ارزیابی داده‌های بارندگی حاصل از ماهواره Trmm، مدل Mm5 و مشاهدات زمینی به صورت مکانی-زمانی در مناطق خشک و نیمه‌خشک کوهستانی

در مواجه با خطر سیل و یا خسارات ناشی از خشکسالی، برآورد میزان بارش و الگوی تغییرات مکانی آن در یک منطقه گسترده، یکی از چالش‌های مهم در علوم هواشناسی، کشاورزی و هیدرولوژی است. اندازه‌گیری محلی بارندگی در مناطق دور افتاده به دلیل هزینه زیاد و محدودیت‌های عملیاتی دشوار است. بدین علت در تحقیق حاضر به‌منظور تعیین الگوی مکانی-زمانی بارش و امکان تلفیق داده‌ها، سه نوع مختلف از تولیدات بارندگی شامل داده‌های ماهواره‌ای (trmm3b42)، داده‌های حاصل از مدل پیش‌بینی عددی جوّی (mm5) و اندازه‌گیری‌های زمینی (نقشه‌های حاصل از روش زمین‌آمار (ked))، مورد مطالعه قرار گرفتند. این مطالعه در بازه زمانی سال‌های 2000 تا 2010 میلادی و برای منطقه شمال شرق ایران به صورت ماهانه، فصلی و سالانه انجام شد. داده‌ها با استفاده از شاخص اعتبارسنجی rmse و الگوریتم تشابه با یکدیگر مقایسه شدند. نتایج نشان دادند که یکی از ضعف‌های روش زمین‌آمار نبودن اطلاعات کافی در ارتفاعات بالای (1500) متر منطقه است. همچنین دقت تصاویر ماهواره‌ای در فصل‌های گرم بیشتر بود؛ بطوریکه در ماه آگوست مقدار 1/7= rmse به دست آمد. در فصل زمستان (ماه ژانویه) بیشترین مقدار 14/02= rmse حاصل شد که این امر عملکرد ضعیف تولیدات ماهواره‌ای trmm در مناطق پوشیده از یخ را نشان می‌دهد. در اعتبارسنجی مدل mm5 بیشترین و کمترین مقدار rmse به ترتیب 6/64 و 1/05 به دست آمد. علاوه بر این مدل mm5 تا حدود زیادی در شبیه‌سازی مقادیر بارندگی سالانه بیش‌برآورد داشت. نتایج تحلیل‌های مکانی زمانی الگوریتم تشابه نیز نشان دادند که عملکرد مدل mm5 در مقیاس ماهانه و فصلی و تعیین مناطق بارندگی بهتر از تصاویر ماهواره‌ای trmm بود. همچنین هر سه محصول الگوی مکانی بارندگی در مقیاس فصلی و سالانه را به‌خوبی نشان دادند.

Evaluation of Rainfall Data Derived from TRMM Satellite, MM5 Model and Ground Observation using SapatioTemporal Analysis in Arid and SemiArid Mountainous Area

1. IntroductionPrecise estimates of rainfall in areas with complex geographical features in the field of climatology, agricultural meteorology and hydrology is very important. TRMM satellite is the first international effort to measure rainfall from space reliably (Smith, 2007). Another set of data that has become available in recent years is the output of numerical prediction models. Akter and Islam (2007) used MM5 model for weather prediction especially for rainfall in Bangladesh. They compared MM5 outputs with 3B42RT production of TRMM, rain gage and radar data and concluded that MM5 is reliable for rainfall prediction. Ochoa et al. (2014) compared 3B42 product of TRMM with simulated rainfall data by WRF model. Their results showed that TRMM data is more applicable for presenting spatial distribution of annual rainfall. In addition to the methods of statistical comparison, the similarity algorithm (Herzfeld & Merriam, 1990) was also used in this study. This algorithm compares a large number of data simultaneously, which can be in the form of maps or models output. In Iran, very few studies have compared the output of numerical prediction models with TRMM products of rainfall. The aim of this study was to evaluate and compare the rainfall data using similarity algorithm for different locations and time periods in order to fill a gap in the spacetime data.2. Material and MethodsThe study area consisted of North Khorasan, Khorasan Razavi and South Khorasan provinces in North East of Iran, which is geographically located between the longitudes of 55 to 61 degrees and latitudes of 30 to 38 degrees. The climate of the area is arid and semi arid. Total area is approximately 313000 square kilometers. In this study, three types of data were used. Groundbased observations used from synoptic and raingauge stations of Meteorology Organization. The seventh series products of TRMM 3B42 sensor containing three hours TRMM rainfall data with a spatial resolution of 0.25 degree were downloaded for free from the site of NASA. MM5 model outputs which were in the form of images with a spatial resolution of 0.5× 0.5 degrees for the period of 20002010 were also obtained from NASA and NOAA .In this study, KED as a geostatistical method was used to interpolate rainfall. For running geostatistics algorithms, GS + and ArcGIS software were used. Similarity algorithm was executed for each grid point map and the similarity values were derived. After standardization by calculating the similarity value for the entire study area, F network model for similar map was created. In similarity algorithm, closest values to zero indicate a good similarity between the input maps in a specific location and higher values indicate weaker similarity. Standardization algorithms, similarity and analytical software programming in MATLAB were performed for each grid point of the map.3. Results and DiscussionRMSE values for MM5 model were higher in the warm months. The highest RMSE values were obtained in late spring and early summer. This result proved that in the summer, rainfall was predicted less accurately than in the cold months in winter. RMSE values for TRMM showed a reverse pattern with MM5 model output. Maximum amount of RMSE for TRMM was obtained in January with 14 mm per month. The reason for this may be because microwave energy scattering from frozen ice on the ground. The scattering from rain or frozen rain in the atmosphere is similar. Similarity values in the area were scattered with uniform distribution that represents the least significant interannual variation is cold seasons. For the warm seasons, in the south and north of the area, similarity values vary from 1 to 2. Results showed that interannual variations of rainfall in warm seasons and in central areas is high. One of the reasons for these results can be errors in the observed data.By examining the time series of TRMM images using similarity algorithm, we found that in the cold season, the south zone of the study area had similarity values 0.05 to 0.1 with a uniform distribution of values. However, higher similarity values were obtained for the northern and central areas where the distribution of similarity values was not uniform.Due to these facts, it can be concluded that rainfall production of TRMM data was relatively good in the cold season in south and relatively week in north and central parts of the region. In the warm season the least amount of similarity could be seen in the northeast part of the study area. But generally, TRMM estimated rainfall fairly in the warm season.4. Conclusion The validation results of MM5 model rainfall and TRMM monthly rainfall images showed that the model predicted rainfall amounts in the cold months better than in the warm months. However unlike the MM5 model, remote sensing images had the highest error in cold months. The reason was the presence of snow and ice on the ground in the cold months of winter. Considering interannual and seasonal changes, it became clear that there is much difference between interannual remote sensing image changes and the actual amounts of rainfall (KED). Nevertheless the model interannual changes were consistent with real data. Interannual changes of the model and the station data (KED) were higher in cold season.KED methods also retained spatial variability of rainfall as well as remote sensing data and model output. The estimates, especially above 1500 meters in the central regions, had low precision in the products. The results showed that in the absence of adequate rain gages in the region, MM5 output model and TRMM data could be used to fill the gaps.