【作者】 王瑞；

【导师】 乔枫雪；

【作者基本信息】 华东师范大学 ， 气象学， 2020， 硕士

【摘要】 本文利用WRF模式(Weather Research and Forecasting model)采用不同的网格嵌套、积云对流表达对上海地区的两次暖区暴雨过程(2018年5月25日、2017年6月10日)和一次梅雨锋暴雨过程(2015年6月27日)进行高分辨率(1 km)单向反馈模拟,基于我国东部地区自动观测站的逐小时降水数据,系统评估了27、15、9、5、3、1 km不同网格尺度上积云对流表达对极端降水模拟的影响。本文所涉及的积云对流表达包括传统的Kain-Fristch(KF)、尺度自适应的Grell-Freitas(GF)对流参数化方案、以及针对“灰色区域”(1-10 km)的显式对流Explicit convection(EC)方案,并着重比较传统三层网格嵌套(15:5:1和9:3:1)与两层大比率网格嵌套(15:1和9:1)方式中,外网格使用不同积云对流表达(KF,GF,EC)对高分辨率1 km内网格极端降水预报的影响。本文基于多种统计方法(Threat score、Bias score、相关系数、均方根误差等)对极端降水的日平均雨量、小时降水演变进行全面评估,评估结果显示,两层大比率嵌套方式(15:1、9:1)对极端降水的模拟优于传统的三层嵌套,并且可以有效规避中间嵌套网格(3-5 km)在“灰色区域”范围内因积云对流表达不确定性引发的问题;在统计分析基础上,本文还从对流降水贡献率、大尺度强迫场、对流发展的热、动力结构等方面,详细阐述了不同积云对流在跨越“灰色区域”时的模拟差异。具体结论如下:(1)对于嵌套母网格(>10 km),在27 km网格上,KF方案系统性地低估了极端降水范围、强度及小时降水峰值,GF方案则高估了极端降水范围,并容易产生虚假的午后降水峰值;在15 km网格上,KF、GF方案模拟结果的系统性偏差均有所改善。随着分辨率提高到9 km“灰色区域”网格,采用积云对流方案比采用显式表达(EC)对于极端降水模拟更有优势,其中,KF方案对于极端降水范围、强度及降水日变化形态的预报比GF方案更有优势,GF方案则对小时降水主峰值预报较好,因此,在9 km网格上对于极端降水的预报,仍需要使用对流参数化方案,KF和GF方案各有优势。(2)对于中间层嵌套网格(5 km,3 km),就极端降水范围、强度及小时降水变化的预报而言,在5公里网格上使用对流参数化方案(KF或GF)比使用EC方案存在优势,其中,尺度自适应的GF由于方案中根据网格间距和云底卷夹率对垂直对流涡动通量输送进行了限制,在5 km下的网格距,其对流降水大幅减少,预报结果与EC方案之间的差异较小。然而,3 km网格上极端降水的预报结果与参数化方案(隐式对流)的选取有关。研究发现使用KF方案对极端降水强度、落区及小时降水变化的预报要比直接使用EC方案存在优势;使用GF方案对小时降水变化的预报虽然优于EC方案,但其对极端降水强度的预报却比EC方案差;这说明“灰色区域”3 km网格上积云对流表达仍存在较大不确定性。(3)对于1 km内网格,采用大比率的双层嵌套(15:1或9:1)方法对于极端降水的模拟更有优势,在母网格(15 km或9 km)与内网格(1 km)之间加入中间嵌套网格(5km和3 km),对于1 km高分辨率网格的极端降水预报并无明显改善,甚至大多数情况下会降低高分辨率网格的预报能力。母网格使用KF方案经双层嵌套直接降尺度到1 km时,比使用GF方案或者EC方案,对极端降水范围、强度及小时变化模拟效果更好。KF方案的模拟优势可能由于其基于有效位能释放的闭合假设,同时在对流触发机制中考虑了格点垂直速度以及云底大尺度辐合对积云卷夹等大尺度强迫场的影响,从而更好地反映了积云与大尺度环境场之间的相互作用。母网格(15 km或9 km)使用KF方案模拟的低层水汽辐合和垂直上升运动的强度变化与小时降水的强度变化较为一致,KF方案中对流降水贡献较大,说明方案中对于与强降水相关的中尺度对流系统具备更强的组织和调控作用,并由此调节高分辨率格点降水小时变化的模拟。而母网格使用GF方案时,由于方案的自适应调整机制,其在15 km到9 km的对流降水贡献率偏小,对1 km内网格的格点降水调控作用较弱,使其对小时降水变化模拟存在一定问题。此外,针对预报优势明显的双层嵌套,使用KF方案降尺度到1 km后,极端降水局部仍存在低估问题,因此,本文进一步比较了三种边界层方案(MYNN,MYJ,YSU)对高分辨率模拟的影响。发现局地的MYNN边界层方案由于其边界层内热、动量垂直混合较弱,容易形成更低的逆温层和不稳定的边界层,进而导致边界层顶到自由大气中高层的垂直上升运动较强,降水较强,能更有效地改善极端降水强度的预报。因此,本研究提出基于大比率网格嵌套(15:1或9:1),在母网格采用传统的KF对流参数化方案,并同时使用局地的MYNN边界层方案,1 km内网格(EC)对上海地区极端暴雨的落区、强度以及小时变化具有较好的预报能力。更多还原

【Abstract】 In this paper,WRF model is used to simulate two extreme precipitation events during warm season and one Meiyu front precipitation event in Shanghai with different grid nesting and cumulus treatment.Based on the hourly precipitation data of automatic observation stations in the eastern part of China,the influence of cumulus treatment on extreme precipitation simulation at different grid scales(27,15,9,5,3,1 km)is systematically evaluated.The cumulus treatment used in this paper includes traditional Kain Fristch(KF),scale-aware Grell Freitas(GF)convection parameterization schemes,and explicit convection(EC)in grey zone(1-10km).At the same time,the traditional triple nesting(3:1 or 5:1)is compared with a new large ratio double nesting(9:1 or 15:1)configuration,and the impact of cumulus convective expressions on high-resolution grid precipitation simulation through different nesting methods is also studied.Based on various statistical methods(Thread score,Bias score,Correlation coefficient,Root mean square error)to evaluate the daily extreme precipitation and its diurnal cycle variations.The simulation results show that the double nesting with large ratio(9:1 or 15:1)is more advantageous for extreme precipitation simulation than the traditional triple nesting(3:1 or 5:1).Using double nesting with large nesting ratio can not only save computing resources,but also effectively avoid a series of problems caused by the uncertainty of cumulus convection expression in the process of crossing the "gray zone".In addition,the in-depth physical mechanism analysis of simulation results is also carried out in the aspects from the contribution rate of convective precipitation,large-scale forcing fields,thermal and dynamic structure of convective development.The conclusions are as follows:(1)In the D1 grids(> 10 km),at 27-km grid,using KF systematically underestimated the range and intensity of extreme precipitation,while using GF systematically overestimated the range of extreme precipitation and resulted a fake precipitation peak in the afternoon;At 15-km grid,the systematic deviation from KF and GF has both been improved.As the resolution increased to 9 km,the grey zone grid,using KF has more advantages than GF and EC in forecasting the range,intensity and diurnal cycle variations of extreme precipitation.Using GF scheme performed better on simulating the time of primary hourly precipitation peak,while EC has poor prediction on intensity of hourly precipitation rate.Therefore,in "grey zone" 9 km grid,using convective parameterization is still needed for the extreme precipitation prediction,KF and GF schemes both have their own advantages.(2)In the D2 grids(5 km & 3 km),using KF or GF has advantages over EC for the prediction of extreme precipitation range,intensity and its diurnal cycle variations on the 5-km grid;However,on the 3 km grid,the prediction results are related to the selection of convection parameterization schemes.It is found that KF scheme has advantages over EC scheme in both intensity and hourly variations of extreme precipitation forecasting.Using GF is better than EC in diurnal cycle forecasting for extreme rainfall,but worse than EC in forecasting the intensity of extreme precipitation.It shows that the expression of cumulus convection on 3 km grid in "grey zone" is still uncertain.(3)In the 1 km grid(D3),it is more advantageous to use the double nesting with large ratio(15:1 or 9:1)than the traditional triple nesting for extreme precipitation simulations.The results of extreme precipitation prediction in the 1-km grid are not significantly improved,or even reduced in most cases,when the intermediate nesting(5km and 3km)is added between the parent grid(15km or 9km)and the inner grid(1km).Using 9KF-1 is more advantageous to simulate the range,intensity and hourly variations of extreme precipitation than using 9GF-1 or 9EC-1.The simulation advantage of KF scheme may be related to the closure hypothesis based on the release of CAPE,and its convective triggering mechanism considering the vertical velocity and the effect of large-scale convergence of the cloud bottom due to the cumulus entrainment.The variations of water vapor convergence and vertical upward motion simulated by KF scheme in the parent grid(15 km or 9 km)are consistent with the hourly variations of extreme precipitation,in which the contribution of convective precipitation is larger.In KF scheme,the hourly precipitation intensity in 1-km grid is affected by the regulation and organization of mesoscale convection activity in D1 grids which are related to heavy precipitation.However,in scale-aware GF scheme,the convective precipitation of 15 km to 9 km in D1 grid is significantly reduced,and the regulation for the precipitation in 1-km grid is weak,which resulted in its primary rainfall peak was underestimated.In addition,in this paper we also try to improve the simulation results from the perspective of the boundary layer scheme.It is found that the local MYNN boundary layer scheme is easy to form an unstable top of boundary layer due to the weak vertical mixing of heat and momentum in the boundary layer,which leads to a strong vertical upward movement from the top of the boundary layer to the middle and upper levels of the free atmosphere and produced stronger precipitation.The local MYNN scheme has more advantages in capturing extreme precipitation than the non-local YSU scheme.Therefore,based on the large ration double nesting grid,we proposed that using MYNN in the 9KF-1 or 15KF-1 configuration produced most realistic intensity and hourly variations of extreme rainfall.更多还原