1.Key Laboratory of Meteorological Disaster of Ministry of Education/Joint International Research Laboratory of Climate and Environment Change/Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters, Nanjing University of Information Science and Technology, Nanjing 210044, China 2.Weather Modification Office of Heilongjiang Province, Heilongjiang Meteorological Bureau, Harbin 150030, China 3.Climate Model Division, National Climate Center, China Meteorological Administration, Beijing 100081, China Manuscript received: 2017-10-11 Manuscript revised: 2018-07-29 Manuscript accepted: 2018-08-13 Abstract:In this study, we evaluate the forecast skill of the subseasonal-to-seasonal (S2S) prediction model of the Beijing Climate Center (BCC) for the boreal summer intraseasonal oscillation (BSISO). We also discuss the key factors that inhibit the BSISO forecast skill in this model. Based on the bivariate anomaly correlation coefficient (ACC) of the BSISO index, defined by the first two EOF modes of outgoing longwave radiation and 850-hPa zonal wind anomalies over the Asian monsoon region, we found that the hindcast skill degraded as the lead time increased. The ACC dropped to below 0.5 for lead times of 11 days and longer when the predicted BSISO showed weakened strength and insignificant northward propagation. To identify what causes the weakened forecast skill of BSISO at the forecast lead time of 11 days, we diagnosed the main mechanisms responsible for the BSISO northward propagation. The same analysis was also carried out using the observations and the outputs of the four-day forecast lead that successfully predicted the observed northward-propagating BSISO. We found that the lack of northward propagation at the 11-day forecast lead was due to insufficient increases in low-level cyclonic vorticity, moistening and warm temperature anomalies to the north of the convection, which were induced by the interaction between background mean flows and BSISO-related anomalous fields. The BCC S2S model can predict the background monsoon circulations, such as the low-level southerly and the northerly and easterly vertical shears, but has limited capability in forecasting the distributions of circulation and moisture anomalies. Keywords: BCC S2S model, boreal summer intraseasonal oscillation, forecast skill, northward propagation 摘要:本研究评估了国家气候中心(Beijing Climate Center, BCC)次季节-季节预测(S2S)模式对北半球夏季节内振荡(boreal summer intraseasonal oscillation, BSISO)的预测技巧,并讨论了影响BSISO预报技巧的关键因素。基于亚洲季风区向外长波辐射(OLR)和低层纬向风(U850)的多变量经验正交分解(MV-EOF),其前两模态可代表北传的BSISO模态;利用这两个模态的时间序列计算双变量相关系数(ACC),发现模式预报BSISO的能力随着模式预报超前时间的增加而降低。提前11天的预报中,ACC下降到0.5以下,显示模式不足以预报出BSISO的强度和北传特征。为了确定导致BSISO在超前预报时间为11天 预报技巧减弱的原因,本研究诊断了观测中导致BSISO向北传播的主要物理过程。结果显示,在提前11天的预报中,模式明显低估了由平均环流和BSISO相关扰动场相互作用所引起的对流北侧低层正涡度、大气加湿和大气增暖的异常信号。进一步诊断显示,BCC S2S模式可以预测低层南风、北风和东风垂直切变等平均季风环流,但对于环流和水汽等扰动场的预报能力有限,无法预报出对流北侧大气不稳定状态,从而限制了BSISO北传的预报能力。 关键词:BCC S2S模式, 北半球夏季季节内振荡, 预报技巧, 北传
HTML
--> --> --> -->
2.1. BCC S2S model
The BCC S2S model is based on BCC_CSM1.1, with the inclusion of atmosphere-ocean-ice-land coupling (Wu et al., 2014). The atmospheric component of BCC_CSM1.1 is the BCC Atmospheric General Model, version 2.1 (Wu et al., 2010), with a horizontal resolution of T106 and 40 layers in the vertical direction. The land component is the BCC Atmosphere and Vegetable Interaction Model, version 1 (Ji, 1995). The oceanic component is the GFDL Modular Ocean Model, version 4 (Griffies et al., 2005), with gradually increased resolutions (1° to 1/3°) from 30°S or 30°N to the equator and 40 levels in the vertical direction. The sea-ice component is the GFDL Sea Ice Simulator (Winton, 2000), which has the same resolution as that in the oceanic component. The S2S hindcast experiment designed by the S2S project is to carry out the prediction by initializing on each day from 1 January 1994 to 31 December 2013 and integrating the model for 60 days. To reduce the uncertainty of initial conditions, for each initial date the strategy of four members using initial conditions at 0000, 0600, 1200 and 1800 UTC is adopted. Since the ensemble mean generally has higher skill, we focus on the outputs of the ensemble mean from these four members. Daily averaged variables are used in this study. Because the hindcast experiment of S2S prediction was conducted each day for the future 60 days during the period of 1994-2013, the model outputs contain a continuous distribution of hindcasts on all dates from 1994 to 2013, at all lead times from 1 to 60 days. Thus, a time series of model data at any fixed lead time can be formed to compare against the observation for understanding the model forecast skill at that lead time.
2 2.2. Observational data and reanalysis products -->
2.2. Observational data and reanalysis products
The observational and reanalysis datasets used for model evaluation include: (2) daily outgoing longwave radiation (OLR) data (Liebmann and Smith, 1996), which are used as a proxy for convection, provided by the National Oceanic and Atmospheric Administration (NOAA); (3) the ERA-Interim dataset (Dee et al., 2011) of the ECMWF; (4) the NCEP-NCAR reanalysis (Kalnay et al., 1996); and (4) the NOAA High-resolution Blended Analysis of Daily Sea Surface Temperature (SST; Reynolds et al., 2007). To reduce the uncertainty of reanalysis products, we averaged the ERA-Interim and NCEP-NCAR fields for horizontal wind components, 3D specific humidity and surface sensible heat flux (SHF). The resolution in the reanalysis datasets and model outputs are unified on a 2.5°× 2.5° grid over eight layers (1000, 925, 850, 700, 500, 300, 200, and 100 hPa). The summer season is defined as May to October.
2 2.3. Evaluation metrics -->
2.3. Evaluation metrics
To better evaluate the real-time forecast capability of the BSISO, we adopt the multi-variate (MV) EOF analysis proposed by (Lee et al., 2013) to derive the real-time BSISO index from different lead forecasts. Our evaluation focuses on the spatial pattern and temporal evolution of the real-time BSISO index in the BCC S2S model output. As in (Lee et al., 2013), we first remove the climatology (the first three harmonics) and the interannual variability (using the running mean of the last 120 days) from daily OLR and 850-hPa zonal wind (U850) fields in the Asian monsoon region (10°S-40°N, 40°-160°E). We then normalize and decompose the two anomaly fields, using the MV-EOF analysis. The first two modes (EOF1 and EOF2) can effectively represent the north-northeastward propagation of the 30-90-day BSISO. Using the time series (PC1 and PC2) of these two modes, we can then divide the BSISO lifecycle into eight phases (Wheeler and Hendon, 2004; Lee et al., 2013), as in Table 1. When (PC12 + PC22)1/2>1.0, it is considered as an active BSISO event. The third and fourth EOF modes represent the northwestward-propagating signal with a shorter period of 10-30 days (Lee et al., 2013). In our model assessments, the filtering and EOF analysis were carried out separately for each lead time. (Jie et al., 2017) noted that the operational models generally show lower skills in predicting this 10-30-day mode. Our analysis also finds that the skill of the 10-30-day BSISO prediction of BCC S2S model is about nine days (not shown), similar to the range of weather forecasts. Therefore, only the 30-90-day BSISO mode is investigated in this study. To quantitatively evaluate the simulation and forecast skills of the 30-90 day BSISO, we calculate the PCC, root-mean-square error (RMSE) and bivariate anomaly correlation coefficient (ACC) of PC1 and PC2. These metrics have been used widely in evaluating MJO/ISO prediction (Lin et al., 2008a; Xiang et al., 2015; Liu et al., 2017). The definitions of these metrics are given below: \begin{eqnarray*} {\rm PCC}&=&\frac{\sum_{j=1}^M\sum_{i=1}^N[(a_{i,j}-\bar{a})(b_{i,j}-\bar{b})]} {\sqrt{\sum_{j=1}^M\sum_{i=1}^N(a_{i,j}-\bar{a})^2\sum_{j=1}^M\sum_{i=1}^N(b_{i,j}-\bar{b})^2}} ;\\ {\rm RMSE}&=&\sqrt{\frac{1}{MN}\sum_{j=1}^M\sum_{i=1}^N(a_{i,j}-b_{i,j})^2} ;\\ {\rm ACC}(\tau)&=&\frac{\sum_{t=1}^T(a_{1t}b_{1t}+a_{2t}b_{2t})} {\sqrt{\sum_{t=1}^T(a_{1t}^2+a_{2t}^2)}\sqrt{\sum_{t=1}^T(b_{1t}^2+b_{2t}^2)}} . \end{eqnarray*} Here, a is observation and b is model output; i and j refer to longitude and latitude points, respectively; M and N indicate the numbers of longitude and latitude points, respectively. An overbar in the definition of PCC represents an area average. The subscript t is time, and τ is the forecast lead time. All the time points in May-October during the period of 1994-2013 were included for the ACC calculation. The subscripts 1 and 2 are for the first and second variables (such as PC1 and PC2), respectively.
2 2.4. Diagnosis of BSISO propagation -->
2.4. Diagnosis of BSISO propagation
Several mechanisms associated with atmospheric internal dynamics and air-sea (land) coupling processes have been proposed to explain the northward propagation of the BSISO [see reviews by (DeMott et al., 2012) and (Wang, 2012)], which are summarized in Fig. 1. In observations, the positive vorticity and moistening in the lower troposphere appear to the north of BSISO convection (Hsu and Weng, 2001; Jiang et al., 2004; Tsou et al., 2005; Bellon and Sobel, 2008), favoring northward development of the convection. Two mechanisms may account for the generation of positive vorticity to the northern flank of the convection over the Indian Ocean (Jiang et al., 2004; Bellon and Sobel, 2008). According to (Jiang et al., 2004), the interaction between background vertical wind shear and the meridional gradient of anomalous baroclinic divergence can induce a positive anomaly of barotropic vorticity (ζ'+) to the north of the convection. This process can be written as \begin{equation} \frac{\partial{\zeta}'_+}{\partial t}\propto\bar{u}_T\frac{\partial{D}'_-}{\partial y} , \ \ (1)\end{equation} where an overbar represents the seasonal-mean component and a prime indicates the anomalous component related to the BSISO, which is extracted from the 30-90-day Lanczos filtering (Duchon, 1979). In Eq. (2), ζ'+ indicates the barotropic mode of the vorticity anomaly; $\bar{u}_T$ is the seasonal-mean vertical wind shear, defined by the difference in zonal wind between 200 and 850 hPa $(\bar{u}_200-\bar{u}_850)$; and D'- represents the baroclinic mode of divergence (anomalous divergence between 200 and 850 hPa). Accompanied by the enhanced convective heating anomaly, anomalous ascending motion appears in the mid troposphere and a divergence (convergence) anomaly can be observed in the higher (lower) level (D'->0). The maximum D'- is located at the convective center; thus, the meridional gradient of baroclinic divergence is negative (positive) to the north (south) of the convection. Under the easterly shear over the Asian summer monsoon region, a positive (negative) barotropic vorticity may be induced in the northern (southern) part of the convective center. Figure1. Schematic diagram summarizing the main mechanisms responsible for the northward propagation of the BSISO. See the text in section 2.4 for detailed explanations of each process.
(Bellon and Sobel, 2008) emphasized the role of vorticity advection by the background meridional flows in the generation of a positive vorticity anomaly to the north of the convection. The process can be written as \begin{equation} \frac{\partial{\zeta}'_+}{\partial t}\propto-\bar{v}_T \frac{\partial{\zeta}'_-}{\partial y} , \ \ (2)\end{equation} where $\bar v_T$ is the seasonal-mean vertical wind shear of meridional wind between 200 and 850 hPa $(\bar v_200-\bar v_850)$ and ζ'- represents the baroclinic mode of vorticity anomaly (difference in the vorticity anomaly between 200 and 850 hPa). Over the BSISO convective area, the positive vorticity anomaly at the low level is coupled with the negative vorticity anomaly at the high level, forming a baroclinic vorticity couplet. As the background monsoonal northerly (southerly) in the higher (lower) troposphere works on the baroclinic couplet of the vorticity anomaly, a positive (negative) vorticity anomaly is induced over the northern (southern) region of the convection. The leading phase of planetary boundary layer (PBL) moisture is also favorable for the northward propagation of the BSISO, as it destabilizes the atmospheric column (Hsu and Weng, 2001; Kemball-Cook and Wang, 2001; Jiang et al., 2004; Hsu and Li, 2012). The PBL moistening can result from the PBL moisture convergence related to the low-level vorticity anomaly and the meridional advection of moisture. Both the seasonal-mean flows and perturbation wind and moisture fields can affect moisture advection. To diagnose the major processes, we decompose the specific humidity and meridional wind fields into the summer-mean component (with an overbar) and the BSISO-related perturbation part (with a prime). Thus, three processes associated with mean flow and perturbation interactions can cause meridional advection of moisture, as expressed by Eq. (4) below: \begin{equation} -\left(v\frac{\partial q}{\partial y}\right)'=-\left(\bar{v}\frac{\partial{q}'}{\partial y}\right)' -\left({v}'\frac{\partial\bar{q}}{\partial y}\right)'-\left({v}'\frac{\partial{q}'}{\partial y}\right)' , \ \ (3)\end{equation} where q is the specific humidity. In addition to the internal atmospheric dynamics, the surface warming induced by anomalous surface heat fluxes to the north of the convection also plays a role in the northward propagation via destabilization of the lower atmosphere (Webster, 1983; Kemball-Cook and Wang, 2001; Fu et al., 2003). (Webster, 1983) pointed out that increased SHF over the South Asian peninsula could reduce the stability of the low-level atmosphere and favor the northward propagation of convection near the equator. (Kemball-Cook and Wang, 2001) and (Fu et al., 2003) found that the leading phase of the SST anomaly related to anomalous evaporation or latent heat flux (LHF) also favors induction of BSISO northward propagation. In section 4, all these processes are diagnosed and compared using the observation and model outputs at different lead times. This will help us identify the key factors governing the northward propagation of the BSISO in the BCC S2S model.