Spatiotemporal Distribution. of Industrial Regions and Impact on LST in the case of Kocaeli, Turkey

Transcription

1 Spatitempral Distributin Presented at the FIG Cngress 2018, May 6-11, 2018 in Istanbul, Turkey FIG Cngress May 2018 f Industrial Regins and Impact n LST in the case f Kcaeli, Turkey Arzu Erener, Gulcan Sarp Kcaeli University, Department f Gematics Engineering, Kcaeli, Turkey, arzu.erener@kcaeli.edu.tr Department f Gegraphy, Suleyman Demirel University, Isparta, Turkey. gulcansarp@sdu.edu.tr

2 Intrductin Mnitring f urban sprawl especially grwth f industrial regins is an imprtant task t maintain the balanced and sustainable develpment, t understand status f urban air pllutin and ecsystems and t supprt urban planning. Additinally, it is als a fact that the increase in industrial areas negatively affects the glbal warming. 2

3 Intrductin Glbal warming ccurs when carbn dixide (CO2) and ther air pllutants and greenhuse gasses cllect in the atmsphere and absrb sunlight and slar radiatin that have bunced ff the earth s surface. Nrmally, this radiatin wuld escape int space but these pllutants, which can last fr years t centuries in the atmsphere, trap the heat and cause the planet t get htter. That's what's knwn as the greenhuse effect. 3

4 Study Regin Due t fast urbanizatin (traffic, ppulatin, building ppulatin etc.) and high dense industrializatin may cause glbal warming. Cal-burning pwer plants the burning f fssil fuels t make electricity the transprtatin sectr is the largest surce f heat-trapping pllutin 4

5 Intrductin Since 1980 s the industrial sectr in Turkey has develped with a remarkable rate. This can be attributed t three different factrs namely a shift frm agriculture twards industry, the mdernizatin f the existing industry, and the effect f internatinal trade (KPMG, 2015). Industrial regin densities may play rle n the Land Surface Temperature (LST) distributin. Remte sensing technlgies use the thermal infrared regin f the electrmagnetic spectrum in rder t bserve LST. 5

6 Intrductin Remte sensing images prvide fast mnitring f multi-tempral spatial data and GIS (Gegraphic Infrmatin Systems) Technlgies has becme an imprtant tl fr handling these spatial data. Satellite images are useful fr: land use mapping, change analysis, bject determinatin, etc., as well as thermal bands prvide pprtunity t determine the land surface temperatures (LST). 6

7 Intrductin Land surface temperature (LST) is defined as hw ht the surface f the Earth wuld feel t the tuch in a specific lcatin. LST is a significant parameter t determine the energy exchange between the surfaces f the earth and the radiant temperature that will help t understand the change f the surface temperature 7

8 Aim f the Study In this paper : spatial and tempral distributin f industrial regins and land surface temperature investigated and the relatinships amng these factrs are discussed. The change in industrial regins was determined with raster-based analysis by handling Landsat 7 ETM+ and Landsat 8 OLI satellite images belng t 2002 and 2016 years respectively. LST analysis was determined by handling: Landsat 7 and Landsat 8 OLI satellite images belng t 2000 and 2016 years respectively. 8

9 THE STUDY AREA AND DATA SETS Kcaeli is ne f the dense industrial prvinces f Turkey thrugh where D-100 and TEM highways pass thrugh the city. The city lies between the 29 22'-31 22' eastern lngitude and 4031'-4242' nrthern latitude and it carries a geplitical imprtance as it is lcated in junctin pint f rads cnnecting Asia and Eurpe. It is the Turkey's secnd largest industrial center after Istanbul. 9

10 THE STUDY AREA AND DATA SETS Kcaeli was mainly a prductin, strage and transfer regin fr mre than 1020 industrial institutins in varius sectrs as: Petrleum Refineries, autmtive, chemistry, textile, machine, fd, paper, wd, tanning, cal, etc.. 10

11 THE STUDY AREA AND DATA SETS The city is divided int 12 urban sub-municipalities. The ppulatin reached

12 THE STUDY AREA AND DATA SETS Landsat7 ETM+ (enhanced thematic mapper plus) and Ladsat 8 OLI images are available fr the study area. The Landsat7 ETM+ fr 2002 and Ladsat 8 OLI fr 2016 were used t btain landuse/landcver map f the regin. The Landsat7 ETM+ fr 2000 and Ladsat 8 OLI fr 2016 were used t btain LST map f the regin. Satellite Year Sensr Spectral Range L7 multispectral ETM µm L7 07/02/2002 thermal ETM+ 18/07/ µm L7 panchrmati ETM+ c 0.90 µm L8 OLI multispectral L8 OLI 22/02/2016 thermal L8 OLI 22/07/2016 panchrmati c Band #s Scene Size Pixel Reslutin 1,2,3,4,5,7 185x m km 6.1, m 8 15 m 1,2,3,4,5,6,7,9 180x m 10, m 8 15 m 12

13 THE STUDY AREA AND DATA SETS Atmspheric changes affect the radimetric reslutin f the satellite image, at the same time substances such as cluds, snw, rainy weather can lead t errrs when classifying and trying t distinguish bjects, and a separate class may be required fr these substances. Fr this reasn, it was nted the satellite images that were belnging t the same mnth were chsen. The images were btained frm the internet adress freely. 13

14 Methd and Analysis The methd f the study cntains f the prcessing f the image data t btain spatitempral change f industrial regins and btain LST values n these regins within 14 years perid in the case f Kcaeli, Turkey. 14

15 Methd and Analysis The clipping/subsetting prcess was carried ut initially in the Kcaeli prvince in rder t cver the industrial areas f Izmit Gulf. Then the 30 m reslutin Landsat images were pansharpaned with pan band t btain 15 m. spatial reslutin images. The images were then enhanced t increase the determinatin f the features. After btaining the landuse/landcver map by using SVM meths the accuracy was tested by using an errr matrix. a Landsat image b Landsat image 15

16 Methd and Analysis Spatitempral Industrial Regin Analysis Supprt vectr machine classificatin methd was used t classify the images. A hyperplane that can best distinguish the training data is btained by using the decisin functin. 16

17 Methd and Analysis Spatitempral Industrial Regin Analysis Fr SVM classificatin training areas were created by chsing plygns that cntain training pixels representing the land cvers. 8 land cver classes including water, industry, green regins, bare sil, urban and rad were selected fr training regins. The sigmid kernel was used. The gamma in kernel functin and penalty parameter was selected as and 100 respectively. 17

18 Methd and Analysis Spatitempral Industrial Regin Analysis Fr accuracy assessment purpses, selectin f grund truth pixels was dne by randm sampling. Accuracy analysis was carried by cmparing the classified pixels with grund truth pixels using a cnfusin matrix. The results were presented in terms f Kappa Cefficient and verall accuracy. Accrding t the results kappa cefficients 2002: kappa cefficients 2016: and verall accuracy 2002: 99.87% verall accuracy 2016: % 18

19 Methd and Analysis The classes ther than the industry in 2002 and 2016 were cmbined int ne class, reducing the number f classes t tw (Figure 4). These images cntaining tw classes were differentiated by change detectin difference analysis. As a result f this prcess 35.15% differences and increases in the field f industry have been bserved fr 14 years The industrial regins btained by classificatin was verlaid with Landsat images a industrial regins b industrial regins 19

20 Methd and Analysis Land surface temperature (LST) estimatin The spectral radiance and quantized calibrated values in DNs frm the relevant header file was btained befre calculating the radiance. The fllwing frmula was used t cnvert DNs t spectral radiance Where; Lʎ is the spectral radiance at the sensrs aperture in W/(m2* sr* μm); DN is the quantized calibrated pixel value, Lminʎ and Lmaxʎ are the minimum and maximum spectral radiance f thermal band respectively, QCalMin and QCalMax are the minimum and maximum quantized calibrated pixel value in DNs, respectively 20

21 Methd and Analysis Land surface temperature (LST) estimatin The spectral radiance (Lʎ) btained by equatin (7) is cnverted t satellite brightness temperature (Landsat prject science ffice, 2002; USGS, 2013) by using the fllwing cnversin frmula Where T is the effective at-satellite temperature in Kelvin (K), Lλ is the spectral radiance; K2 and K1 are the thermal band pre-launch calibratin cnstants. Fr Landsat7 ETM+ band 6.1 image, K2 = , and K1 = Fr Landsat8 OLI band 10 images, K2 = and K1 =

22 Methd and Analysis Land surface temperature (LST) estimatin The satellite temperature in Kelvin are then transfrmed t Celsius by using the equatin 22

23 Results and Discussin 3.Relatinship analysis between LST and industrial regin In rder t investigate the relatinship between the industrial regins and land surface temperature, the LST fr 2000 and 2016 was verlaid with industrial regin bundaries. 100 randm pints is emplyed in and arund the industrial regins t prduce a radiant temperature difference after the surface radiant temperature f each year has been nrmalized. 23

24 Results and Discussin 3.Relatinship analysis between LST and industrial regin The temperatures were resampled t database f generated randm pints frm the 2000 and 2016 LSM maps. Then the temperature values were cmpared by calculatin f the difference. 24

25 Results and Discussin The temperature difference f %70 f the distributed randm pints were higher than 0 which indicated 1 t 9 degrees f increase in the temperature. On the ther hand %30 indicates negative difference means decrease in temperature between -1 t -4. Figure 7. The frequency distributin fr LST 2000 nd LST

26 Results and Discussin Descriptive Statistics Min. Max. Mean Std. Dev. LST ,97 1,911 LST ,72 2,223 Table 2. The statistics fr sample pints Figure 7. The frequency distributin fr LST 2000 nd LST

27 Cnclusin The cmbined use f remte sensing and GIS allws fr an examinatin f the impact f industrial regin expansin n surface temperature. In this study, an integrated apprach f remte sensing and GIS was develped fr evaluatin f spatitempral distributin f industrial regin and its impact n surface temperature in Kcaeli city, Turkey. The results indicate that 35.15% increase in the field f industry has been bserved fr 14 years. 27

28 Cnclusin The LST values were cmpared in and arund industrial regins. Spatitempral difference f LST values were cmpared by emplying 100 sample pints. The results shwed that industrial develpment raised surface radiant temperature by 1 t 9 degrees fr 70% f randm sampled pints f the study regin. The min, mean and maximum temperature increased 1, 0.5 and 3 degrees respectively. As a result the integratin f remte sensing and GIS prvides an efficient way t detect industrial regin expansin and t evaluate its impact n surface temperature. 28

29 Thank yu s much fr yur cncern. 29