr
1 Geomatics Engineering Department, Faculty of Civil
Engineering, Istanbul Technical University, Maslak, 34469 Istanbul, Turkey
2 Department of Soil Science and Ecology, Faculty of Forestry,
Istanbul University - Cerrahpasa, Bahcekoy, 34473 Istanbul, Turkey
3 Environmental Engineering Department, Faculty of Civil
Engineering, Istanbul Technical University, Maslak, 34469 Istanbul, Turkey
4 ENVIS Energy and Environmental Systems Research
and Development Ltd, ITU ARI Technocity, Maslak, 34469 Istanbul, Turkey
5 Faculty of Civil and Environmental Engineering, Department
of Environmental Engineering, Near East University, Nicosia, 99138 Mersin 10, Turkey
Introduction
Spreading across two continents, Europe and Asia, Istanbul is a unique historical landmark and a major cosmopolitan with a population over 15 million. In the last 50–60 years, Istanbul Metropolitan area has evolved into the major center of industrial activities, services, manufacturing along with education and culture. The city currently covers an area of 5340 km2, which represents less than 1% of the land area in
Turkey; however, it accommodates nearly 19% of the coun-try’s population. Since 1950, Istanbul has been receiving sizable migration, boosting its population from 1 million in 1950 to over 15 million in 2018 (TUIK 2020).
The existing geomorphologic characteristics significantly define two ecologically different natural regions within the Metropolitan area of Istanbul. The first one is the southern part of the Metropolitan area involving the bulk of the urban activities, where the climate is relatively warm and dry. The second region covers the northern part of the Metropolitan area, which is mainly wet, humid and under the effect of northerly winds in winter. This region is commonly known as “Northern Forests” and literally constitute “the lungs” of the city. The natural vegetation of this region largely con-sists of forests of broad-leaved trees dominated by oaks-hornbeam and beech. The beech forests are well developed on the north-facing slopes (IBB 2009). The wet woodlands are abounding along river valleys; flood plains are mainly dominated by ash, lime, alder, poplar, willow, etc. (Gürel and Gündüz 2011). The total area of woodlands and for-ests in the metropolitan area is estimated around 2500 km2.
In this framework, the linear urban development of Istan-bul relies on a fragile ecosystem structure, including for-est areas, water basins and agricultural lands in the north and ever-increasing human activities in the southern belt. This fragile ecological balance should be protected to secure Istanbul’s sustainability (Orhon 2014a). Unfortunately, mega construction projects implemented in the Northern Forest zone inflicted a significant damage to the ecological balance of the region.
Istanbul, like all similar cities worldwide, requires reli-able information to assess the impact of rapid urbanization and land use on a sustainable environment (Griffiths et al.
2010; Taubenböck et al. 2012). Satellite remote sensing techniques have offered an efficient and timely approach to the mapping and collection of basic land use and land cover (LULC) data over large areas and now have become one of the most effective tools for LULC change detection (Mayomi
2009). Temporal remote sensing data can be used for pro-ducing change map and examine LULC changes (Zhu and Woodcock 2014; Kharazmi et al. 2018).
The determination of LULC change is particularly effec-tive for the identification of the natural and artificial impacts on the environment. Numerous studies examined the poten-tial of satellite remote sensing to obtain accurate and timely geospatial information describing changes in LULC of urban environments (Bauer et al. 2003; Dogru et al. 2006; Zhu and Woodcock 2014; Kharazmi et al. 2018). In this context, the change of carbon stocks can be determined and analyzed due to the change in land cover (Wang et al. 2014; Al-Reasi et al. 2018).
Environmental impacts of the natural or artificial LULC changes occurred in Istanbul have been studied by several researchers for the last decade. In this context, Sanli et al. (2008) defined temporal-spatial patterns in Istanbul for intro-ducing the impacts of increasing population on the environ-ment in Istanbul by determining the LULC changes through remotely sensed images. Their results highlighting the rapid increase in settlement areas and decrease in forest area were also confirmed by Goksel et al. (2018) with their study exe-cuted in Beykoz and Sariyer, which are two districts located at the Asian and European side of northern. Additionally, Dogan and Stupar (2017) analyzed the dynamics of urban transforma-tion using three mega projects planned in Istanbul as a case study. This study mainly focused on the three estimated impact areas, which are the urban structure, environment/ecology and community. Besides, the study gave an idea of the contextual background of the projects and their implementation mecha-nisms (Dogan and Stupar, 2017). In another study conducted by Ayazli et al. (2015), potential urban growth to be observed as a result of the 3rd Bridge project was examined based on the LULC through a cellular automata (CA) based simulation. The study resulted with an urban growth simulation model for 2030, highlighting a continuous growth through northern Istanbul affecting 41% of forested areas and 28% of fragile ecosystem areas (Ayazli et al. 2015). Akin et al. (2015) also confirmed and improved this result by estimating the LULC in 2040. Their study resulted in a rapid urban expansion in North-ern Istanbul in the period of 2013-2040, which will mostly influence water basins and forest areas (Akin et al. 2015).
In this study, satellite remote sensing technology has been used as an accurate and reliable data acquisition tool to determine the changes in LULC characteristics of Istan-bul. The study essentially explored the detrimental impact of the large-scale projects carried out in the last decade, leading to the deterioration of Northern Forest zone in Istan-bul. It was basically implemented by assessing changes in land classification detected by multi-temporal Landsat ETM + (Enhanced Thematic Mapper) and OLI (Operational Land Imager) data of Istanbul Metropolitan Area between 2009 and 2016. In this context, (i) it evaluated the mag-nitude of destroyed forest and green area through detected LULC changes during the study period; (ii) it quantified the negative impact of the large- scale projects in terms of
carbon loss by estimating the reduction in the carbon dioxide absorption potential inflicted by massive land use.
Study area
Characteristics of the area
In 2002, the municipality started comprehensive studies for
a strategic plan for Istanbul. Natural, economic and social
sustainability was the prime objective of this plan, which was unanimously approved in 2009. The plan defined a population range of 16–17 million as the upper limit for the optimal size of Istanbul, especially to prevent the inva-sion of forest zones and water resources basins for urban expansion. Additionally, urban development restricted to
west–east direction and preservation of natural areas and
ecological resources basically focused on restricting spatial development to protect forest areas, water basins and agri-cultural land in Istanbul. The location of the 3rd airport was also selected as Silivri–Gazitepe in the plan.
All the mega projects implemented in the metropolitan area of Istanbul since 2009 were in direct contradiction to the recommendations of the unanimously approved strategic plan: (i) The construction of The Yavuz Sultan Selim Bridge (commonly known as the 3rd bridge) was started in 2013; it was opened to traffic on 29 August 2016. The bridge crossed the Bosphorus strait between Garipçe village on the Euro-pean side and Poyraz on the Asian side. It was constructed as part of the Northern Marmara motorway, which included a total of 115 km of motorway and access roads. (ii) The 3rd airport, which was planned as the largest airport in the world, with 150 million annual passengers in its last planned expansion capacity, was suddenly moved from its optimal location to a new location in the heart of the Northern Forest zone; it was built in the Arnavutköy region on the European side near lake Terkos, an important water supply reservoir. Construction works started on 7 June 2014 and the con-struction zone covered an area of 7659 ha, mainly located in a state-owned forest area, affecting more than 2.5 mil-lion trees either cut or presumably relocated. It was opened on 29 October 2018. (iii) The Istanbul Canal project, still in the planning phase, which would involve a 500 m wide and 25 m deep artificial waterway that would stretch nearly 50 km to connect the Sea of Marmara with the Black Sea (Orhon 2014a, b). These mega-projects, either planned and/ or implemented, completely challenged the sustainability of the fragile ecological structure of water basins, forests and agricultural lands in the northern part of Istanbul.
Istanbul, the most populated city in Turkey with a surface area of 5340 km2 is one of the mega-cities in Europe. As
shown in Fig. 1, it is located in the north-west Turkey, along both sides of the Bosphorus strait, where the western side of
the city is in Europe and its eastern side is in Asia. The city is surrounded by the provinces Kocaeli on the east, Tekirdağ and Kırklareli on the west, the Black Sea on the north and the Marmara Sea on the southern side.
In this study, related parts of both the European and Asian sections, which included forest lands outside of urban areas of Istanbul, have been selected as the study area as presented in Fig. 1. The study area covered an area of approximately 398,000 ha. The selected area is under the pressure of two mega projects, namely, the 3rd Bridge across the Bosphorus and 3rd Airport projects, as they are located in a very pre-carious ecosystem in the Istanbul’s northern region (Orhon
2014a).
Materials and methods
Data and methodology
In this study, 29 June 2009 dated Landsat (ETM +) image and 10 July 2016 dated Landsat 8 OLI image were used to determine the LULC characteristics of the study area. Landsat 7 satellite is equipped with the ETM + (Enhanced Thematic Mapper Plus) which collects data in 8 bands involving the VNIR (Visible and Near Infrared), SWIR (Shortwave Infrared) range with 30 m resolution and pan-chromatic wavelength regions with 15 m resolution, while TIR (thermal infrared–thermal infrared range) collects data with 60 m resolution. Six bands of Landsat 8 and Landsat 7 ETM + data which have closer bandwidth were used in this study. The Landsat 8 contains 11 bands, is equipped OLI and Thermal Infrared Sensor (TIRS). OLI collects data for the visible, near-infrared (NIR) and shortwave infrared (SWIR) with 30 m resolution and panchromatic wavelength regions with 15 m resolution, while TIRS collects data for the two thermal wavelength regions with 100 m resolution. Both of the remotely sensed data were downloaded from the United States Geological Survey (USGS) Earth Resources Observation Systems Data Center (URL 1). All visible and infrared bands (except the thermal infrared) were integrated into the analysis. In addition to Landsat imageries, ancillary data including (1:25,000 scale) standard topographic maps, (1:5000 scale) orthophoto maps and google earth, were used as reference data for the training site selection, classifica-tion process, accuracy assessment of classified images and change detection. A schematic flow chart of the methodol-ogy adopted was given in Fig. 2. The methodology applied for detecting the LULC changes using remote sensing tech-nology occurred from 2009 to 2016 consisted of three main stages as image preprocessing, image classification and change detection. Accuracy assessment of image classifi-cation and change detection processes were also provided for making decisions on the usability of the outputs of each
process considering the required accuracy levels. The main output of the change detection process was the change map and it is used to produce change matrix for the use of carbon loss determination through the changes carbon stocks in the study area.
Image preprocessing
Since the remote sensing data provided by USGS was atmos-pherically corrected, image preprocessing stage of the study only included geometric correction of images used in the study. Accurate spatial registration of multi-temporal images is essential for change detection (Gong 2012). It has been suggested that a root mean squared error (RMSE) of 0.5 pixels is the maximum tolerable error for change detection (Jensen 2005). High accurate geometric registration of a multi-temporal Landsat images, with root mean square error (RMSE) of 0.25–0.5 pixel is necessary for accurate change detection (Jensen 2005; Mouat et al. 1993).
In this study, Landsat Images were registered to a UTM (Universal Transversal Mercator) projection system based on
1:25,000 scaled topographic maps of the study area using a first-order polynomial method. During image rectification, a nearest neighbor resampling algorithm was used, and the process followed with approximately 15 m (0.5 pixel) of RMSE for both images as required.
Image classification
This study used per-pixel supervised classification, which grouped satellite image pixels with the same or similar spectral reflectance features into the same information cat-egories (Campbell 2002). In the classification process, land cover types were defined in accordance with the CORINE (Coordination of Information on the Environment) land cover nomenclature for interpretation of remotely sensed data at various scales and resolutions (Bossard et al. 2000), however, land cover types were modified to fit the study objectives. Supervised classification method was performed by applying maximum likelihood algorithm which assumes that the statistics for each class in each band are normally distributed and calculates the probability that a given pixel Fig. 1 Location of the study area
belongs to a specific class (Lu and Weng 2007). All pixels were classified and assigned to the classes (water, forest, green area, agricultural land, urban & build-up, roads, sand dune, bare land, open mining area and destroyed area) hav-ing the highest probability (i.e., the maximum likelihood).
Accuracy assessment of image classification process
Classification accuracy is typically defined as the degree to which the derived image classification agrees with reality or conforms to the ‘truth’. The ground truth data should be selected precisely to represent the characteristics of the land in practice. A minimum of 85% remote sensing data should be achieved when interpreting the accuracy in the identification LULC change categories (Congalton
1991; Jensen 2005). Overall accuracy, user’s accuracy, producer’s accuracy and kappa coefficient are important measures in accuracy evaluation (Foody 2002). The clas-sification accuracy of each image is basically assessed through error matrix.
Change detection
Change detection analysis describes and quantifies differ-ences between images of the same scene at different times. The change detection process started with detecting the changes between two data sets and then the characteris-tics of the changes were identified. Once the identification
process was completed spatial extends of the changes were measured and the pattern to be assigned to these changes were assessed (Singh 1989). Following image classifi-cation, a post-classification comparison, which is also referred as delta classification, change detection algorithm was used to observe and determine the LULC changes in the study area (Coppin et al. 2004). In this study, two clas-sified image pairs were compared using cross-tabulation to determine qualitative and quantitative aspects of the LULC changes.
Accuracy assessment of the change detection process
This study adopted the approach of change/no change vali-dation to evaluate the accuracy of the change map produced for 2009–2016 period. The change detection accuracy was obtained by randomly sampling the study area to calculate an error matrix for the two classified imageries. Sample size was determined using the standard formula as following:
where Z value is a constant value determined based on the confidence level (e.g. 1.96 for 95% confidence level), P stands for the expected accuracy and e represents the error tolerance. A total of 650 samples were used in the change detection assessment, points on the boundaries of change areas (e.g. mixed pixels) were not included in the analysis resulting in 398 samples of the “change” and 191 of the “no change” category. All sample points were extensively analyzed by visual inspection, compared and validated with reference data of the study. Additionally, an intensive analy-sis for the change detection accuracy assessment was carried on-site by the experts familiar with the study area.
Carbon loss through carbon stock changes
The carbon stock basically involves five different compo-nents: the stock in the tree aboveground and belowground, deadwood, litter and soil. The carbon densities (ton/ha) of these pools are required to estimate the carbon emissions and removals.
The stock change of carbon depending on the land-use changes between 2009 and 2016 was calculated using the methodology described in IPCC (2006).
where ΔCConversion is the change in carbon stocks of different
carbon pools (biomass, litter, deadwood and soil organic carbon) on land converted to another land category, ton C. (1) N= Z2× P × (1 − P) e2 (2) ΔCConversion= ∑ i
(CAfteri− CBeforei) × ΔATo_othersi
Fig. 2 Flow-chart of the methodology for land use/land cover and
CAfter
i is the carbon stocks of different carbon pools on land
type i after the conversion, ton C/ha. CBeforei is the carbon
stocks of different carbon pools on land type i before the conversion, ton C/ha. ΔATo_othersi is the area of land use i
converted to another land-use category between 2009 and 2016, ha. i = type of land use converted to another land-use category.
The carbon densities given in Table 1 were developed for different LULC categories. Roads and residential areas were kept in their related classifications whereas gardens, parks within the residential areas were classified as green area. Other areas in the residential group without plant cover were categorized as destroyed areas and their carbon stocks were accepted as 0 ton C/ha. Details of the calculations were provided as the Supplementary Material.
Experimental results
Land cover classifications and changes
As a result of image classification, a total of nine LULC classes were produced for 2009 dated image (water, for-est, green area, agricultural land, urban & build-up, roads, sand dune, bare land and open mining area). A total of ten classes were identified for 2016 dated image, i.e. the same nine classes and an additional class for the “destroyed area”, which has top significance for the study, as it is directly
linked for the mega projects in the area. Images were then compared in terms of the total area of each land cover cat-egory. Figure 3a, b illustrated classification maps generated for the year 2009 and 2016 respectively.
The numerical analyses of the results illustrated in Fig. 3, yielding the fractionation of the study area among differ-ent land cover classes and the magnitude of changes that occurred between 2009 and 2016 were summarized in Table 2. As presented in the table green area, agricultural land, sand dune and bare land were considered as natural and semi-natural lands while urban & buildup, roads, open mining area and destroyed area were assigned as artificial surfaces. Change rates of the LULC classes were presented in terms of areal change (ha) and its ratio per each LULC class together with per total study area. Change per LULC class demonstrates that an ultimate change in forest area was observed as −7% which represented a reduction of the existing area, whereas 12.8% of increase was discovered in agricultural land. Considering the LULC in the whole area indicated a decrease of 3.8% in the forest area, of 1.6% in the green area and an increase of 1.6% in agricultural land.
Assessment of the classification accuracy for 2009 and 2016 imageries was carried out by randomly selecting approximately 300 points for each LULC class to deter-mine the quality of information derived from the data. As a result of the accuracy assessment process, the overall accu-racies for 2009 and 2016 were determined as 92.4% and 91.3%, respectively, which are above the limit set as 85%. Table 1 Carbon density of different carbon pools on land use category (ton C/ha)
1 NIR Turkey (2018)
2 Calculated for İstanbul based on National Forest Inventory 2009
3 Carbon stocks for roads and settlements are accepted as 0 because they are impermeable surfaces
4 Tolunay and Çömez (2008)
5 Tolunay (2011)
6 General Directorate of Combating Desertification and Erosion unpublished data (GDCDE 2018)
7 Tolunay et al. (2017)
8 Tecimen (2000), carbon stocks in open mining area and destroyed area accepted as the same value
Land cover type Aboveground
biomass Belowground biomass Total biomass Litter Deadwood Soil organic carbon (for 0–30 cm) Total carbon
stocks Water 01 01 0 0 0 01 0 Forest 24.542 8.232 32.77 5.804 0.605 43.196 82.36 Green area 0.491 1.371 1.86 0.061 0 38.376 40.29 Agricultural land 0.751 01 0.75 0.271 0 35.176 36.19 Sand dune 0 0 0 0 0 1.457 1.45 Bare land 0 0 0 0 0 33.856 33.85
Urban and build up 03 03 03 03 03 03 0
Roads 03 03 03 03 03 03 0
Open mining area 0 0 0 0 0 0.738 0.73
The kappa statistics were estimated as 0.90 for 2009 and 0.89 for 2016, indicating that 89% of the classification was better than a random classification. These values reflected reasonably good overall accuracy and were accepted for the subsequent analysis and change detection (Congalton 1991; Jensen 2005). Users and producers’ accuracies of individual classes were consistently high, ranging from 73% to 100%
for both 2009 and 2016. The lowest accuracy values were achieved as 73% while determining the open mining areas since they were mixed with bare land. However, this result did not have a significant effect on the further analysis of the study because open mining areas covered such a limited area (approximately 7500 ha) which corresponded less than 2% of the study area.
Fig. 3 Maps of supervised clas-sification results a in 2009 and
The change detection accuracy was also performed by randomly sampling the study area for the calculation of an error matrix for two classified imageries. For this purpose, 398 samples were classified as “change” out of 589 random samples, while the remaining were assigned as “no-change” between 2009 and 2016. The overall accuracy of change detection was 92%, with Kappa statistic of 0.90. False detec-tion or commission error was detected as 6% and omission error of the change detection process was assessed as 2%.
These results confirmed the outcomes of the previous research studies provided by Göksel et al. (2018), which highlighted a rapid change in forest and green area in Sarıyer and Beykoz districts in Istanbul during the same study period, and Ayazli et al. (2015), which estimated a rapid decrease in forest and green area based on the construc-tion of the third bridge in Istanbul. Obtained LULC changes were mostly parallel with the environmental outcomes of the study by Dogan and Stupar (2017) which highlighted the effects of the anticipated expansion of the city towards the north in consequence of the mega projects applied in Istanbul.
Land cover conversions
A matrix of LULC changes from 2009 and 2016 was created and displayed in Table 3, to further clarify and visualize con-versions among all LULC classes considered. The table out-lined the unchanged pixels along the major diagonal of the matrix. One of the most significant results was the change in an area of 6422 ha, covered by a new LULC class of destroyed areas. This LULC class appeared as a result of 3rd bridge and 3rd airport constructions. Furthermore, the lands transferred to agricultural land were mostly from the green area (18,543 ha) and bare land (6486 ha). Urban and built-up
class was increased from 18,676 ha in 2009 to 25,322 ha in 2016 with an increase rate of %35.6 as presented in Table 2. This change mainly affected the agricultural land and green area. The change observed in the forest and green area was evaluated in the following part.
Impact on forest area
The change matrix given in Table 3 presented that 176,507 ha of 213,424 ha forest area existing in 2009 remained unchanged in 2016. The rest of the area (36,917 ha) was converted to different uses in 2016, which counted as 17.3% of the total forest area and 9.3% of the whole study area. Considering the recoveries throughout the years till 2016 the ultimate loss was reduced down to 7% of the total forest area as given in Table 2. 2117 ha of this loss was the change from forest to roads and destroyed area caused by construction works of 3rd bridge and 3rd airport. This amount was 5.7% of the total forest loss and 1% of the total forest area. Figure 4a presented the spatial distribution of the total forest loss and unchanged forest lands in the study area. As this Figure indicated, the for-est loss was mostly concentrated at construction sites and along the connecting roads.
Figure 4b presented the main dynamics of the observed forest loss by identifying the general characteristics of the conversions from the forest area. In this context, the changes from forest area occurred in two major directions: namely, as artificial surfaces and natural/semi-natural areas. This observation is quite significant as it directly reflected the impact of the construction project to the for-est area. This impact could be clearly visualized in the figure as “purple colored” areas.
Table 2 LULC classification results and change in 17 years
Land cover type 2009 (TM) area (ha) % Ratio in
total area 2016 (OLI) area (ha) % Ratio in total area Areal change (ha) % Change % Overall change
Forest 213,424 53.6 198,466 49.9 −14,960 −7.0 −3.8
Water 10,889 2.7 9905 2.5 −984 −9.0 −0.2
Natural and semi natural lands
Green area 72,759 18.3 66,324 16.7 −6435 −8.8 −1.6
Agricultural land 49,277 12.4 55,607 14.0 6330 12.8 1.6
Sand dune 1249 0.3 928 0.2 −321 −25.7 −0.1
Bare land 19,148 4.8 20,724 5.2 1575 8.2 0.4
Artificial surfaces
Urban and build up 18,676 4.7 25,322 6.4 6646 35.6 1.7
Roads 5088 1.3 5814 1.4 726 14.3 0.2
Open mining area 7490 1.9 8488 2.1 998 13.3 0.3
Destroyed area 0 0 6422 1.6 6423 100.0 1.6
When the change matrix in Table 3 examined in more detail, it accounts for a significant recovery from green land (18,722 ha) and bare land (2935 ha) into the forest area. However, this result should not be considered as forest recovery, because the indicated areas were already plan-tation areas in 2009. They were not identified accurately during the classification process due to similar reflectance values of young trees and the green area leading to mixing class problems in practice. In fact, the forestation target of the 2009 Environmental Management Plan was limited to approximately 3500 ha, a much smaller area compared to the above recovery values.
Impact on carbon balance Destruction of carbon stocks
Forest and green area are usually considered as lungs of adjacent urban developments. In this context, the studied LULC change, which resulted in a significant reduction of forest and green area in the northern part of Istanbul was also evaluated in terms of related carbon/carbon dioxide balance. The impact of LULC changes on CO2 emission or
absorption was estimated within two different perspectives: (i) assessment of destroyed growing carbon stocks and the magnitude of CO2 emission due to the resulting chance in
carbon stocks; (ii) annual loss of CO2 that the destroyed
for-est and green area could have absorbed if they had remained intact and continued their functions.
Table 4 reflects and expresses the conversion of previ-ously presented LULC changes (from to analysis) as carbon stock changes. The values in Table 4 was calculated based on the LULC conversion values presented in Table 3 using carbon density values of different carbon pools on land use category (see Table 1) as an index. Negative values repre-sented the carbon loss due to the loss in LULC class (for example −1,179,012 ton C lost due to the loss in the green area (28,025 ha presented in Table 3)) and positive values represented carbon gain due to recovery of the LULC class (for example 787,635 ton C gain due to recovered forest area (18,722 ha in Table 3) from the green area) in the study area. Related date in this table indicated a total reduction of 1.2 million tons of C. This basically means that 4.4 mil-lion tons CO2 were additionally emitted to the atmosphere, due to observed reduction of carbon stocks in 7 years, i.e. between 2009 and 2016. The greenhouse gas inventory study for Istanbul in 2015 reported a unit CO2 emission of 3.22
tons CO2/ca.year, including all related activities in the city (CCAP 2015). To have a concrete idea about the magnitude of CO2 emission related to lost carbon stock, it should be
noted that it is equivalent to more than that of 1.35 million people. Table 3 Matr ix of land co ver con versions fr om 2009 t o 2016 Land co ver type W ater For es t Natur
al and semi natur
al lands Ar tificial sur faces To tal 2016 (ha) Gr een ar ea Ag ricultur al land Sand dune Bar e land U
rban and build up
Roads Open min -ing ar ea Des tro yed ar ea W ater 9062 323 59 187 51 16 18 12 177 NA 9905 For es t 0 176,507 18,722 0 3 2935 82 180 37 0 198,466 Natur
al and semi natur
al Gr een ar ea 88 28,025 23,528 9617 62 3642 358 467 537 0 66,324 A gr icultur al land 58 99 18,543 26,911 173 6486 1480 826 1031 0 55,607 Sand dune 177 66 59 96 248 29 62 28 163 0 928 Bar e land 113 3949 6650 5397 81 2984 499 310 741 0 20,724 Ar tificial sur faces U
rban and build up
61 856 1876 3774 205 1568 14,000 1442 1540 0 25,322 R oads 140 323 730 983 54 398 1119 1251 816 0 5814 Open mining ar ea 672 1482 1251 1416 251 714 636 386 1680 0 8488 Des tro yed ar ea 518 1794 1341 896 121 376 422 186 768 0 6422 To tal 2009 (ha) 10,889 213,424 72,759 49,277 1249 19,148 18,676 5088 7490 0 398,000
As given in Table 4, the carbon stock decrease related to the destroyed forest land was calculated as 1.77 million tons C (6.5 million tons CO2) for the period of 2009 to 2016. On
the other hand, restored carbon stock of due to reforestation in the same period was approximately 0.95 million tons C (3.5 million tons CO2) so that the net carbon stock decline could be corrected as 0.82 million tons C or 3.0 million tons
CO2. This amount indicated that 70% of the total C/CO2
emission associated with LULC changes could be attributed to the loss of forest land.
The higher carbon exchange between land cover types occurred between forest land and green area. The second-largest carbon stock loss from forest area was 191,566 tons C to bare land and the next one was 146,444 tons C to the Fig. 4 a Total forest loss from
2009 to 2016. b Change from the forest to artificial and natu-ral/semi-natural lands
destroyed area. The total carbon/CO2 emission associated with bare land, urban & build-up, roads, destroyed area amounted to 871,360 tons C (3,195,000 tons CO2), which
corresponded to 73% of the total carbon emission related to LULC changes.
As previously explained, carbon stocks include five different components. While the tree biomass is the pre-dominant component, deadwood, litter and soil may have a significant role in the calculation of the carbon stock; hence, they need to be considered as separate carbon pools. Table 5, summarizes the distribution of the carbon emissions among these pools, generated by changes in LULC between 2009 and 2016.
Annual loss of carbon absorption
There is no reliable information about the size and nature of trees lost during the destruction of the forest area. Hence, only an approximate can be made for estimating the decrease in CO2 absorption associated with the reduction of forest area. General studies on the subject indicated that the yearly CO2 absorption rate may be assumed to vary in the range of
3–10 kg CO2/tree depending on the type of trees based on
a 15 cm diameter (Tennyson 2014). Assuming that a tree requires an area of 4.0 m2, a forest as dense as in the
north-ern part of Istanbul would contain 2500 trees/ha. There-fore, the destroyed forest area, assessed as approximately 15,000 ha is this study, would house 35 million trees. Based on a unit absorption value of 8 kg CO2/tree.year, this land cover change would correspond to a CO2 absorption loss
of 0.3 million tons CO2/year. Using the generally accepted
CO2 emission value of 1 kg CO2/person.day, (0.365 tons CO2/person.year) based on a daily respiration of 500 L of air, the yearly loss in the CO2 absorption capacity would be
equivalent to the emission of 830,000 people in Istanbul.
Conclusion
The results presented above should be regarded as the miss-ing essential and decisive components of the environmental impact assessment studies that should have been carried out prior to the implementation of the mega land-use projects that drastically destroyed the ecosystem in the northern for-est area of Istanbul. They clearly showed that if properly executed as in this study, related environmental impact state-ments had a chance to generate a compelling and scientific evidence to prevent these projects way before their imple-mentation mainly because:
(i) The observed results detected an overall land cover change of 11.5% in the study area of approximately
Table 4 Chang es in t ot al carbon s toc ks (t on C) due t o con versions be tw een differ ent land co vers in t he per iod of 2009–2016 Land co ver types W ater For es t Gr een ar ea Ag ricultur al land Sand dune Bar e land U
rban and build up
Roads Open mining ar ea To tal 2016 W ater 0 −26,602 −2377 −6768 −74 −542 0 0 −129 −36,492 For es t 0 0 787,635 0 243 142,377 6754 14,825 3020 954,853 Gr een ar ea 3546 −1,179,012 0 39,430 2408 23,454 14,424 18,815 21,244 −1,055,691 Ag ricultur al land 2099 −4571 −76,026 0 6010 15,177 53,561 29,893 36,559 62,703 Sand dune 257 −5340 −2292 −3335 0 −940 90 41 117 −11,402 Bar e land 3825 −191,566 −42,826 −12,629 2624 0 16,891 10,494 24,542 −188,645 U
rban and build up
0 −70,500 −75,584 −136,581 −297 −53,077 0 0 1124 −334,915 Roads 0 −26,602 −29,412 −35,575 −78 −13,472 0 0 596 −104,544 Open mining ar ea 491 −120,976 −49,490 −50,211 −181 −23,648 464 −282 0 −243,832 Des tro yed ar ea 378 −146,444 −53,050 −31,772 −87 −12,453 308 −136 0 −243,256 To tal 2009 10,595 −1,771,613 456,578 −237,441 10,568 76,877 92,492 73,650 87,073 −1,201,221
400,000 ha; this change mainly reflected the destruc-tion of forest and green area: Nearly 15,000 ha, cor-responding to 7.0% of the forest land and 8.8% of the green area amounting to 6.350 ha were lost due to mega land-use projects implemented in this area. (ii) In addition, they indicated that an equivalent of more
than 4.4 million tons of CO2 was additionally emitted
to the atmosphere, due to observed reduction of car-bon stocks. More than 70% of the total CO2 emission
associated with LULC changes could be attributed to the loss of forest land. Also, the destroyed forest land corresponded to a CO2 absorption loss of 0.3 million
tons CO2/year, equivalent to the respiratory emission
of 830,000 people in Istanbul.
More important than the numbers presented, the results of the study provided conclusive evidence that the sustain-ability of the ecosystem in the area was severely damaged because the biological environment would no longer be able to thrive and support itself, since the outside influence will be detrimental, especially in the long run.
Observed deforestation and loss of green area should not be compared to those occurring during natural disasters, such as fire; earthquake, etc., since they took place as irre-versible conversions to the airport and auxiliary commercial and industrial services, residential areas, roads, other barren areas due to human activities and increased agricultural land. In similar situations, this type of human activities always exhibits an increasing trend, which would further impair environmental sustainability.
In this context, it is strongly recommended that further projects and manmade activities be prevented in the cumu-lative sense, to safeguard and restore to the extent possible
the sustainability of this area of vital importance for Istanbul and in all regions of similar character around the world.
References
Akin A, Sunar F, Berberoglu S (2015) Urban change analysis and
future growth of Istanbul. Environ Monit Assess 187(8):506. https
://doi.org/10.1007/s1066 1-015-4721-1
Al-Reasi HA, Kwarteng A, Sulaiman H, Amoatey P (2018) Above-ground carbon dynamics in different arid urban green spaces.
Environ Earth Sci 77:431. https ://doi.org/10.1007/s1266
5-018-7613-1
Ayazli IE, Kilic F, Lauf S, Demir H, Kleinschmit B (2015) Simulat-ing urban growth driven by transportation networks: a casestudy of the Istanbul third bridge. Land Use Policy 49(2015):332–340 Bauer ME, Yuan F, Sawaya KE (2003) Multi-temporal Landsat
image classification and change analysis of land cover in the twin cities (Minnesota) metropolitan area. MutiTemp-2003. In: Second international workshop on the analysis of multi-temporal remote sensing images, July 16–18, 2003. Ispra, Italy Bossard M, Feranec J and Otahel J (2000) CORINE land cover tech-nical guide—Addendum 2000 Techtech-nical report No 40. Euro-pean Environment Agency
Campbell JB (2002) Introduction to remote sensing. Taylor&Francis, London. ISBN: 0898627761
CCAP (2015) Istanbul climate change action plan: greenhouse gas inventory, Project Report, Istanbul Climate Change Action Plan
Preparation Project. https ://www.iklim .istan bul/wp-conte nt/
uploa ds/serag aziev anter i.pdf. Accessed 25 Feb 2020
Congalton RG (1991) A review of assessing the accuracy of classi-fications of remotely sensed data. Remote Sens Environ 37:35– 46. https ://doi.org/10.1016/0034-4257(91)90048 -B
Coppin P, Jonckheere I, Nackaerts K, Muys B (2004) Digital change detection methods in ecosystem monitoring: a review. Int J Remote Sens 25(9):1565–1596
Dogan E, Stupar A (2017) The limits of growth: a case study of three mega-projects in Istanbul. Cities 60:281–288
Dogru AO, Bektas Balçık F, Goksel C, Ulugtekin NN (2006) Monitoring coastal dunes by using remote sensing and GIS
Table 5 Distribution of carbon emissions among different carbon pools due to changes in land uses between 2009 and 2016
Land cover types Carbon pools Total
Aboveground
biomass Belowground biomass Total biomass Litter Deadwood Soil
Water −8096 −2739 −10,835 −1927 −194 −23,536 −36,492 Forest 529,700 155,073 684,774 126,239 13,175 130,665 954,853 Green area −673,976 −172,015 −845,991 −162,574 −16,815 −30,311 −1,055,691 Agricultural land 10,006 −26,219 −16,212 6061 −59 72,913 62,703 Sand dune −1721 −624 −2345 −412 −40 −8605 −11,402 Bare land −104,215 −41,611 −145,825 −24,760 −2369 −15,690 −188,645
Urban and build up −24,756 −9615 −34,371 −6096 −514 −293,934 −334,915
Roads −9021 −3658 −12,680 −2183 −194 −89,488 −104,544
Open mining area −38,043 −13,911 −51,954 −9053 −889 −181,936 −243,832
Destroyed area −45,354 −16,602 −61,956 −10,728 −1076 −169,497 −243,256
integration in North West part of Turkey: a case study of Kilyos dunes. Fresenius Environ Bull 15(9b):1227–1231
Foody GM (2002) Status of land cover classification accuracy
assess-ment. Remote Sens Environ 80(2002):185–201. https ://doi.
org/10.1016/S0034 -4257(01)00295 -4
GDCDE (2018) Soil organic carbon project. Unpublished project final report by TÜBİTAK-BİLGEM and General Directorate of Combating Desertification and Erosion
Goksel C, David RM, Dogru AO (2018) Environmental monitoring of spatio-temporal changes in northern Istanbul using remote
sensing and GIS. Int J Environ Geoinf 5(1):94–103. https ://doi.
org/10.30897 /ijege o.41094 3
Gong P (2012) Remote sensing of environmental change over China: a review. Chin Sci Bull 57:2793–2801
Griffiths P, Hostert P, Gruebner O, van der Linden S (2010) Mapping megacity growth with multi-sensor data. Remote Sens Environ
114(2):426–439. https ://doi.org/10.1016/j.rse.2009.09.012
Gürel A, Gündüz AE (2011) İstanbul’un ekolojik yapısı üzerine bir araştırma—a research on the ecological structure of Istanbul. Marmara Sosyal Araştırmalar Dergisi- The Journal of Marmara Social Research, 1, December 2011 (in Turkish)
IBB (2009) Istanbul Environmental Plan. http://www.ibb.gov.tr/tr-TR/
kurum sal/Birim ler/Sehir Planl amaMd /Docum ents/100.000%20 %C3%96l%C3%A7ekl i%20%C3%87evr e%20D%C3%BCzen i%20Pla n%C4%B1%20Sun um%20TR.pdf Accessed 1 Mar 2020 IPCC (2006) IPCC Guidelines for national greenhouse gas inventories.
Prepared by the National Greenhouse Gas Inventories Programme, Eggleston HS, Buendia L, Miwa K, Ngara T and Tanabe K. (eds). Published: IGES, Japan
Jensen JR (2005) Introductory digital image processing: a remote sens-ing perspective, 3rd Edn. Prentice Hall, Upper Saddle River Kharazmi R, Tavili A, Rahdari MR, Chaban L, Panidi E, Comino RJ
(2018) Monitoring and assessment of seasonal land cover changes using remote sensing: a 30-year (1987–2016) case study of
Ham-oun Wetland, Iran. Environ Monit Assess 190:356. https ://doi.
org/10.1007/s1066 1-018-6726-z
Lu D, Weng Q (2007) A survey of image classification methods and techniques for improving classification performance. Int J Remote
Sens 28(5):823–870. https ://doi.org/10.1080/01431 16060 07464 56
Mayomi I (2009) Assessment of human impacts on landuse and veg-etation cover changes in Mubi region, Adamawa State, Nigeria; remote sensing and GIS approach. Global J Environ Sci 8(2):1B. https ://doi.org/10.4314/gjes.v8i2.53778
Mouat DA, Mahin GC, Lancaster JS (1993) Remote sensing techniques in the analysis of change detection. Geocarto Int 2:39–50 Orhon D (2014a) Istanbul canal needs environmental study. Nature
513(7519):461
Orhon D (2014b) Will the Marmara Sea survive? The struggle against pollution. Government Gazette, March 2014, pp 14–15. ISSN: 1042-4167
Sanli FB, Bektas FB, Goksel C (2008) Defining temporal spatial patterns of mega city Istanbul to see the impacts of increasing
population. Environ Monit Assess 146:267–275. https ://doi.
org/10.1007/s1066 1-007-0078-4
Singh A (1989) Digital change detection techniques using
remotely-sensed data. Int J Remote Sens 10(6):989–1003. https ://doi.
org/10.1080/01431 16890 89039 39
Taubenböck H, Esch T, Felbier A, Wiesner M, Roth A, Dech S (2012) Monitoring urbanization in mega cities from space. Remote
Sens Environ 117(2012):162–176. https ://doi.org/10.1016/j.
rse.2011.09.015
Tecimen HB (2000) Ağaçlı (İstanbul) kömür ocakları artıkları üstündeki ağaçlandırmanın ham matertyaldeki organik madde ve azot birikimine etkileri. MSc Thesis, İstanbul University, Institute of Science and Engineering (in Turkish)
Tennyson R (2014) Ontario trees that sequester carbon dioxide and other atmospheric pollutants, Environmental benefits from energy
conservation. https ://www.toron to.ca/legdo cs/mmis/2014/pe/bgrd/
backg round file-70898 .pdf. Accessed 25 Feb 2020
Tolunay D (2011) Total carbon stock and carbon accumulation in liv-ing tree biomass in forest ecosystems of Turkey. Turk J Agric For
35:265–279. https ://doi.org/10.3906/tar-0909-369
Tolunay D, Çömez A (2008) Türkiye ormanlarında toprak ve ölü örtüde depolanmış organik karbon miktarları—amounts of organic car-bon stored in forest floor and soil in Turkey. Hava Kirliliği ve Kontrolü Ulusal Sempozyumu 2008. 22–25 October 2008, Hatay (in Turkish)
Tolunay D, Makineci E, Şahin A, Özturna AG, Pehlivan S (2017) İstanbul-Durusu kumul alanlarındaki sahil çamı (pinus pinaster ait.) ve fıstık çamı (pinus pinea L.) ağaçlandırmalarında karbon birikimi. National Project supported by TÜBİTAK TOVAG, Pro-ject No: 114O797 (in Turkish)
TUIK (2020) Official website for Address Based Population
Registra-tion System in Turkey by Turkish Statistical Institute. https ://birun
i.tuik.gov.tr/medas /?kn=95&local e=en. Accessed 2 Mar 2020 NIR Turkey (2018) National greenhouse gas inventory report of the
Turkey 1990–2016. https ://unfcc c.int/docum ents/65716 . Accessed
25 Feb 2020
URL 1: Official website of USGS Earth Resources Observation
Sys-tems Data Center, https ://earth explo rer.usgs.gov/. Accessed 25
Feb 2020
Wang J, Jiao Z, Bai Z (2014) Changes in carbon sink value based on RS and GIS in the Heidaigou opencast coal mine. Environ Earth Sci
71:863–871. https ://doi.org/10.1007/s1266 5-013-2488-7
Zhu Z, Woodcock CE (2014) Continuous change detection and classifi-cation of land cover using all available Landsat data. Remote Sens
Environ 144:152–171. https ://doi.org/10.1016/j.rse.2014.01.011
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
