ارزيابي و مقايسه شبيه‌سازي‌هاي مدل‌هاي HARMONIE و WRF در مقياس همرفتي در منطقه غرب ايران

Inter-comparison of HARMONIE and WRF model simulations in convective-permitting scale over western area of Iran

استفاده از مدل­ هاي پيش­بيني عددي وضع هوا براي پيش ­بيني پديده­ هاي هواشناسي در مقياس همرفتي توجه زيادي را به خود جلب كرده است. بدين‌ منظور استفاده از مدل­هاي مختلف پيش­بيني عددي وضع هوا و مقايسه نتايج شبيه­ سازي اين مدل­ها در درك بهتر مشكلات مرتبط با اين مقياس ­ها و شناخت خطاهاي سامانمند مدل ­ها كمك بسياري خواهد كرد. در اين تحقيق شبيه­ سازي­هاي مختلف با استفاده از دو مدل ميان ­مقياس WRF و HARMONIE و با شرايط اوليه و مرزي يكسان (ECMWF ERA-Interim) به مدت 15 روز در ماه دسامبر 2013 انجام شده است. تمامي شبيه­ سازي­ ها با تفكيك افقي در مقياس همرفتي 2/5 كيلومتر و آغازگري شده در ساعت ­هاي 00UTC و 12UTC مي­باشند. اجراي مدل به مدت 72 ساعت در منطقه ­اي با كوهساري پيچيده در نيمه غربي ايران انجام شده است. نتايج نشان داد كه دو مدل از عملكرد مشابهي در شبيه­ سازي متغيرهاي مختلف برخوردار هستند. براي شبيه­ سازي بارش تجمعي، نتايج نشانگر عملكرد بهتر مدل WRF نسبت به مدل HARMONIE در دوره زماني شبيه ­سازي است. براي بارش تجمعي 24 ساعته، مدل WRF داراي همبستگي اندكي بيشتر و خطاي ميانگين مربعات كمتر در همه زمان­هاي پيش ­بيني بين مقادير مشاهداتي و شبيه­ سازي شده است، اما براي اريبي بارش، HARMONIE داراي عملكرد بهتري است. به ­واسطه به‌كارگيري داده­ هاي بازتحليل به‌ عنوان شرايط مرزي جانبي، افزايش زمان پيش ­بيني اثر قابل توجهي بر امتيازهاي ارزيابي بارش نداشت. براي متغيرهاي هواشناسي سطح زمين، اختلاف قابل توجهي در ارتباط با اريبي رطوبت نسبي تراز 2 متر براي مدل WRF و HARMONIE وجود داشت (فراتخمين رطوبت براي HARMONIE و فروتخمين براي مدل WRF). براي متغيرهاي ترازهاي بالا، بيشترين شباهت­ ها در اريبي و انحراف معيار خطا مرتبط با نيم ­رخ ­هاي قائم دما و بيشترين تفاوت­ها در اريبي رطوبت نسبي در تراز 850 هكتوپاسكال مشاهده شد.

Ever increasing attention is being paid to the use of Numerical Weather Prediction (NWP) models in the convection-permitting mode for providing high-resolution forecasts. In such applications, the use of NWP models and comparison among the simulations of models help us to understand the problems associated with these scales and to unravel the systematic errors of the models. In this study, two weeks of “model simulation experiments” have been conducted with the HARMONIE-AROME and the WRF-ARW meso-scale NWP models at 2.5 km horizontal resolution, in order to partly resolve convective phenomena on the same domain over the mountainous areas of the west of Iran for the period of 1–15 December 2013. All experiments have been conducted by using the ECMWF ERA-Interim reanalyses for the lateral boundary conditions, and for this reason, they are called “model simulation experiments”. The HARMONIE Verification System has been used for the validation, and operational radiosonde and SYNOP observations from the ECMWF have been used for the verification. The precipitation observations from some climatological stations of Iran have also been used. The model simulations described in this study were run up to +72 h. The motivation for this long simulation time is to investigate any possible systematic model problems that could hide possible impact of data assimilation in the planned data assimilation forecast experiments. Generally, the WRF and HARMONIE have a comparable performance, both of which have similar results for some variables at all forecast lead times. For 24-hour accumulated precipitation forecasts, the correlation coefficient, the bias and the root mean square error (RMSE) were used to compare the performance of both models over the same area. For the correlation coefficient and the RMSE, the WRF has slightly better verification scores at all lead times. The results for the temperature at 2 m, wind speed and direction at 10 m, and specific humidity (mixing ratio) at 2 m are verified by using different verification scores. A similar behavior is found for both models in the error standard deviation (STDV) verification score; although some minor differences are observed at some lead times and for some variables. A more significant difference is related to the bias of specific humidity at 2 m for the WRF and HARMONIE as over-estimation of moisture for the HARMONIE and its under-estimation for the WRF. Considering the upper air profiles of the bias and the STDV of the error, both similarities and differences were shown for the vertical structures of various quantities as obtained by the two model simulations. While the strongest similarity was seen in the bias and the STDV of the temperature error profiles, the relative humidity at 850 hPa exhibited the largest differences in both measures of error. A dry bias, which increased with the forecast time, was noticed for the WRF at low levels (850 hPa) as verified against the radiosonde data as well as the SYNOP data at 2 m level.