ارزیابی و مقایسه دقت منطقه ای محصولات جهانی سنجش از دوری در ایران، مطالعه موردی: محصولات پوشش اراضی در استان مازندران (مقاله علمی وزارت علوم)
درجه علمی: نشریه علمی (وزارت علوم)
آرشیو
چکیده
سنجش از دور به عنوان یکی از مهم ترین منابع داده مکانی در عصر حاضر محسوب می شود که روز به روز شاهد توسعه آن در ابعاد مختلف هستیم. انتشار محصولات (پروداکت های) جهانی این داده ها در سال های اخیر با هدف دسترسی و استفاده راحت تر متخصصان علوم مکانی یکی از ابعاد این توسعه محسوب می شود. محصول پوشش اراضی یکی از این محصولات محسوب می شود که سهم بیشتری در استفاده نسبت به سایر محصولات سنجش از دوری دارد. هنگام ارائه این محصولات ویژگی های کیفی و کمّی آن ها از جمله دقت جهانی آن ها نیز منتشر می شود. بیان دقت این محصولات به صورت جهانی، ارزیابی مجدد دقت آن ها را به صورت منطقه ای برای کاربران این محصولات در مناطق مختلف دنیا لازم و ضروری می نماید. در این تحقیق پوشش اراضی سرویس جهانی اراضی برنامه کوپرنیک آژانس فضایی اروپا (CGLS)، محصول پوشش اراضی GlobeLand30 و محصول پوشش اراضی Esri از جمله محصولات جهانی پوشش اراضی هستند که ضمن مقایسه ویژگی های اسمی، از لحاظ کمّی برای استفاده به صورت منطقه ای در کشور (استان مازندران) مورد بررسی و ارزیابی قرار گرفتند. نتایج، دقت منطقه ای پوشش اراضی CGLS، GlobeLand30 و Esri را به ترتیب برابر با 84، 81 و 75 درصد نسبت به دقت جهانی آن ها (80، 83 تا 85 و 86 درصد) نشان می دهند. در ارزیابی دقت منطقه ای کلاس ها، هر سه محصول مورد مطالعه دقتی بالای 90 درصد در کلاس های برف و یخ، جنگل، پهنه های آبی و ساخت وساز انسانی داشته اند. برای کلاس اراضی کشاورزی دقتی برابر با 92، 69 و 84 درصد برای پوشش های اراضی CGLS، GlobeLand30 و Esri به دست آمد. در سه کلاس بوته زار، پوشش علفی و تالاب، نتایج دقتی کمتری را نسبت به سایر کلاس ها برای هر سه محصول پوشش اراضی نشان می دهد.Evaluation and comparison of regional accuracy of global remote sensing products in Iran - Case study of land cover products in Mazandaran Province
Extended IntroductionRemote sensing is considered as the most important source of spatial data in the current era, which we witness its increasing development in different dimensions. The release of global products of these data in recent years with the aim of easier access and use by experts in geospatial science is one of the dimensions of this development. The land cover product is one of these products that is used more than other remote sensing products. When presenting these products, their qualitative and quantitative characteristics, including their global accuracy, are also published. Expressing the accuracy of these products globally makes it necessary and necessary to re-evaluate their accuracy regionally for the users of these products in different regions of the world.Materials & MethodsIn this research, the accuracy of the European Space Agency's Copernicus Global Land Service (CGLS), GlobeLand30 and Esri's land cover product were evaluated for regional use in the north of Iran - Mazandaran province. After calculating the area of the classes for each of the land cover products, Pearson's correlation coefficient was used to calculate the correlation between them. For quantitative evaluation, the error matrix was used as one of the most common ways to evaluate the accuracy of land cover products. This method is based on the comparison of classified data and ground reality data. Also, the categorized random sampling method was used to select 1329 evaluation samples in Mazandaran province. For visual evaluation, three areas with dimensions of 6 x 6 km were selected.Results & DiscussionThe regional accuracy evaluation of the studied products shows opposite results compared to the global accuracy of these products. Based on the global accuracy reported for the studied products, the highest accuracy is calculated for the Esri product at 86%, followed by GlobeLand30 and CGLS at 83-85 and 80%. Meanwhile, based on the regional accuracy obtained from the results of this research, the highest regional accuracy for the CGLS product has been calculated at 84% and then for GlobeLand30 and Esri products at 81 and 75%. In evaluating the regional accuracy of the classes, all three studied products (CGLS, GlobeLand30 and Esri) have acceptable accuracy (above 90%) in the classes of snow and ice (100, 100 and 100%), forest (90, 95 and 98 percent), water (96, 94 and 90 percent) and impervious surface (94, 91 and 90 percent). For the agricultural class, accuracy equal to 92, 69 and 84% was obtained for CGLS, GlobeLand30 and Esri land covers.In the 3 classes of shrubland, Impervious surface and wetland, the accuracy results are less than other classes for all three land cover products and in the amount of (29, 0 and 13 percent), (65, 66 and 42 percent) and (67, 38 and 0 percent).Conclusion By evaluating and comparing the regional accuracy of three CGLS products, GlobeLand30 and Esri, this research answered the question of whether the accuracy stated in global land cover products can be trusted for regional studies and planning. The results show that the regional accuracy of CGLS, GlobeLand30, and Esri are 84, 81, and 75 percent, respectively, compared to their global accuracy (80, 83, 85, and 86 percent). These results show the difference obtained for the Esri product more than the two products CGLS and GlobeLand30. Meanwhile, the remote sensing data used for the Esri product (Sentinel-2 data) and its pixel size (10 meters) are of higher quality and quantity than the other two products. In fact, these results show that only paying attention to the type of data used and the global accuracy is not enough to use products in regional scales and requires evaluations before using them.In addition, by evaluating the classes of each product and comparing them, the need for this evaluation before using these products seems necessary. The results showed that in the evaluation of the regional accuracy of the classes, all three studied products had an accuracy of over 90% in the classes of snow and ice, forest, water areas and human construction. For the agricultural land class, accuracy equal to 92, 69 and 84% was obtained for CGLS, GlobeLand30 and Esri land covers. In the 3 classes of shrubland, herbaceous cover and wetland, the results show lower accuracy than other classes for all three land cover products. Significant results were also obtained in the visual evaluation, and it seems necessary to pay attention to this evaluation before the applications where it is important to pay attention to a particular class.
تبلیغات
