ارزيابي دقت آزمايي مدل هاي رقومي ارتفاعي(DEM)ماهواره هاي ASTER و SRTM با مشاهدات دقيق زميني(DGPS) مطالعه موردي: از سد آزاد به دشت قروه-دهگلان، سنندج

Accuracy assessment of Digital Elevation Models (DEM) of SRTM and ASTER satellites with precise ground observations (DGPS) Case study: from the Azad dam to the plain of Ghorveh-Dehgolan, Sanandaj

مدل‌هاي رقومي ارتفاعي براي بسياري از اهداف، مهم بوده و در بسياري از كاربردها و مطالعات جزء الزامات اوليه مي‌باشند. هدف اين مقاله بررسي ميزان دقت و صحت مدل‌هاي رقومي ارتفاعي حاصل از تصاوير ماهواره ASTER و داده‌هايSRTM با ابعاد پيكسل 30 و 90 متر و همچنين مدل رقومي ارتفاعي به دست آمده از نقشه‌هاي توپوگرافي 1:25000 با مشاهدات دقيق زميني (DGPS) در لندفرم‌هاي مختلف شامل دشت، تپه‌ماهور و كوهستان مي‌باشد. ميزان انطباق اين داده‌ها با استفاده از تحليل همبستگي پيرسون آزمون شد. دقت و صحت مدل‌هاي رقومي ارتفاعي مختلف مورد بررسي با استفاده ازRMSE، خطاي ميانگين و انحراف استاندارد بررسي شد. براساس نتايج ضريب تعيين رابطه داده‌هاي زميني با مدل‌هاي رقومي ارتفاعي بين 97 تا 99 بود. بيشترين انطباق مربوط به مدل رقومي مستخرج از داده‌هاي توپوگرافي 1:25000 و مدل رقومي ASTER30 متر و كمترين انطباق مربوط به داده‌هاي SRTM90 متر بود. در مجموع با دشوارتر شدن شرايط عرصه يعني از دشت به كوهستان، انطباق مدل‌هاي رقومي ارتفاعي با داده‌هاي زميني برداشت شده كاهش مي‌يافت. نتايج بررسي صحت و دقت مدل‌هاي رقومي نشان داد كه كمترين خطا در وهله اول مربوط به مدل رقومي ارتفاعي استخراج شده از خطوط‌ميزان نقشه 1:25000 (6/27=RMSE ) و پس از آن مدل رقومي ارتفاعي ASTER30 متر (7/43=RMSE ) است. همواره اندازه پيكسل 30 متر نتايج بهتري نسبت به پيكسل 90 متر داشته است. بر اساس معيار خطاي ميانگين، كمترين اريبي مربوط به ASTER30 متر (2 متر اريبي) و پس از آن مربوط به مدل رقومي 1:25000 (17/2) است. بيشترين اريبي مربوط به مدل‌هاي 30 و 90 متري استخراج شده از داده‌هاي SRTM بود. نتايج خطاي انحراف استاندارد منطبق بر نتايج RMSE بود كه تأييد كننده بهتر بودن مدل‌هاي رقومي ارتفاعي مستخرج از داده‌هاي توپوگرافي 1:25000 و ASTER30 متر بود.

Digital elevation model (DEM) is the raster representation of the ground surface so that the information of each cell on the image has a value equal to the altitude from the sea level corresponding to the same spot on the ground. DEM is an appropriate tool for the generation of topographic maps and contour lines, access to the information of surface roughness, three dimensional vision, etc. (Jacobsen, 2004). The accuracy of the digital elevation model is effective on the accuracy of the information from which it is obtained. This is why researchers are always looking for a way to increase the accuracy of digital elevation models. Among the information resources that are used to generate this model are ground mapping, aerial photography, satellite images, radar data, and Lidar. Some of these data generate the digital elevation model with little accuracy due to the insufficiency of the elevation information. The aim of this paper is to investigate the accuracy of DEMs derived from ASTER satellite images and SRTM data with 30 and 90-meter pixel dimensions and the digital elevation model derived from the topographic 1:25000-scale maps with Differential Global Positioning System (DGPS) in different landforms including plains, hills and mountains. Materials and Methods The study area is a part of the project of dam and water transfer system from the Azad dam to the plain of Ghorve-Dehgolan (with the goal of transferring water from the catchments of Sirvan River into the country) in the province of Kurdistan and the city of Sanandaj. In this study, the Real-Time kinematic method (RTK) was used to locate the points. In this method, assuming that the coordinates of the reference station are known and comparing it with the location obtained from the GPS receiver, a correction value is obtained that is applied to the coordinates obtained for the Rover Station, which is known as the relative or differential method. In this method, the corrections are calculated asreal-time during the observations and are considered in the determination of the Rover location.The Leica GS10 GNSS receivers were used in this study. First, two reference stations were determined using the Fast Static method and then, the Real-Time kinematic (RTK) method was used. In order to investigate the extent of the data compliance and relation, the Pearson linear correlation analysis was used and the accuracy assessment of the extracted digital elevation models was carried out using the RMSE, mean error and standard deviation. Results & Discussion The statistical parameters such as root mean square error (RMSE), bias (µ) and standard deviation () were used to assess the accuracy of each one of the investigated digital models. By comparing different sources that create DEMs, it can be seen that the minimum error is first related to the digital elevation model extracted from the contour lines of the 1:25000-scale map (27/6 = RMSE) and then to the ASTER digital elevation model with the pixel size of 30 meters (RMSE=7.43). The 30-meter pixel size DEM has always led to better results than the 90- meter pixel size DEM. Based on the mean error standard, the minimum bias is related to ASTER30 m (bias of 2 m) and then to the 1: 25,000 DEM (2.17). The maximum bias was related to 30-and 90-meter models extracted from the SRTM data. The results of standard deviation error were in compliance with the RMSE results, which confirmed the superiority of 1:25000-scale map and ASTER30 m DEMs. The results showed that the determination coefficient of relationship between the ground data and digital elevation models is between 97 and 99. The maximum compliance is related to the digital elevation model extracted from the 1:25000-scale topographic data and the ASTER30 m DEM, while the minimum compliance is related to the SRTM90 m data. In general, the compliance of the digital elevation models with the ground data decreased as the field's conditions became more difficult, i.e. from plain to mountain. Conclusion The results of DEMs accuracy assessment showed that the minimum error was primarily related to 1:25000 contour lines DEM (RMSE=6.27) and then, to the ASTER30 m DEM (RMSE=7.43). The pixel size of 30 meters has always been better than the pixels size of 90 meters. Based on the mean error standard, the minimum bias is related to the ASTER 30 m (bias of 2 m) and then, to the 1: 25,000 DEM (2.17). The maximum bias was related to 30-and 90-meter models extracted from the SRTM data. The results of the standard deviation error were consistent with the RMSE results, which confirmed the superiority of the digital elevation models extracted from the topographic 1:25000-scale maps and the ASTER30 m DEM.