تعیین LST در تصاویر سنجش از دور و افزایش دقت آن با استفاده از ادغام الگوریتم های مختلف و روش های تصمیم گیری چند معیاره (مقاله علمی وزارت علوم)
درجه علمی: نشریه علمی (وزارت علوم)
آرشیو
چکیده
-چکیده:دمای سطح زمین (LST) یکی از معیارهای مهم در برنامه های کاربردی است و پایش دقیق زمانی و مکانی آن جهت مطالعات محیطی و مدیریت و برنامه ریزی امری ضروری محسوب می شود. با توجه به محدودیت هایی که در ایستگاه های هواشناسی برای تعیین این پارامتر ضروری وجود دارد، به کمک الگوریتم های مختلف و به کمک سنجنده های حاوی باندهای مادون قرمز حرارتی، می توان این پارامتر را در سطح گسترده ای تعیین کرد. دقت الگوریتم های مختلف تعیین دمای سطح زمین با استفاده از تصاویر سنجش از دور، در مناطق مختلف و با استفاده از سنجنده های مختلف تغییر می کند و تاکنون الگوریتم مشخصی با دقت بالا برای تمام مناطق در نظر گرفته نشده است. در این مقاله هدف تعیین دمای سطح زمین با استفاده از الگوریتم های تک کاناله، پنجره مجزا، پلانک، تک پنجره و معادله انتقال تابشی و همچنین استفاده از روش ادغام الگوریتم های تعیین LST به صورت وزن دار و ساده است. در روش وزن دار، وزن هر روش به کمک الگوریتم های تصمیم گیری چندمعیاره TOPSIS و SAW مشخص شده است. همزمان با عبور ماهواره لندست8 از منطقه مورد مطالعه، دمای سطح زمین برای 25 نقطه برداشت شده است. برای ارزیابی عملکرد روش پیشنهادی ادغام الگوریتم های تعیین LST، از معیار آماری ریشه میانگین مربعات خطا (RMSE) استفاده شده است تا مقایسه ای بین برداشت های زمینی و مقادیر محاسبه شده به وسیله الگوریتم ها انجام شود. نتایج نشان می دهد روش ادغام الگوریتم ها که ضریب هر الگوریتم با استفاده از روش تصمیم گیری چند معیاره TOPSIS محاسبه شده است دارای بیشترین دقت است(RMSE=0.552oK). با استفاده از این الگوریتم ترکیبی، به روش های دارای دقت بیشتر وزن بیشتری تعلق می گیرد. از بین پنج الگوریتم به طور مجزا، الگوریتم تک کاناله دارای بیسترین دقت (RMSE=0.5623oK) و الگوریتم تک پنجره دارای کم ترین دقت می باشد (RMSE=1.0046ok).Determining LST in Remote Sensing Images and Increasing its Accuracy Using the Fusion of Different Algorithms and Multi-Criteria Decision-Making Methods
Land surface temperature (LST) is one of the important criteria in applications, and its accurate time and place monitoring is considered essential for environmental studies and management as well as planning. Considering the limitations that exist in meteorological stations to determine this necessary parameter, with the help of different algorithms and sensors containing thermal infrared bands, this parameter can be determined on a wide scale. The accuracy of different algorithms for determining the LST using remote sensing images varies in different regions, using different sensors, and so far, no specific algorithm with high accuracy has been considered for all regions.In this article, the aim is to determine the temperature of the LST by using Single channel, Split window, Planck, Mono Window and RTE algorithms, as well as using the fusion method of LST determination algorithms in a weighted and simple way. In the weighted method, the weight of each method is determined with the help of TOPSIS and SAW multi-criteria decision-making algorithms.At the same time as the Landsat 8 satellite passes through the study area, the LST is taken for 25 points. To evaluate the performance of the proposed method of fusion of the LST determination algorithms, the root mean square error (RMSE) statistical criterion is used to make a comparison between the ground measurements and the values calculated by the algorithms. The results show that the algorithm fusion method, where the coefficient of each algorithm is calculated using the TOPSIS multi-criteria decision-making method, has the highest accuracy (RMSE=0.552oK). By using this combined algorithm, more weight is given to more accurate methods. Among the five algorithms separately, the single-channel algorithm has the best accuracy (RMSE=0.5623oK) and the single-window algorithm has the lowest accuracy (RMSE=1.0046ok).Extended IntroductionLand surface temperature (LST) is an important parameter related to surface energy; as a result, researchers intend to find an accurate Algorithm to estimate it. In addition, researchers in recent decades have used various methods to determine LST, including the methods of Split window algorithm, Single channel algorithm, Mono window algorithm, Radiative transfer equation and Planck equation and etc. The performance of these methods is different when compared to each other. While Radiative transfer equation and Single channel algorithm are more sensitive to atmospheric water vapor content than Mono window algorithm and Split window algorithm especially in hot and humid conditions, Split window algorithm has the highest sensitivity to Land surface emissivity and does not require accurate atmospheric profiles. On the other hand, high quality atmospheric transmission/radiation is required in accurate determination of LST using Single channel algorithm and is not required in Mono window algorithm. Satellite-based LST retrieval is still a challenging process due to the great variability of the Earth surfaces and a priori knowledge about input parameters such as the atmospheric transmittance, LSE, the meteorological conditions, and the sensor specifications are necessary. Data fusion methods can be used to take advantage of multiple data. In this method, fused data will be produced in which more comprehensive information can be obtained. This study focuses on developing two reliable multi-criteria decision analysis method to LST determination, based on the weighted average of several LST determination approaches called TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution) and SAW (Simple Additive Weighting) methods. The main idea of these techniques, the preferred alternative which is the closest to the positive ideal solution and the farthest to the negative ideal solution. During 2019, data collection experiments were performed on the Iranian plateau in semi-arid region of Marvdasht (N 29°49′0.26"_29°50′44.5", E 52°42′51"_52°46′14") located in Fars province. On November 13, 2019, coinciding with the passage of the Landsat8 satellite through the region, the surface temperature of 25 points was measured by a mercury thermometer and the position of those lands was recorded by a manual GPS devise. These points were used in this research to determine the RMSE of different methods of LST retrieval.In this study, an attempt has been made to estimate the LST, by using Landsat-8 TIRS, OLI satellite data. MethodologyDifferent algorithms have been defined to determine LST. Each of these algorithms has advantages and disadvantages, and one method is not considered as the most accurate algorithm for determining LST in all regions. Therefore, in this article, in addition to the five methods of LST determination, simple average and weighted average methods are also used. The weight of each method in the weighted average has been calculated using multi criteria decision making TOPSIS and SAW methods. Results and Discussion In this research, several methods in LST determination have been used. Due to the calculation of RMSE in the desired area, the most accurate algorithm among these algorithms is Single channel algorithm. The amount of RMSE in the Weighted mean method by TOPSIS approach has the lowest number of 0.552oK which is more accurate than the most precise methods in this region and image (Single channel algrithm). The least accurate method of retrieval of the LST in this region is Mono window algorithm, and the Weighted mean method by TOPSIS approach is 0.450oK more accurate than this method. By using the weighted mean, more weight is granted to the less-error methods. ConclusionEstimation of LST is an important topic of research. Because, these days, global climate is changing fast. Therefore, it is vital to investigate ways to predict change in LST change. In the study area, among the 5 algorithms of determining LST, Single channel algorithm has the most accuracy, and the weighted average method, which calculates the weight of the methods or the use of TOPSIS method, has the most accuracy.In a weighted mean methods, the weight of each method is calculated by using the TOPSIS & SAW methods and entering the RMSE of different LST determination algorithms (results of comparing the output of each algorithm with ground measurements) and simple mean is obtained from the average of the methods.
تبلیغات
