پديد آورندگان :
فرزندي، محبوبه نويسنده دانشگاه فردوسي مشهد,ايران farzandi, mahbobeh , رضائي پژند، حجت نويسنده دانشگاه آزاد اسلامي واحد مشهد,ايران rezaee, hojat , ثنايي نژاد، حسين نويسنده دانشكده كشاورزي,گروه مهندسي آب,دانشگاه فردوسي مشهد,ايران Sanaeinejad, H.
كليدواژه :
رگرسيون خطي و غيرخطي , آسيب شناسي , روند , همگني , 127 سال دماي ماهانه مشهد
چكيده فارسي :
طول دوره آماربرداري اهميت زيادي در تحليل و پيش بيني دماي سالانه دارد. آمار طولاني مدت مي تواند نوسانات سري زماني دما شامل روند، تغييرات فصلي و تغييرات دوره اي را بهتر نمايان كند يا دقت تحليل فراواني را افزايش دهد. حداكثر طول دوره آماربرداري دماي هوا در ايستگاه هاي كشور حدود 60 سال است كه آمار كوتاه مدتي محسوب مي شود. اين آمار نمي تواند روند خطي يا غير خطي و نوسانات دوره اي را به خوبي نشان دهد. مشهد داراي 127سال آمار بلندمدت دماي ماهانه است (2011 -1885). 60 سال اخير آن توسط ايستگاه همديد مشهد (2011-1951) و 67 سال باقي مانده (1940-1885) توسط كنسولگري آمريكا واقع در مشهد آماربرداري شده است. اين آمار داراي ماه هاي مفقود است (حدود 20%) كه بايد ترميم شود. هدف اين مقاله تكميل آمار 127 سالانه دماي ماهانه مشهد است. چند ايستگاه با آمار طولاني مدت وجود دارد كه همبستگي خوبي با دماي مشهد دارند كه شامل جاسك (ايران، تاسيس 1893)، سرخس و كيزيل (تركمنستان، تاسيس به ترتيب 1900 و 1893)، تركستان (قزاقستان، تاسيس 1885) و بغداد (عراق، تاسيس 1893) است. الگوهاي خطي و غيرخطي چندگانه متنوعي براي ترميم آمارهاي مفقود به كار رفت. سه الگوي خطي در نهايت پذيرفته و استفاده شد. ضريب تعيين اين الگوهابه ترتيب 0/98، 0/96و 0/93 كه نشان از قدرت خوب آنهاست. آزمون هاي كلاسيك و آسيب شناسي الگوها نشان از قبول آنهاست. آماره F براي اين سه به ترتيب 23160، 31080 و 14480 است. Pvalue براي تمام آزمون ها نزديك به صفر است. VIF براي هرسه الگو كمتر از 10 است. همچنين آماره دوربينواتسن نيز در ناحيه قبول قرار مي گيرد. لذا 127سال دماي ماهانه (1885 تا 2011) ايستگاه مشهد به اين ترتيب كامل شد. آزمون همگني، استقلال، تصادفي بودن و نبود دادهپرت انجام شد كه نتايج رضايت بخش است. بررسي 127سال دماي سالانه مشهد نشان داد كه يك نقطه شكست و ايجاد روند صعودي از 1986 به وجود آمده است.
چكيده لاتين :
Introduction
The length of data is very important in forecast and analysis of annual temperature. The long term data may have fluctuations. The time series of temperature as a trend, can show seasonality and cyclical changes and will appear better and frequency analysis result could be improved. Maximum Statistical period of temperatures data in the stations of Iran is about 60 years that is a shortterm. Mashhad has 127 years longterm monthly temperatures (20111885). The recent 60 years (20111951) have been observed by The Mashhad Synoptic Station and the remaining 67 years (19401885) observed by the America#039;s consulate which was located in Mashhad. These 127years longterm data have missing values that should be repaired. There are some stations that have longterm temperatures data and have a good correlation with temperatures of Mashhad. This station consists of Jask (Iran, founded in 1893), Serakhs and Kyzyl (Turkmenistan, founded in 1900 and 1893 respectively), Turkestan (Kazakhstan, founded in 1885) and Baghdad (Iraq, founded 1893).
Materials and methods
Mashhad a city in the center of Khorasan Razavi Province, is located in the northeast part of Iran at a latitude of , longitude of and an altitude of 946 MSL. Mashhad has 127 years longterm monthly temperatures (20111885). The recent 60 years (20111951) have been observed by The Mashhad Synoptic Station and the remaining 67 years (19401885) by the America#039;s embassy which has been located in Mashhad observed.
In linear regression, the model specification is that the dependent variable, is a linear combination of the parameters (but need not be linear in the independent variables). For example, in simple linear regression for modeling n data points there is one independent variable: , and two parameters, and : according to equation (1).
(1)
In multiple linear regressions, there are several independent variables or functions of independent variables, according to equation (2).
(2)
Results and discussion
Multiple linear and nonlinear patterns were used for estimating missing data in this paper. Three Linear models finally accepted and were used for this purpose.
Multiple linear models for restoration of data in years 19041895, 19491941, January, February and July 1905, December 1910, January and February 1914, January 1913, May 1932, September to December 1933 and February and March 1937, by two base stations: Jask and Turkestan, according to equation (3).
(3)
Univariate linear pattern according to equation (4) using data from the years 18851890 by using Turkistan stations to extending data and reconstruction of June, July and August 1892, October 1894 and March 1951 were used.
(4)
Univariate linear pattern by using the temperature of Jask station to reconstruct July 1918 to December 1919 were used. This pattern is according to equation (5).
(5)
The coefficients of determination of these models are 0.98, 0.96 and 0.93 respectively. Classical tests and diagnostic of models showed acceptance of them. Fstatistics of these models are 23160, 31080 and 14480, respectively. The Pvalue for all tests is closed to zero. VIF for all models are less than 10. DurbinWatson statistic located in accepted regions.
Conclusion
Multiple linear and nonlinear models were used for estimating missing data in this paper. Three Linear models finally accepted and were used for this purpose. The coefficients of determination of these models are 0.98, 0.96 and 0.93 respectively. Classical tests and diagnostic of models showed acceptance of them. Fstatistics of these models are 23160, 31080 and 14480, respectively. The Pvalue for all tests is closed to zero. VIF for all models are less than 10. DurbinWatson statistic located in accepted regions .Therefore 127 years of monthly temperature for Mashhad (1885 to 2011) was completed. The test of homogeneity, independence, randomness and outliers were done and accepted. The analyzing of 127 Mashhad annual temperatures showed that there is a change point and an increasing trend from 1986.
