مقايسه روش‌هاي برآورد منطقه‌اي دماي كمينه هوا (مطالعه موردي: استان اصفهان)

Comparison of different regional estimation methods for daily minimum temperature (A case study of Isfahan province)

در اين پژوهش ارزيابي و مقايسه ميزان خطاي مدل‌هاي مختلف در تخمين درون‌يابي منطقه‌اي دماي كمينه روزانه در استان اصفهان در دو سال مختلف با تعداد ايستگاه هاي متفاوت صورت گرفته است. سال 1992 به عنوان سال با تعداد ايستگاه كم (19 ايستگاه همديدي و 11 ايستگاه اقليم‌شناسي) و سال 2007 به‌عنوان سالي با تعداد ايستگاههاي بيش‌تر (31 ايستگاه همديدي و 23 ايستگاه اقليمشناسي) انتخاب شدند. از هر ماه اين دو سال به‌طور تصادفي دو روز (هر سال 24روز) انتخاب و عمليات درون‌يابي روي داده‌هاي منتخب انجام شد. براي منطقه‌اي نمودن داده‌هاي نقطه‌اي دماي كمينه روزانه شش روش درون‌يابي به كار گرفته شد. نتايج نشان مي‌دهند كه در سال1992 روش وايازي چندگانه خطي (ريشه ميانگين مربعات خطا از 33/2 تا 12/5 و ضريب همبستگي از 37/0 تا 85/0) و در سال 2007 روش كريجينگ- وايازي و روش وايازي چندگانه خطي (ريشه ميانگين مربعات خطا از 36/2 تا 5 و ضريب همبستگي از 38/0 تا 83/0) داراي كم‌ترين خطا و بيش‌ترين ضريب همبستگي بوده و روش هاي كريجينگ، كوكريجينگ، وزني عكس فاصله و اسپيلاين در مراتب بعدي قرار گرفتند. افزايش تعداد ايستگاه سبب بهبود دقت روش هاي درون يابي شده است.

The main objective of this study is to evaluate different interpolation methods for estimation of regional minimum temperature in Isfahan province, Iran. In order to explore the effect of stations number on the accuracy of the interpolation methods, two years 1992 and 2007 with different number of meteorological stations have been selected. The daily minimum temperature (Tmin) data of 30 meteorological stations (17 synoptic and 13 climatological stations) for year 1992 year and 54 meteorological stations (31 synoptic and 23 climatological stations) for 2007 year were collected from Isfahan and neighboring provinces. In order to regionalize the point data of Tmin, several interpolation methods, including inverse distance weighted (IDW), Kriging, Co-Kriging, Kriging-Regression, Multiple Regression and Spline were worked out. To evaluate the performance of these methods, 2 days from each month, i.e. total 24 days for both years were chosen randomly. The obtained results were compared using statistical measures including: RMSE, MBE, MAE and correlation coefficient (r). The findings revealed that the application of multiple regression method for interpolation produced the least error in estimation of minimum temperature in 1992 (with RMSE ranging from 2.33 to 5.12 and r from 0.38 to 0.85). For 2007 year, the best estimation was achieved by multiple regression and Kriging-Regression (RMSE from 2.36 to 5 and r ranging from 0.38 to 0.83) respectively. The overall performance of Kriging, Co-Kriging, IDW, and Spline methods was also acceptable and they were in next ranks respectively. In general, with increasing number of study the overall accuracy of model performance in estimation of daily minimum temperature has been improved.