بررسي روش هاي تركيبي زمين آمار در افزايش دقت طبقه بندي اقليمي و نيز پهنه بندي عناصر اقليمي شمال شرق ايران

An investigation into using of combined geostatistical methods to increase precision in climatological classification and climatic parameters zoning in great Khorasan

متغير هاي اقليمي به يكديگر و نيز به وضعيت سطح زمين مانند ارتفاع و پوشش گياهي وابسته اند. اين در حالي است كه اين متغير ها به صورت نقطه اي در ايستگاه هاي هواشناسي اندازه گيري مي شوند. براي انجام مطالعات محيطي و تحقيقات كشاورزي، داشتن درك صحيحي از تغييرات پيوسته مكاني و زماني اين متغير ها از اهميت بسزايي برخوردار است. از طرفي طبقه بندي هاي اقليمي به دليل استفاده از روابط ساده و متغير هاي كم از دقت بالايي برخوردار نيستند، از اين رو در اين تحقيق دو هدف دنبال شده است: اول اينكه از الگوريتم هاي زمين آمار براي درون يابي، ارزيابي و تهيه نقشه هاي تغييرات مكاني و زماني متغير هاي اقليمي در شمال شرق ايران استفاده شد و آنگاه روش تورنت وايت براي طبقه بندي اقليمي انتخاب و درجه تاثير هر متغير اقليمي در افزايش دقت طبقه بندي اقليمي با استفاده از روش هاي چند متغيره بررسي شد. روش ها ي درون يابي در اين تحقيق كريجينگ معمولي (OK) ، كو كريجنگ (COK)، روش وزن دهي عكس فاصله (IDW) و روش (KED) بود. با استفاده از روش هاي چند متغيره (COK,KED)، وابستگي متغير هايي مانند (تبخير، دماي هوا، بارندگي و رطوبت نسبي) به ارتفاع به عنوان متغير ثانويه با گام هاي زماني ماهانه و سالانه مورد بررسي قرار گرفت. مقدار MSE براي مقايسه نتايج مدل ها استفاده شد و نتايج متفاوتي براي هر متغير به دست آمد. روش COK براي دماي هوا نتايج بهتري را نشان داد، در حالي كه روش KED براي بارندگي نتايج دقيق تري را حاصل كرد. به عنوان مثال MSE براي براي دما از روش هاي K، COK و KED در ماه ژانويه به ترتيب مقادير 2/19، 0/004 و 1 ، در ماه فوريه 2/63، 0/005 و 1/27 و در ماه مارس 2/51، 0/004 و 1/33 به دست آمد. همچنين نتايج نشان داد كه مقادير MSE از ماه مارس تا جولاي افزايش مي يابد، بدين معني كه استفاده از ارتفاع در اين مدل براي تخمين دما در اين ماه ها دقت كمتري دارد. همچنين مشاهده شد كه توزيع زماني و مكاني بارندگي نسبت به ساير متغير هاي مورد مطالعه، بيشترين تاثير پذيري را از تغييرات ارتفاع دارد. قابل ذكر است كه بر اساس اين تحقيق تبخير در طول ماه هاي سرد از ارتفاع تاثير مي پذيرد( اكتبر تا مارس). و از ميان متغير هاي محيطي به ترتيب تبخير، ارتفاع، رطوبت نسبي و بارندگي در تغيير پذيري زماني و مكاني اقليم در منطقه مورد مطالعه بيشترين تاثير را دارند. دما نتايج متفاوتي بسته به شاخص اقليمي مورد استفاده براي پهنه بندي اقليمي حاصل كرد.

Introduction: Climatic parameters’ modeling is very important in environmental data processing. This is a consequence that climatic parameters vary dramatically in time and space. Moreover, the climate variables are dependent to each other and also to earth surface conditions such as height. The other problem is that climatic parameters are measured as a point based variables in weather stations. However, for environmental studies it is crucial to have continues spatial and temporal perception for these parameters. There are different methods to provide such perceptions from climatic variables. Some geostatistical models are used to interpolate the data. The ability of these models for spatial interpolation increases significantly, if covariables are used (Daly et al. 1994). In Kriging methods the sparsely sled variables can be completed by secondary attributes that are more densely sled. Topography and weatherradar observations could be used as secondary information in these models. Material and methods: The study area is Khorasan province (Northeast of Iran, longitude 55◦W to 61◦E and latitude 38◦S to 30◦N). The area is approximately 248,000 km2 in a semiarid climate.  The monthly and annual precipitation has been averaged for the climate normal period of 1993 – 2009. We were very strict in data selection, only keeping weather stations with complete years. After assuring the raw data quality, monthly and annual climate data averages were calculated. This information was loaded to the spatial database and used as the source of input data for the gridding process. We used geostatistic algorithms for assessment, interpolation and preparing spatial and temporal maps for climatic parameters in North East of Iran. Different interpolation methods including ordinary Kriging (OK), Inverse Distance Weighted (IDW), CoKriging (COK) and Kriging with External Drift (KED) were examined. The dependence of the variables (including solar radiation, evaporation, air temperature and precipitation) to height as ancillary variable was also investigated in different monthly and annual time scales. Thornthwaite climate classification method was used for climate zoning. Then the effect order of each climatic variable in the climate zoning precision was assessed by using multivariate methods such as COK and KED. Mean Squared Error (MSE) was used to compare the models results. Different results were obtained for different variables. Results and discussion: According to MSE values, COK and IDW had the highest and lowest accuracy among the methods for temperature respectively. The pattern of MSE changes were also similar for all of the four methods when MSE values increased from January to June showing that the accuracy of the models decreased from cold to warm season. OK and IDW showed more errors in the warm months than in cold months for precipitation, while KED and COK with elevation as ancillary variable showed better results. Considering all of the variables, KED provided the most accurate spatial interpolation among all of the applied models. However, COK was more accurate for evapotranspiration interpolation with the minimum MSE with increasing toward warm months as other variables. There is only one exception in applying COK method for evapotranspiration and temperature where MSE is almost the same in cold and warm seasons. For interpolating of relative humidity, there was not a substantial difference between K and KED, while COK and IDW showed smaller values of MSE. In this case MSE values decreased from clod to warm months. All of the four interpolation methods were used for climatological zoning based on Thornthwaite Climatological Index values. MSE values decreased in order of IDW, K, KED and OCK respectively. Using meteorological parameters such as temperature and evapotranspiration as ancillary variables in multivariable methods such as COK and KED showed a substantial improvement in the accuracy of climatological zoning. COK model provided better results for air temperature, while KED method showed more precision for precipitation. For exle the resulted MSE from K, COK and KED methods for temperature in January was 2.19, 0.004 and 1, in February was, 2.63, 0.005 and 1.27 in March was 2.51, 0.004 and 1.33 respectively. The results also showed that MSE values substantially increased from March to July which means that using elevation in this model for estimating temperature during these months provides less precision. Conclusion: It was concluded that temporal and spatial distribution of precipitation is affected more by elevation among all of the climatic parameters, followed by air temperature, evaporation and relative humidity respectively. It should be noticed that evaporation is affected by elevation during cold season (from October to March). Among the environmental parameters, evaporation, elevation, relative humidity and precipitation had the most effect on spatial and temporal climate variability in the area of study respectively. Temperature provided different results depending on the climate index that was used for classification and zoning.