Uydu verileri ile istanbul Boğazı ve Haliç'de su kirliliğinin makro düzeyde belirlenmesi

Abstract

Dünya nüfusunun hızla artmasına paralel olarak tükenen doğal kaynaklar, enerji yetersizliği, çevre sorunları, yerleşim konuları, yeryüzü hakkında yeni bilgiler edinme isteği gibi sorunlara çözüm amacıyla uzay teknolojisindeki gelişme hızla ilerlemektedir. Arada fiziksel bir temas olmaksızın bir cisimden yayılan ısınımın nitelik ve nicelik yönünden değerlendirilmesi ile cismin özelliklerinin, uzaktan ortaya konması ve ölçülmesi şeklinde tanımlanan uzaktan algılama, son yıllarda bilgisa yarların hızlı gelişmeleri ile askeri amaçlar dışında bir çok bilim dalı içerisinde uygulama alanı bularak geniş bir şekilde kullanılmaktadır. Günümüzde çok sayıda uydu, çeşitli amaçlarla faali yetini sürdürmektedir. Bu uydular içerisinden yeryüzünün bilimsel araştırılmasına yönelik olarak algılama yapan LANDSAT (land-satellite) uydularından ikinci jenerasyon olarak adlandırılan Landsat-5 uydusu çok spektrumlu tarayıcı (MSS) ve tematik haritalama (TM) sistemleri ile yeryüzünü algılamaktadır. Landsat uydularından faaliyetini sürdüren, 1.3.1984 tarihinde yörüngeye oturtulmuş olan Landsat-5 uydusu, diğer algılayıcılara göre üstünlüğü olan TM algılayıcısını taşımaktadır. Bu üstünlük TM algılayıcısının çözünürlüğünün 30 m ve kanal sayısının 7 olmasından kaynaklanmaktadır. Bu çalışmada 12.6.1984 ve 24.10.1986 tarihlerinde Landsat-5 uydusu, TM algılayıcısı ile algılanan, istanbul'un güneyini, İstanbul Boğazı ve Haliç'in görüntü verilerini içeren manyetik band (CCT), görüntü işleme sis teminde incelenerek, çalışma için gerekli renkli ve sınıflandırılmış görüntüler oluşturulmuştur. Oluşturulan renkli görüntüler kontrolsüz ve kont rollü olarak çeşitli sınıflandırma yöntemine göre farklı kanallarda sınıflandırılmıştır. Sınıflandırma ile oluşan görüntülerde sularla kaplı alanlar olan, İstanbul Boğazı ve Haliç 'de kirliliğe neden olan evsel ve sanayi atıklarının oluşturduğu kirletici unsurlar olan hidrokimyasal maddeler dikkate alınarak yorumlanmıştır. Yorumlar bölgenin coğrafi durumu, hidrografik ve deşarjlarla ilgili mevcut veriler, dip topografyası ele alınarak irdelen miştir. Piksel parlaklık değerleri ile kirletici unsurlar olan TAM, HM, KOI ve PAH arasında her bir kanal için istatistiksel bağıntıyı belirlemek amacıyla regresyon analizi yapılmış, sonuçları irdelenmiştir. Sularla kaplı alanlarda kirliliğin uydu verileri ile makro düzeyde belirlenebileceği gösterilmiştir. Farklı yöntemlerle oluşturulan görüntülerin yorumları kendi aralarında da kıyaslanarak sonuçların benzerliği ortaya konmuştur.

Remote sensing of the earth's surface began with the use of aerial photography in the early 1900s. Aerial mapping cameras and photointerpretation were the tools used until the late 1960s. The instantaneous record of the terrain not only provided an overall view of the ground surface, but also enabled specific sites to be examined in great detail. Remote Sensing is the acquisition of information about an object without physical contact. The term "remote sensing" was coined in the early i960' s by geographers in the office of Naval Research to apply to the information derived from photographic and nonphotographic instruments. In the simplest case the human eye can be considered a remote sensor be cause it visually sensed information from the surroun ding world Special techniques in the remote sensing are applied to process and interpret at many scientific branch, resource surveys, water resource, environmental engineering applications, highway engineering projects, forestry, geography, geology and other, Remote sensing is acquired data with a sensor. Sensor systems has made available enormous quantities of photographic and other forms of data about the surface of the earth, data which have great potantial for helping to solve many human problems: for relieving critical food shortages; for monitoring and controlling environmental pollution, for water resources; for augmenting shrinking supplies of natural resources ; and for planning the orderly growth of cities. In view of these reguirement, these satellite date are of great human value, provided they can be re duced to useful information both quickly and economi cally. Modern, high-speed digital computers are equiped to this data-reduction task. Remote sensing aplications in generally including two step: firstly that data acquisition which use of different remote sensors and, secondly: data analysis as computer image processing, classifications and interpret the data. Remote sensing data-acquisition system is conside- dered to have four basic parts: the radiation source, the atmospheric path, the target, and the sensor. In a passive remote sensing system the primary radiation so urce is the sun, whose energy is spectrally distributed trought the electromagnetic spectrum. The energy then interacts with the target and is reflected, transmitted, -VII- and/or absorbed by it. A portion of the energy absorbed in one region of the spectrum may be emitted in another region of the spectrum. The reflected/emitted energy then passes back through the atmospher and again is sub jected to spectral and intensity modifications. Finally, the energy reaches the sensor where it is measured and converted into data for subsequent processing. Surface of earth reflects and/ or absorbs and emits incident energy is the signal that is to be detected and placed in a quantitative format. In generally assumed that the format is digital data suitable for processing on a computer. The processing of the data is done in such a way as to enhance the signal. Radiation from the sun, when viewed from outside the atmosphere is greatest between the wavelengths of approximately 0.38 and 3.0 um, the reflective region of the spectrum. Energy sensed in these wavelengths is pri marily radiation originating from the sun and reflected by objects on the earth. The reflective portion of the spectrum is further divided into the visible wavelengths and the reflective- infrared wavelengths. Since the human eye responds to radiation between wavelengths of approximately 0, 38 and 0.72 um, these wavelengths are referred to as the visib le wavelengths. The region between 0.72 and 3.0 um is referred to as the reflective-infrared portion of the spectrum, which in turn is further subdivided into the near-infrared (0.72 to 1.3 um) and the middle-infrared (1.3 to 3.0 um) wavelengths region. In fact, these lat ter two designations are preferred to the designation "reflactive infrared". Electromagnetic energy in wavelengths from 7.0 to 15 um is in the far-infrared region of the spectrum. The terms "emissive and thermal" are sometimes used to de signate this portion of spectrum. In this region, both reflection and solar radiation also accur as well as emission. In remote sensing applications involving the acqui sition of data from either aircraft or spacecraft sensor platforms, the atmosphere between the sensor and the target and between the radiation source and the target has an effect on the data. Generally, unless the atmos- feric effect is constant over the frame of data, correc tions for the effect may have a significant impact on the final analysis of the frame. The atmosphere may effect remote sensing data in two ways, trough scattering and absorbtion of energy. Scattering occurs when radiation is reflected or refrac ted by particles in the atmosphere which may range from molecules of the constituent gases to dust particles and large water droplets. The meteorological characteristics of the atmosphere strongly affect the relative dominance of scattering and absorption mechanisms. -VÜİ- Pure scattering is said to occur in the absence of all absorption; there is no loss of energy-only a redi rection of energy. In the remote sensing context, scat tering causes some of the energy to be directed outside the field of view of the sensor. If the field of view is very large, some scattered radiation will still be ac cepted; however, if the field of view is small, virtu ally all the scattered radiation will be rejected. In the latter case, scattering produces an apparent attenu ation or dimming of the image, whereas in the former ca se scattering increases the signal being received by the instruments due to the additional radiation entering the instrument aperture. In either case, however, the scat tering degrades the quality of the received data. Data acquisition in remote sensing is photographic or non-photographic systems. Photographic remote sensing systems are capable of sensing visible and near infrared regions of the electromagnetic spectrum (A = 0.4 to 0.9 urn). There are, however, other langer-wavelength regions of the electro magnetic spectrum, which are potentially useful for remote sensing purposes. Unfortunately, howe ver, conventional glass lenses and photographic emulsi ons are not able to detect and record these regions of the electro magnetic spectrum. An alternative method of detecting and recording both the visible, near infrared and longer-wavelength bands, such as thermal infrared and microwave energy, is the airborne scanner. Scanning is carried out by a rota ting mirror, which directs the incoming radiation on to a detector. The scanner detector is analogous to the emulsion of a conventional film camera. Its purpase is there fore to detect measure and record electro magnetic energy. However, in contrast to film emulsions, scanner detectors are available which are sensitive to wave- lenghts longer than 1 um, such as the thermal infrared and microwave regions of the electro magnetic spectrum. Airborne scanning systems can be broadly classified as being either passive, that is sensing naturally ocurring electro magnetic energy, or alternatively active that is, sensing system-generated electro magnetic energy. A type of non-photographic systems satallite-based systems. The development of satellite-based MSS systems for commercial rather than military use is a very recent phenomenon. Although a great deal of satellite photog raphy had been taken from space from the early manned space programes of the 1960s, it was not until the early 1970s that MSS imagery become available. Although some experimental imagery has been obtained from manned sa tellites, such as skylab in 1973, the majority of satel lite MSS imagery has been obtained from unmanned plat forms, in particular from the Landsat serries of satel lites. -IX- The Landsat Satellite System known as the Earth Resources Technology Satellite (ERTS), was initiated in 1967 by NASA in conjunction with the US Department of the Interior. Landsat-1 was launched in 1972 and was equipped with a RBV (Return Beam Vidicon) television camera system and a four waveband Multispectral Scannig System (MSS). The latter operates in two channels of the visible spectrum and two in the view by means of onboard detectors that converts this, electronically, on to mag netic tape (it can also be telemetred directy to ground receiving stations). Landsat 4 and 5 is called second generation Landsat series. Landsat 5, the second of the new generation Landsat satellites, was launched in March, 1984, following the failure early in 1983 of the Thematic Mapper sensor on board Landsat 4. The design of this second-ge neration series of satellites began in the early 1970s and has led to the development of a system which differs considerably from the first-generation Landsat satelli tes (number 1 to 3). By for the most significant change has been the introduction of a new sensor with improved spatial resolution, spectral seperation, geometric fide lity and radiometric accuracy known as the Thematic Map per (TM). In addition, and in common with the previous satellites, Landsat 5 also has a four-channel MSS with identical spatial and spectral characteristics to the previus Landsats ' MSS. In terms of basic design, the TM differs fundamen tally from the MSS in five ways. Firstly,. the TM obtains data on both the west to east and east to west scans, in contrast to the single-scan method emplayed on Landsat 1 to 3. This bidirectional approach was introduced in or der to further reduce the mirror scan rate. Secondly, the TM detector arrays are located within the primary focal plane of the instrument, allowing incoming light to be reflected directly onto the detectors without transmission trough the fibre optics emplayed in the MSS. By the introduction of this scanner design feature, it was hoped to minimize any loss in the intensity of the incoming radiation. Detector arrays for different spectral bands are spaced apart in the focal plane by a small amount; hence the same point on the ground is not simultaneously scanned in all spectral bands. Precise band-to-band registration is therefore an extremely im portant aspect of the subsequent data processing. A third distinguishing design feature of the TM is use of seven spectral bands in contrast to four for the MSS. The band designations, spectral ranges and principal applications as follows: -x- Band number Spectral range (um) Applications i 0.45 to 0.52 Water penetration 2 0. 52 to 0. 60 Measurement of visible green reflectance 3 0.63 to 0.69 Water pollution, vegetation discrimination 4 0.76 to 0.90 Delineation of water bodies 5 1, 55 to 1. 75 Water resources, 6 10.40 to 12.50 Thermal mapping 7 2.08 to 2.35 Geological map ping In addition to increase in the number of spectral bands, the TM also has an increased radiometric sensi- vity. The range of discrete radiometric levels which can be sensed had been increased from 64 to 256. Finally, the spatial resolution of the TM is much improved over the MSS. A pixel size of 30 m is used in all bands ex cept band 7 which has a pixel of 120 m. One consequence of the lower orbit of Landsat 5 has been significant change in the earth coverage cycle. Landsat 5 the orbital strip adjacent to the initial or bital strip is scanned after a time lapse of seven days. A further alteration is the reduction in the repeat co- varege period from 18 days to 16 days. Typical Landsat scene is the parallelogram shape of the final processed image. This is a consequence of the effect of the earth's rotation during the 25s interval required to image a 185 kmxl85 km Landsat scene. All imagery acquired by Landsat satellites is re corded and archived at the Earth Resources observation System (EROS) Data Center at Sioux Falls, South Dakota. After the scene has been imaged, the data is then either telemetered to one of the twelwe ground receiving stati ons in real time if the receiving station is within ran ge, or recorded on to two on-board tape recorders. The recorded data can then be transmitted to one of the ground receiving stations when the satellite passes wit hin range, the receiving range of the Landsat ground stations. Landsat scenes of Turkey are recorded at Faci- no in Italy and also at Kiruna in Sweden. Digital images in remote sensing consist of discre te picture elements, called pixels. Associated with each pixel is a number that is the average radiance (bright ness) of a relatively small area within a scene. The si ze of this area affects the scene. A number representing the average intensity of scene within the square is sto red. Thus the image is a vast matrix of numbers. These images are very often held on magnetic tape and in par ticular on computer-compatible tape (CCT). Each image -XI- including about 2340 scan lines and each line holds 300-3450 pixels. Multispectral images require special consideration because they contain separate images that may be treated logically as a single image. They are stored in one of tree interleaved formats: band- interleaved by pixel (BIP), band-interleaved by line (BID, or band sequen tial (BSQ). The density of bits along the tape is com monly 800, 1600, or 6250 bits/inch and determines the amount of data that can be stored on the tape. Image processing and image classifications techni ques are the main stages of remote sensing. Image pro cessing techniques that assist the analyst in the quali tative, i.e, visual, interpretation of images. Digital image classification techniques that assist the analyst in the quantative interpretation of images, both appro aches to extracting information from images. When the image analyst supervises the process by choosing the in formation catagories or classes he desires and then se lecting training areas that represent each catagory, the method is called "supervised analysis". An alternative to this supervised approach is unsupervised classifica tion which the analyst emplays a computer algorithm that locates naturally occuring concentrations of feature vectors from a heterogeneous sample of pixels. Supervised and unsupervised training thus comple ment each other ; the former imposes the analyst ' s know ledge of the area on the analysis to constrain the re sults, and the latter determines the inherent structure of the data, unconstrained by external knowledge about the area. A combination of the two techniques is often used to take advantage of the characteristics of each. Clustering is widely employed in remote sensing da ta analysis for unsupervised classification, primarily as part of the training process. Unsupervised classifi cation allows the analyst to utilize any natural struc ture which may be present in the data as an aid to app ropriately partitioning the feature space into regions corresponding to the classes. The Gaussian maximum-likelihood statistical classi fier is by for the classifier most commonly used in re mote sensing. The mean vectors and covariance matrices of classes required to compute the class-conditional density fonctions as part of the recognition process are estimated by supervised or unsupervised methods. The minimum distance classifier is the simplest of the statistical classifiers commanly used for remote sensing applications. Class mean are determined from training data by either supervised or unsupervised met hods. Each pixel is then assigned to the class with the nearest class mean. This classifier requires a moderate amount of training data, that is sufficient to compute reliable estimates of the multivariate class means, and -xii- the results it produces are not optimal if the class covariance matrices are not equal. Parallelpiped classifier is most often used for qu alitative or "quick look" scene analysis, particularly when a good interactive color image display system is available for training the classifier. In a supervised mode of operation, the analyst locates a few regions in the image representative of each of the apparent ground cover typs. To define the feature-space parallelpipeds corresponding to each ground cover class, the system de termines the maximum and minimum data values in each of the specified spectral bands for each of the selected regions. Before write to application of the study will men tion the water resources engineering in remote sensing. Water, a fundamental substance for sustaining life it self, is a key factor in maintaning agricultural produc tion, energy production, and other activities at optimum levels. Assessing the quantity and quality of water is main aspect of water resources assesment that is of gre at and growing importance as the population of the Earth increases, there by placing greater stress on existing water supplies. At present for using remote sensing in hydrologic studies and water resources, in the first category, simple qualitative observations are made. A visual ob servation or interpretation from a photo that water from factory effluent into a stream has a different color than stream water would be an example. In the second ca tegory of remote-sensing utilization, geometric from di mensions, patterns, geographic location, and distributi on are the types of information derived. A third cate gory is the use of remote sensing for direct estimation of a hydrologic paremeter through the development of correlation between the remotely sensed observation and a corresponding in-situ measurement technique. The data provided by remote sensing still represent new measurement techniques for much of hydrology and wa ter resources assessment. In the i960' s or even in the early 1970' s, most of the published data fit into the first category of remote-sensing utilization. However, in the 1970' s, many applications of remotely data fal ling in to the second category were developed. In addi tion, the literature also shows progress in the third category of remote sensing utilization. Several data sources in the United States, which indicate the kinds of remote-sensing data available, or potentially avai lable, for use in water resources assessment. Remotely sensed data can be used effectively moni tor various hydrologic aspects of lakes, rivers and wetlands. Differences in the appearance of surface water are readily apparent on aerial photographs and multi- spectral scanner imagery. -Xlft- Solar energy that is not speculary reflected is refracted downward at the water surface and is affected by absorption and scattering. Absorption and backscatte- ring are highly influenced by inorganic and organic substances within the water body and produce distinctive spectral signatures. The recorded remotely-sensed signal is that part of the backscattered energy that returns to the water surface. If a water body is relatively clear and shallow, solar energy is reflected from the bottom and can be detected. The water depth that permits detec tion of the bottom depends on water color, turbidity, bottom reflactance characteristics, and the intensity of incident light. Penetration of radiation in to pure water is described by means of an extinction coefficient, k, which takes into account the effects of both scattering and absorption. A paralJel (unscattered) beam of radiation of wavelength ( >. ) passing through water a distance, dx, is reduced in intensity by an amount of dl, which is proportional to the intensity and to the distance, dx or dl= -kldx where k has dimensions of L-i (L is length and is depen dent on wavelenghth (%)). If the intensity of the radi ation at x= 0 is Io, then for some distance, x, 1=1 e-**4 (Figure 2.7) shows the values shown in this figure, it can be seen that the reduction in intensity of sunlight after passing through very thin layers of water will be quite high. Theories incorporating the concept of the extinc tion coefficient suggest that the basic color of water of an infinitely deep pure-water body would be blue. The inroduction of suspended organic and inorganic materials introduces further scattering, absorbtion, and reflecti on by these particles. Scattering by small particles is wavelenghth-selective and Rayleigh's law applies. Small wavelenghts are scattered most, and the basic blue color of the water remains re]atively unchanged. As particle size increases, scattering becomes independent of wave length so that a color shift of water toward the green begins to occur. With increasing turbidity, water color shifts even more toward yellowish greenK until the color of the water approaches or becomes that of the natural color of the particles creating the turbidity. Discolaration of the water can also occur as a re sult of the presence of life forms, such as algae. The spectral responce of the predominant life form wiJl ba sically determine the water color. In shallow waters, the natural color of the bottom can be detected, and wa ter color will be altered toward bottom color. Theoretical considerations involving the concept of extinction coefficient suggest that the best depth penetration of water in the visible and near infrared range could be made by data obtained near 0.5 um. In more turbit waters or in waters containing substance such as dissolved organic materials and plankton, the -XIV- peak wavelenght of light transraittance shifts toward the longer wavelengths. Remote sensing imagery, in general, beyond estimating the area covered by water would be to assess conditions within the water body as expressed in turbidity or changes. Pollutants (both inorganic and or ganic constituents) may effect the reflective or emmis- sive properties of water bodies and thus become detec table by multispectral scanning instruments. Both quali- tive and quantitative estimates of water quality may be derived from satellite data. When qualitative or relative estimates of water qu ality are being made, it is adequate to simply detect and delineate the area of pollution. Relative turbidity expressed as tonal change, is easily detected on Landsat TM images. Due to the frequent success of measuring sus pended sediment in both fresh and marine waters using Landsat MSS and TM data, it is widely held that broad intervals of water turbidity can be discriminated with Landsat data. In wiew of fact that remote quantitative measure ment of turbidity is considered operational by some in vestigators (the same cannot yet be said for types of sediment and/or phytoplankton species), it is surprising that remote sensing of turbidity is not in wider use. Remote sensing for infrared plume analysis is not fully operational (in part because it does not yield a profile by depth), it has contributed to the study of plumes and has yielded useful data. For example, a prohibited waste head discharge was found at one site during constructi on. Current studies focus on using digital infrared line scanner data for waste-head dispersion models, including the study of plume behavior in winter. To rationalize the aim of this study defination at macro level water pollution which was to interpret the Landsat-5 TM data belonging to the Old Istanbul which is including Bosforus, Golden Horn Estuary and part of Marmara Sea. In this study computer compatible magnetic tape (CCT) wich held the image values for the resource regi on. Landsat-5 TM data dated 12. June. 1984 and 24 October 1986 were evaluated to water resources classifications by means of computer-assisted techniques. While the analyses of TM data were accomplished by the collection of hydrological archived data. Training sites were chosen within the study area which was covered both by TM digital data and actual wa ter depth map, hydrological data. Detailed water resour ces were carried out to determine actual water data. For this purpose, training sites were carefully checked in the water, then their characteristics were transfered to the hydrological data. Different classification techni ques applied at study area. Finally showed that remote sensing imagery within the water body can determine as expressed at macro level turbidity or changes.