Advances in Meteorology

Advances in Meteorology / 2019 / Article

Research Article | Open Access

Volume 2019 |Article ID 1010858 |

Tien Du Duc, Cuong Hoang Duc, Lars Robert Hole, Lam Hoang, Huyen Luong Thi Thanh, Hung Mai Khanh, "Impacts of Different Physical Parameterization Configurations on Widespread Heavy Rain Forecast over the Northern Area of Vietnam in WRF-ARW Model", Advances in Meteorology, vol. 2019, Article ID 1010858, 24 pages, 2019.

Impacts of Different Physical Parameterization Configurations on Widespread Heavy Rain Forecast over the Northern Area of Vietnam in WRF-ARW Model

Academic Editor: Pedro Jiménez-Guerrero
Received29 Mar 2019
Revised12 Jul 2019
Accepted18 Jul 2019
Published18 Aug 2019


This study investigates the impacts of different physical parameterization schemes in the Weather Research and Forecasting model with the ARW dynamical core (WRF-ARW model) on the forecasts of heavy rainfall over the northern part of Vietnam (Bac Bo area). Various physical model configurations generated from different typical cumulus, shortwave radiation, and boundary layer and from simple to complex cloud microphysics schemes are examined and verified for the cases of extreme heavy rainfall during 2012–2016. It is found that the most skilled forecasts come from the Kain–Fritsch (KF) scheme. However, relating to the different causes of the heavy rainfall events, the forecast cycles using the Betts–Miller–Janjic (BMJ) scheme show better skills for tropical cyclones or slowly moving surface low-pressure system situations compared to KF scheme experiments. Most of the sensitivities to KF scheme experiments are related to boundary layer schemes. Both configurations using KF or BMJ schemes show that more complex cloud microphysics schemes can also improve the heavy rain forecast with the WRF-ARW model for the Bac Bo area of Vietnam.

1. Introduction

Elongated to 15 degrees (from 8 to 23 degrees north), Vietnam weather is affected by a variety of extratropical and tropical systems such as cold surges and cold fronts, subtropical troughs and monsoon, tropical disturbances, and typhoons [13, 4]. An approximate number of 28–30 heavy rain events occurring over the whole Vietnam region every year were determined from local surface stations. According to Nguyen et al. [4], the rainy season in the Vietnam area usually lasts from May to October with the peak rainy months starting in the north and then moving to the south with time because of the movement of the subtropical ridge and Intertropical Convergence Zone (ITCZ). In the northern part of Vietnam, the peak rainy months can be seen during the June–September period, particularly in July and August, which is partly related to the activity of the ITCZ [5].

Apart from single weather patterns causing heavy rain for Vietnam, it is common to observe a mix of different patterns simultaneously. A good example for this is the historical rain and flood over Central Vietnam in the Hue city in 1999 due to the combination of cold surges and tropical depression-type disturbances [6]. In 2008, extreme heavy rain was witnessed in Hanoi as a result of the dramatic intensification of the easterly disturbance through the midlatitude-tropical interactions [7]. Likewise, a surface low pressure which coincided with the subtropical upper-level trough was the main factor causing historical rainfall over northeastern Vietnam in Quang Ninh Province in 2015 [8].

In the northern part of Vietnam (hereinafter referred to as Bac Bo area)—the focus area of this study—tropical cyclones or surface low-pressure patterns are one among several key factors leading to heavy rain for different regions in Vietnam. ITCZ plays the vital role in facilitating the development of low-level vortexes which are likely to develop into tropical disturbances and typhoons in the summer time, particularly in July and August. As a result, rain may fall over the Bac Bo area with excessive amount, and rain duration directly depends upon the lifetime of ITCZ and the disturbances itself. The other causing reasons are related to the effects of cold surge from the north (Siberian area) or the combinations of different patterns (a cold surge with a tropical cyclone, a cold surge with a trough, a surface low-pressure system with a trough, etc.). Table 1 provides the observed precipitation from SYNOP stations in the Bac Bo area with the annual rainfall mostly ranging from 1500 to 2000 mm and a high number of daily heavy rain occurrences (over 50 mm/24 h and 100 mm/24 h).

Station nameLong.Lat.Annual rain#50 mm#100 mm

Muong Te102.8322.372423112
Sin Ho103.2322.372787132
Tam Duong103.4822.42235592
Muong La104.0321.52141251
Than Uyen103.8821.95192081
Quynh Nhai103.5721.85167871
Mu Cang Chai104.0521.87174151
Tuan Giao103.4221.58158651
Pha Din103.521.57184061
Van Chan104.5221.58148161
Song Ma103.7521.5111820
Co Noi104.1521.13130840
Yen Chau104.321.05122930
Bac Yen104.4221.23151940
Phu Yen104.6321.27149651
Minh Dai105.0521.17169561
Moc Chau104.6820.83158151
Mai Chau105.0520.65169871
Pho Rang104.4722.23164871
Bac Ha104.2822.53161140
Hoang Su Phi104.6822.75173951
Bac Me105.3722.73175360
Bao Lac105.6722.95117930
Bac Quang104.8722.54295249
Luc Yen104.7822.1191781
Ham Yen105.0322.07185881
Chiem Hoa105.2722.15163161
Cho Ra105.7322.45141750
Nguyen Binh105.922.65168661
Ngan Son105.9822.43185381
Trung Khanh106.5722.83182172
Dinh Hoa105.6321.92174982
Bac Son106.3221.9169372
Huu Lung106.3521.5157271
Dinh Lap107.121.53186572
Quang Ha107.7521.452883165
Phu Ho105.2321.45148061
Tam Dao105.6521.472585134
Hiep Hoa105.9721.35160161
Bac Ninh106.0821.18160872
Luc Ngan106.5521.38140661
Son Dong106.8521.33168682
Ba Vi105.4221.15194292
Ha Dong105.7520.97163571
Chi Linh106.3821.08153671
Uong Bi106.7521.03182492
Kim Boi105.5320.33211592
Chi Ne105.7820.48184182
Lac Son105.4520.45204092
Cuc Phuong105.7220.25191892
Yen Dinh105.6719.98150971
Sam Son105.919.75175993
Do Luong105.318.89187192
Lai Chau103.1522.072178102
Sa Pa103.8222.352581102
Lao Cai103.9722.5170471
Ha Giang104.9722.822372122
Son La103.921.33145641
That Khe106.4722.25148261
Cao Bang106.2522.67143961
Bac Giang106.2221.3151871
Hon Ngu105.7718.82007114
Bac Can105.8322.15142151
Dien Bien Phu10321.37153651
Tuyen Quang105.2221.82165482
Viet Tri105.4221.3160182
Vinh Yen105.621.32148961
Yen Bai104.8721.7176881
Son Tay105.521.13161271
Hoa Binh105.3320.82182882
Huong Son105.4318.522135104
Ha Noi105.821.03164782
Phu Ly105.9220.55173182
Hung Yen106.0520.65138161
Nam Dinh106.1520.39156471
Ninh Binh105.9720.23166282
Phu Lien106.6320.8157271
Hai Duong106.320.93156171
Hon Dau106.820.67150181
Van Ly106.320.12162092
Lang Son106.7721.832874122
Thai Nguyen105.8321.6126651
Nho Quan105.7320.32172882
Bai Chay107.0720.97170082
Co To107.7720.981933103
Thai Binh106.3520.451930103
Cua Ong107.3521.02159181
Tien Yen107.421.332189123
Mong Cai107.9721.522168103
Bach Long Vi107.7220.132670155
Huong Khe105.7218.18121661
Thanh Hoa105.7819.752516125
Hoi Xuan105.1220.37170183
Tuong Duong104.4319.28172061
Ha Tinh105.918.351896104
Ky Anh106.2818.072507135
Bai Thuong105.3819.9183072
Nhu Xuan105.5719.63175883
Tinh Gia105.7819.45177793
Quy Chau105.1219.57166171
Quy Hop105.1519.32157972
Tay Hieu105.419.32151772
Quynh Luu105.6319.17155982
Con Cuong104.8819.05164772

In Vietnam National Center for Hydro-Meteorological Forecasting (NCHMF), several global model (NWP) products are mostly applied in operational forecast to predict the occurrence of heavy rain. These include the models from National Centers for Environmental Prediction (NCEP), European Centre for Medium-Range Weather Forecasts (ECMWF), Japan Meteorological Agency (JMA), and Germany’s National Meteorological Service (DWD). In addition, some regional NWP products including the High Resolution Regional Model (HRM) (the HRM’s information can be found at DWD’s internet link [9], the Consortium for Small-Scale Modeling (COSMO) [10] models from DWD, and the Weather Research and Forecasting (WRF-ARW) model [11] from the National Center for Atmospheric Research (NCAR) are also a useful reference source in the operational heavy rainfall forecast of Vietnam [12]. Despite the predictability of these models in some certain cases, they may fail to predict extreme events because of several reasons. Lorenz [13] pointed out the three main factors causing uncertainties in NWP are the initial conditions, the imperfection of the models, and the chaos of the atmosphere. While the initial condition problem for NWP can be reduced by data assimilation methods, the imperfection of models, which relates to many subgrid processes, can be alleviated by using proper physical parameterizations. For regional weather forecasting centers with limited capabilities in data assimilation and resource computation to provide the cloud resolved resolution forecast, choosing correct physical parameterization schemes still plays the most important role in downscaling the processes in regional NWP models [14].

To illustrate the dependence of heavy rainfall forecast on physical parameterizations, the typical heavy rainfall event relating to the activities of ITCZ over the South China Sea—the East Sea of Vietnam—from 27 to 30 August 2014 in the Bac Bo area was simulated by the WRF-ARW model (see mean sea level pressure analysis of the GFS model in Figure 1(a)). The 3-day accumulated rainfall was mostly from 100 mm to 150 mm, and some stations recorded more than 200 mm such as Kim Boi which recorded 229 mm (Hoa Binh Province; marked with a square in Figure 1(b)), Dinh Lap 245 mm (Lang Son Province, the northeast area; marked with a star in Figure 1(b)), and Tam Dao 341 mm (Vinh Phuc Province, center of the domain; marked with a circle in Figure 1(b)).

Figure 1 illustrates 72 h accumulated rainfall forecasts from the WRF-ARW model at 5 km horizontal resolution, issued at 00UTC 27/08/2014 with different physical parameterization configurations by combining the Betts–Miller–Janjic (BMJ) or Kain–Fritsch (KF) cumulus schemes, the Lin or WRF single-moment three- or five-class (WSM3/WSM5) cloud microphysics schemes, the Dudhia or Goddard shortwave radiation schemes, and the Yonsei University (YSU) or Mellor–Yamada–Janjic (MYJ) boundary layer schemes.

With the KF cumulus parameterization scheme in Figures 1(c)1(f), the results show that extreme rainfall over the northeast area (Lang Son Province) can be predictive with the amount of rainfall over >150 mm but was forecasted overestimation for the mountainous regions over the northwest area (Hoang Lien Son mountain ranges). The extreme rainfall over the center of the domain (Vinh Phuc Province) was not simulated well using the KF scheme for this situation.

For the experiments using the BMJ scheme in Figures 1(g)1(j), the extreme rainfall over the Lang Son Province is underestimated, but cases using Lin or WSM3 for microphysics, Dudhia for shortwave radiation scheme, and the MYJ surface boundary layer (Figure 1(g) and 1(h)) can reduce the overestimation related to Hoang Lien Son mountain ranges in KF scheme experiments. The extreme heavy rainfall over the center of the domain (in Vinh Phuc Province) can be quite well forecasted when using WSM5 with Goddard Hoang Lien Son mountain ranges and the MYJ surface boundary layer (Figure 1(i)). The same configuration in Figure 1(i) but using another surface boundary layer (YSU) scheme provided totally different results (Figure 1(j)) for the southern area of the domain (overestimation).

Many studies have been carried out for validating the effects of physical parameterization schemes in the WRF-ARW model. Zeyaeyan et. al. [15] evaluated the effect of various physic schemes in the WRF-ARW model on the simulation of summer rainfall over the northwest of Iran (NWI). The result shows that cumulus schemes are the most sensitive and microphysics schemes are the least sensitive. The comparison between 15 km and 5 km resolution simulations does not show obvious advantages in downscaling. These investigations showed the best results for both the 5 and 15 km resolutions with model configurations from the newer Tiedtke cumulus scheme, MYJ scheme, and WSM3/Kessler microphysics scheme. Tan [16] processed sensitivity tests with microphysics parameterization of the WRF-ARW model in the concept of quantitative extreme precipitation forecasting for hydrological inputs. About 19 bulk microphysics parameterization schemes were evaluated for a storm situation in California in 1997. The most important finding was that the extreme and short-interval simulated precipitation of the WRF-ARW model, which is very important for hydrological forecast input, can be improved by the choice of microphysics schemes. Nasrollahi et al. [17] showed that different features from hurricane (track, intensity, and precipitation) in the WRF-ARW model can be improved with suitable selections of microphysics and cumulus schemes. These results showed that the best simulated precipitation can be achieved by using BMJ cumulus parameterization combined with the WSM5 microphysics scheme, but the hurricane’s track was best estimated by using the Lin or Kessler microphysics option with BMJ cumulus parameterization. Other validations of parameterization of physical processes on the tropical cyclone of Pattanayak et al. [18] with WRF for the NMM dynamical core showed the important role of cumulus parameterization in track forecasts, thereby contributing to rainfall induced by landfall of tropical cyclones.

In the other aspects related to using cumulus parameterization schemes at high-resolution simulations, Gilliland et al. [19] compared simulations using different cumulus parameterization schemes in summer-time convective activities. Compared with the model simulations at below 5 km horizontal resolution, which resolve the convection, the study showed that depending on the strength of the synoptic-scale forcing, the use of a cumulus parameterization scheme can still be warrant for representing the effects of subgrid-scale convective processes (the Kain–Fritsch scheme in this study).

Thus, the application of regional models to certain regions such as Vietnam will be influenced significantly by local factors (topography and microclimate mode) as well as the effects of physical combinations. Among regional models which are applied in Vietnam, the WRF-ARW model is the most useful tool because of the capabilities of providing different configuration physical/dynamical models for implementation to various scientific communities for both research and operation. For this reason, this study will focus on the impact of some physical parameterization configurations of the WRF-ARW model combining cumulus, cloud microphysics, and shortwave radiation parameterization schemes on heavy rainfall forecast. Two typical cumulus parameterizations (adjustment and mass-flux approaches) will be investigated with the combination of simple to complex cloud microphysics schemes. The verification of dependence of the typical synoptic situations/weather patterns causing heavy rainfall for the Bac Bo area on parameterization schemes is also carried out. For the real-time adaptation forecasting verification purposes, the lateral boundary condition will be used from the Global Forecast System (GFS) of NCEP.

The remainder of this paper is organized as follows: In Section 2, experimental design and validation methods will be presented. Section 3 discusses the impacts of different physical parameterization configurations on heavy rainfall over the Bac Bo area, and conclusions are given in Section 4.

2. Experiments

2.1. Model Description

This study used the recently released version of the Weather Research and Forecasting model with the ARW dynamical core (WRF-ARW model; version with multinested grids and two-way interactive options. The WRF model has been integrated many advances in model physics/numerical aspects and data assimilation by scientists and developers from the expansive research community, therefore becoming a very flexible and useful tool for both researchers and operational forecasters (

For the purpose of investigating the impact of physical parameterization schemes, similarly to the study of Kieu et. al. [20, 21], a set of combination of physical parameterizations has been generated based on (a) the modified KF and BMJ cumulus parameterization schemes; (b) the Goddard and Dudhia schemes for the shortwave radiation; (d) the YSU and MYJ planetary boundary schemes; and (e) the Lin, WSM3, WSM5, and WSM6 schemes for the cloud microphysics. There are a maximum of 32 different configuration forecasts for each heavy rainfall case listed in Table 2. The other options are the Monin–Obukhov surface layer scheme and the Rapid Radiative Transfer Model scheme for longwave radiation. Note that, with the MYJ scheme, the surface layer option will be switched to Janjic’s Eta–Monin–Obukhov scheme which is based on similar theory with viscous sublayers over both solid surfaces and water points. Skamarock et al. [22] provided the detailed description of the WRF-ARW model, and various references for physical parameterizations of the WRF-ARW model can be found from various listed references [11, 2329, 30].

AbbreviationMicrophysicsShortwave radiationBoundary layer

Betts–Miller–Janjic (BMJ) cumulus parameterization

Kain–Fritsch (KF) cumulus parameterization

The WRF-ARW model is configured with two nested grid domains consisting of 199 × 199 grid points in the (x, y) dimensions with horizontal resolutions of 15 km (denoted as d01 domain) and 5 km (denoted as d02 domain). All domains share 41 similar vertical σ levels with the model top at 50 hPa. The higher resolution domain covers the northern part of Vietnam with a time step of 15 seconds. All validations will be carried out with forecasts from the d02 domain. Figure 2 shows the terrain used in d01 and d02 domains.

2.2. Boundary Conditions

The GFS model of NCEP used to provide boundary conditions for the WRF-ARW model in this study has a 0.5-degree horizontal resolution and be prepared every three hours from 1000 hPa to 1 hPa. The GFS data for this study were downloaded from the Research Data Archive at the National Center for Atmospheric Research via website link More information on GFS data can be found at

2.3. Observation Data

The number of observation stations in Vietnam increased from 89 in 1988 to 186 in 2017, with 4 or 8 observations per day (black dots are 8 observations/day for stations in Vietnam and nearby countries in Figure 2), but only 24 stations are reported to WMO. The difference in location and topography results in the significant change from one climate to another; therefore, Vietnam has been divided into several climate zones. The highest station density is in the Red River Delta area (the southern part of northern Vietnam) with approximately 1 station per 25 km × 25 km area. The coarsest station density is in the Central Highlands area (latitude between ∼11 N and 16 N) with approximately 1 station per 55 km × 55 km area. On average, the current surface observation network density of Vietnam is about 1 station per 35 km × 35 km for flat regions and 1 station per 50 km × 50 km for mountainous complex regions. In this paper, in order to verify model forecast for the Bac Bo area, we used observation data from the northern SYNOP stations for the period from 2012 to 2016 listed in Table 1.

In this study, 72 cases of typical widespread heavy rains which occurred in the northern part of Vietnam in the period 2012–2016 were selected. For each case, the forecasting cycles are chosen so that the 72-hour forecast range can cover the maximum duration of heavy rain episodes.

With respect to causes of heavy rainfall events, there are four main categories: (i) activities of ITCZs or troughs: type I, (ii) affected by tropical cyclones or surface low-pressure system (staying at least more than 2 days over the Bac Bo area): type II, (iii) related to the cold surge from the north: type III, and (iv) the complex combinations from different patterns: type IV. The list of forecasted events is given in Table 3 with the station name, the maximum value of daily rainfall, and the type of each heavy rainfall case. The sample number of type I, type II, type III, and type IV has 37, 21, 5, and 9 forecast cycles, respectively.

Forecast cycle (year month day hour)Maximum 24 h accumulation observation (mm)Station with maximum observationTypes of rain events

2012 05 18 0037Bac YenType I
2012 05 19 0044Lac SonType I
2012 05 20 00114Bac QuangType I
2012 05 21 00195Bac QuangType I
2012 05 22 00131Ha DongType I
2012 05 23 00186Quang HaType I
2012 07 20 0041Bac QuangType I
2012 07 21 0024.6Lai ChauType II
2012 07 22 00116Vinh YenType II
2012 07 23 0075VinhType II
2012 07 25 00229Tuyen QuangType II
2012 07 26 00104Ham YenType II
2012 07 27 00112Bac QuangType I
2012 08 03 0058.1Con CuongType I
2012 08 04 0034.1Tinh GiaType I
2012 08 05 0070Lang SonType I
2012 08 06 00153Muong LaType I
2012 08 07 00163Quang HaType I
2012 08 13 0063.6Muong LaType II
2012 08 14 0064.5Nho QuanType II
2012 08 15 0074Do LuongType II
2012 08 16 0076Tam DaoType II
2012 08 31 0049.1Sin HoType IV
2012 09 01 0067Nhu XuanType IV
2012 09 02 00124Quynh LuuType IV
2012 09 03 0084Moc ChauType IV
2012 09 04 00157Huong KheType IV
2012 09 16 00151.4Muong TeType IV
2012 09 17 00104.6Tam DuongType III
2013 06 19 00109Bac QuangType II
2013 06 20 0072.2Huong KheType II
2013 06 21 0059Nam DinhType II
2013 06 22 00218Ha TinhType II
2014 08 25 0091Lao CaiType II
2014 08 26 0089Van LyType II
2014 08 27 00157Hon NguType II
2015 05 18 0090Hoi XuanType III
2015 05 19 0058Sin HoType III
2015 05 20 00106That KheType III
2015 05 21 0072Tuyen QuangType III
2015 06 21 00124Luc YenType II
2015 06 22 0072.7Do LuongType II
2015 07 01 0084.5Cao BangType I
2015 07 02 00123.0Van ChanType I
2015 07 03 0074.0Huong KheType I
2015 07 21 0054.5Hoa BinhType I
2015 07 22 0019Sin HoType I
2015 07 23 0092Sin HoType I
2015 07 24 00180Quynh NhaiType I
2015 07 25 00181Cua OngType I
2015 07 26 00432Cua OngType I
2015 07 27 00347Quang HaType I
2015 07 28 00224Mong CaiType I
2015 07 29 00247Cua OngType I
2015 07 30 00145Quang HaType I
2015 07 31 00239Quang HaType I
2015 08 01 00157Phu LienType I
2015 09 18 0073Hai DuongType IV
2015 09 19 0078.0Dinh HoaType IV
2016 05 20 00118Tinh GiaType I
2016 05 21 00107.0Ha NamType I
2016 05 22 0092Sa PaType I
2016 05 23 00190.4Yen BaiType I
2016 07 24 0060Phu LienType II
2016 07 25 0017Thai BinhType II
2016 07 26 0046Dinh HoaType II
2016 07 30 0018.1Mu Cang ChaiType II
2016 08 01 0019Sin HoType I
2016 08 02 00150Ninh BinhType I
2016 08 10 0081Quynh NhaiType I
2016 08 11 00117Thanh HoaType I
2016 08 12 0087Tay HieuType I

Type I: activities of trough or ITCZ; type II: affected by tropical cyclone or low-pressure system; type III: related to cold surge from the north; type IV: combinations of different patterns.
2.4. Validation Methods

By finding the nearest grids to each station position (listed in Table 1), the daily accumulated rainfall for these heavy rainfall cases from WRF-ARW model forecasts can be assigned. The verification scores used in this study are frequency bias (BIAS), probability of detection (POD), false alarm ratio (FAR), threat score (TS), and equitable threat score (ETS). If we denote H for the hit rate of occurred rainfalls (at a given threshold) for both forecast and observation, M for the missed rate of occurred rainfall forecast, and F for the false alarm rate of the forecast, the BIAS, POD, FAR, and TS are calculated by the following equations:

If we set Hitsrandom = (H + F) (H + M)/T, where T is the sum of H, M, F, and the number of nonoccurred rainfalls for both forecast and observation, the ETS is calculated by

Other meanings of these scores can be found in Wilks’ study [31]. The verification will be carried out for the 5 km domain and for 24 h accumulated rainfall at 24 h, 48 h, and 72 h forecast ranges. Other analysis charts include the histogram of precipitation occurrences at given thresholds (>25 mm/24 h, >50 mm/24 h, and >100 mm/24 h) at the observation stations.

3. Results

3.1. General Performance

The histogram charts (Figure 3) show the number of observations or forecasts that occurred at stations for given ranges—or bins. We divided rainfall into 4 main bins (0–25 mm, 25–50 mm, 50–100 mm, and >100 mm) for different rainfall classes. For all 24 h, 48 h, and 72 h forecasts, it is quite clear that most forecasts from the BMJ scheme are in the 0–25 mm range, higher than the number of observations, while the KF scheme tends to have less forecasts than observations. In contrast, at the thresholds greater than 25 mm, for the BMJ scheme, the number of forecasts is less than the number of observations, while the KF scheme tends to have more forecasts than the number of observations.

Figure 4 shows the BIAS score at different thresholds (>25 mm/24 h and >50 mm/24 h) and separated for KF and BMJ scheme combinations. Overall assessment through the BIAS score is quite similar to the results from the evaluation through histograms: simulations with the BMJ scheme tend to be lower than observations at most forecast ranges and different thresholds (BIAS < 1), while the simulations with the KF scheme tend to be higher than the observations (BIAS > 1). The BIAS tends to decrease significantly when the forecast ranges increase. When the validation thresholds are increased, the simulation with the BMJ scheme tends to decrease BIAS, whereas in combination with the KF scheme, BIAS increases with both the forecast ranges and the evaluation thresholds.

The individual assessment in each combination of BMJ and KF schemes shows that when combined with the Goddard radiation scheme, the BIAS increases from 0.1 to 0.2 compared to the Dudhia scheme. These results are similar for all validating thresholds and for 24 h, 48 h, and 72 h forecast ranges. The difference in simulations with different boundary layer schemes is unclear while evaluating with thresholds below 25 mm/24 h; however, at higher thresholds, it is apparent: the change of BIAS when combining with the KF scheme is much higher. For example, compared to the BMJ scheme, the BIAS score of KF-WSM3-God-MYJ at the >100 mm/24 h threshold at the 24-hour forecast range is 1.6458 and that of KF-WSM3-God-YSU is 2.0833, while BMJ-WSM3-God-MYJ and BMJ-WSM3-God-YSU had approximately equal BIAS scores (0.8229). Thus, the combinations of the boundary layer schemes are different (here only the two schemes YSU and MYJ) with the KF scheme being much more sensitive to its combinations with the BMJ scheme and especially at the high rainfall thresholds. Details of numerical values of BIAS can be found in Tables 4, 5, and 6.

>25 mm/24 h>50 mm/24 h


>25 mm/24 h>50 mm/24 h


>25 mm/24 h>50 mm/24 h


3.2. Skill Score Validation

The charts for the skill scores at the two thresholds in Figures 5 and 6 show that the TS value at the 24 h forecast range is about 0.2 to 0.27 for >25 mm/24 h and 0.1 to 0.2 for >50 mm/24 h. At 48 h, the TS is around ∼0.18 to 0.3 and ∼0.1 to 0.19 corresponding to two thresholds >25 mm/24 h and >50 mm/24 h. At 72 h, the TS is around 0.15 to 0.25 and ∼0.08 to 0.15 corresponding to two thresholds >25 mm/24 h and >50 mm/24 h.

The probability of detection decreases and the false alarm rate clearly increases with forecast ranges and the validating thresholds. In addition, when the threshold increases, the difference between the TS and the ETS decreases which means that the amount of Hitsrandom is too small or the cause of very small hit rate (Hitsrandom rate decreases by 90% when changing the threshold from >25 mm/24 h to >50 mm/24 h).

Specific comparisons between the combinations of the BMJ or KF scheme show that the KF scheme model’s skills in heavy rain forecast in the northern region of Vietnam are better than BMJ scheme model’s skills. The average TS with the KF scheme can be about 15–25% larger than that using the BMJ scheme. If the difference of skills in a regional model is insignificant when changing the physical parameterization schemes, the lateral boundary conditions (from global forecasts) will greatly affect the quality of the dynamical downscaling forecasts after 24 h integration. However, here the skill difference when combining the two different cumulus schemes in the longer forecast range (such as 72 h) shows the importance of convection simulation capability contributing to the forecasting quality of the model. Detailed evaluation of the combination with the radiation physical schemes of KF or BMJ does not show any difference compared to changing the boundary layer schemes when looking at the skill scores TS or ETS.

The combinations with YSU boundary layer schemes have better skills compared to those with MYJ schemes. In addition, when changing the complexity of the cloud microphysics schemes, the more complex the microphysical processing simulation, the better the TS and ETS (at 24, 48, and 72 h forecast ranges and two validating thresholds; see Figure 7 for comparison of the change in TSs and ETSs with the cloud microphysics scheme).

For the skill comparisons of different event types (I, II, III, and IV), Figure 8 shows the TSs at thresholds over 25 mm/24 h and over 50 mm/24 h for 24 h and 48 h forecast ranges. For type I, which is associated with the activity of ITCZ and low-pressure trough over the Bac Bo area, the KF scheme proved its forecast skills in almost forecast ranges and thresholds mentioned in this research. However, with the rain caused by tropical cyclone in type II, the difference between KF and BMJ schemes within 24 h was smaller than that in type I. In 48 h and 72 h, the BMJ scheme showed more skilled forecast with the skill score for threshold 25 mm ranging from 0.25 to 0.35 of the BMJ scheme, compared with 0.2 to 0.3 of the KF scheme, and the skill score for threshold 50 mm ranging from 0.2 to 0.3 of the BMJ scheme, compared with 0.2 to 0.25 of the KF scheme. In type II, both KF and BMJ schemes combined with the simple cloud microphysics Lin scheme showed lowest skill score. For type III, which is associated with the activity of cold surge and its role in squeezing the low-pressure trough from the north towards Bac Bo, the KF scheme was only skillful in threshold 25 mm in 24 h. Particularly in type IV, which contains heavy rain events caused by a complex combination of situations resulting in a trough in Bac Bo, the KF scheme still showed skilled forecast compared to very low forecast (no skill with threshold over 50 mm/24 h and 25 mm/24 h) of BMJ scheme experiments. More details of TSs for different types are listed in Tables 7 and 8.

>25 mm/24 h>50 mm/24 h
Type IType IIType IIIType IVType IType IIType IIIType IV


>25 mm/24 h>50 mm/24 h
Type IType IIType IIIType IVType IType IIType IIIType IV


4. Conclusions

For the purpose of investigating the effects of physical schemes in the WRF-ARW model on the operational heavy rainfall forecast for the Bac Bo area, 32 different model configurations have been established by switching two typical cumulus parameterization schemes (BMJ and KF), the cloud microphysics schemes from simple (Lin) to complex (WSM with 3/5/6-layer closure assumptions), and boundary layer (YSU and MYJ) and shortwave radiation (Dudhia and Goddard) schemes. The 72 experiments of widespread heavy rainfall occurring in the Bac Bo area used boundaries from the GFS model and had the highest horizontal resolution of 5 km × 5 km.

The model verification with local observation data illustrated the limited capabilities in heavy rainfall forecast for the northern part of Vietnam. On average, for the threshold over 25 mm/24 h, the TSs are from 0.2 to 0.25, 0.2 to 0.3, and 0.2 to 0.25 for 24 h, 48 h, and 72 h forecast ranges, respectively. For the threshold over 50 mm/24 h, TSs are 0.1–0.2 for 24 h and 48 h forecast ranges and about 0.1–0.15 for the 72 h forecast range. At above the 100 mm/24 h threshold, a very low skill value (below 0.1) is validated for most forecast ranges.

The change of the microcloud physics from simple to complex closure assumptions shows that complex schemes give very positive results for both the BMJ and KF schemes. The model configurations with the KF scheme showed higher skills compared to BMJ scheme configurations, and these higher skill scores (TS and POD) mainly come from a higher hit rate (H) and a lower missed rate (M) but also a higher false alarm rate (F). More preliminary assessment for the KF scheme configurations showed the most sensitivity of boundary layer schemes compared to microphysics or shortwave radiation schemes and some initial comments that boundary layer interaction during the application of the KF scheme is an important factor to find the appropriate parameters for heavy rainfall forecast over the Bac Bo area in the WRF-ARW model.

In terms of sample size of each category, the first two types are comparable because of the similar sample size. The other two types, however, are limited in sample size and need to be further analyzed in the subsequent research. A detailed assessment related to the origin of the mechanisms causing heavy rain illustrates that the KF scheme showed more skilled forecast with the BMJ scheme during trough- or ITCZ-related heavy rain events, whereas it was less skilled in the events caused by tropical cyclones. With the events caused by the cold surge and a combination of different patterns, the skill of the BMJ scheme was quite low. More verification with the last two types needs to be further investigated in the subsequent research because of the limited sample size studied.

Data Availability

The data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest

The authors declare that they have no conflicts of interest.


This paper describes the results of the National KC.08.06/16-20 Project “Developing an operational heavy rainfall forecast system for the northern part of Vietnam” funded by the Vietnamese Ministry of Science and Technology and the Project VT-CN.04/17-20 funded by the National Program on Space Science and Technology, Vietnam Academy of Science and Technology. L. R. Hole was sponsored by the Norwegian Agency for Development Cooperation and the Norwegian Ministry of Foreign Affairs. The authors would like to thank Ms. Huyen Khanh Luu for her data preparations.


  1. V. T. Phan, T. Ngo-Duc, and T. M. H. Ho, “Seasonal and interannual variations of surface climate elements over Vietnam,” Climate Research, vol. 40, pp. 49–60, 2009. View at: Publisher Site | Google Scholar
  2. V. T. Tran, “Variation of some atmospheric circulation factors affecting Vietnam climate,” Earth Sciences, vol. 27, no. 2, pp. 107–111, 2011. View at: Google Scholar
  3. T. Ngo-Duc, C. Kieu, M. Thatcher, D. Nguyen-Le, and T. Phan-Van, “Climate projections for Vietnam based on regional climate models,” Climate Research, vol. 60, no. 3, pp. 199–213, 2014. View at: Publisher Site | Google Scholar
  4. D.-Q. Nguyen, J. Renwick, and J. McGregor, “Variations of surface temperature and rainfall in Vietnam from 1971 to 2010,” International Journal of Climatology, vol. 34, no. 1, pp. 249–264, 2014. View at: Publisher Site | Google Scholar
  5. K. C. Nguyen, J. J. Katzfey, and J. L. McGregor, “Downscaling over Vietnam using the stretched-grid CCAM: verification of the mean and interannual variability of rainfall,” Climate Dynamics, vol. 43, no. 3‐4, pp. 861–879, 2014. View at: Publisher Site | Google Scholar
  6. S. Yokoi and J. Matsumoto, “Collaborative effects of cold surge and tropical depression-type disturbance on heavy rainfall in central Vietnam,” Monthly Weather Review, vol. 136, no. 9, pp. 3275–3287, 2008. View at: Publisher Site | Google Scholar
  7. T.-C. Chen, M.-C. Yen, J.-D. Tsay, N. T. Tan Thanh, and J. Alpert, “Synoptic development of the Hanoi heavy rainfall event of 30-31 october 2008: multiple-scale processes,” Monthly Weather Review, vol. 140, no. 4, pp. 1219–1240, 2012. View at: Publisher Site | Google Scholar
  8. R. V. D. Linden, A. H. Fink, J. G. Pinto, and T. Phan-Van, “The dynamics of an extreme precipitation event in northeastern Vietnam in 2015 and its predictability in the ECMWF ensemble prediction system,” Weather and Forecasting, vol. 32, pp. 1041–1056, 2017. View at: Google Scholar
  9. F. P. Härter, R. B. Silveira, and G. R. Bonatti, “An assessment of the DFI on DWD-HRM,” Far East Journal of Applied Mathematics, vol. 48, pp. 25–38, 2010. View at: Google Scholar
  10. D. Majewski, “Contributions of the DWD/Germany to SWFDP in Southeast Asia,” in Proceedings of the Meeting the Regional Subproject Management Team (RSMT) of the Severe Weather Forecasting Demonstration Project (SWFDP) in Southeast Asia, World Meteorological Organization, Colombo, Sri Lanka, November 2017, View at: Google Scholar
  11. D. Majewski, “Contributions of the Germany to SWFDP in Southeast Asia,” in Proceedings of the World Meteorological Organization meeting Regional Subproject Management Team in the Severe Weather Forecasting Demonstration Project (SWFDP) for Southeast Asia, Ha Noi, Viet Nam, August 2015, View at: Google Scholar
  12. D. D. Tien, “Lars robert hole, duc tran anh, cuong Hoang duc, and thuy nguyen Ba, verification of forecast weather surface variables over Vietnam using the national numerical weather prediction system,” Advances in Meteorology, vol. 2016, Article ID 8152413, 11 pages, 2016. View at: Publisher Site | Google Scholar
  13. E. N. Lorenz, “Atmospheric predictability as revealed by naturally occurring analogues,” Journal of the Atmospheric Sciences, vol. 26, no. 4, pp. 636–646, 1969. View at: Publisher Site | Google Scholar
  14. A. Pérez-Bello, I. Mitrani-Arenal, O. O. Díaz-Rodríguez, C. Wettre, and L. R. Hole, “A numerical prediction system combining ocean, waves and atmosphere models in the inter-American Seas and Cuba,” Revista Cubana de Meteorología, vol. 25, no. 1, 2019. View at: Google Scholar
  15. S. Zeyaeyan, E. Fattahi, A. Ranjbar, M. Azadi, and M. Vazifedoust, “Evaluating the effect of physics schemes in WRF simulations of summer rainfall in north west Iran,” Climate, vol. 5, no. 3, p. 48, 2017. View at: Publisher Site | Google Scholar
  16. E. Tan, “Microphysics parameterization sensitivity of the WRF model version 3.1.7 to extreme precipitation: evaluation of the 1997 new year’s flood of California,” Geoscientific Model Development Discussions, pp. 1–29, 2016. View at: Google Scholar
  17. N. Nasrollahi, A. AghaKouchak, J. Li, X. Gao, K. Hsu, and S. Sorooshian, “Assessing the impacts of different WRF precipitation physics in hurricane simulations,” Weather and Forecasting, vol. 27, no. 4, pp. 1003–1016, 2012. View at: Publisher Site | Google Scholar
  18. S. Pattanayak, U. C. Mohanty, and K. K. Osuri, “Impact of parameterization of physical processes on simulation of track and intensity of tropical cyclone nargis (2008) with WRF-NMM model,” The Scientific World Journal, vol. 2012, Article ID 671437, 18 pages, 2012. View at: Publisher Site | Google Scholar
  19. E. Gilliland and C. Rowe, “A comparison of cumulus parameterization scheme in the WRF model,” in Proceedings of the 21st AMS Conference on Hydrology, San Antonio, TX, USA, October 2007, View at: Google Scholar
  20. C. Q. Kieu, N. M. Truong, H. T. Mai, and T. Ngo-Duc, “Sensitivity of the track and intensity forecasts of typhoon megi (2010) to satellite-derived atmospheric motion vectors with the ensemble kalman filter,” Journal of Atmospheric and Oceanic Technology, vol. 29, no. 12, pp. 1794–1810, 2012. View at: Publisher Site | Google Scholar
  21. C. Kieu, P. T. Minh, and H. T. Mai, “An application of the multi-physics ensemble kalman filter to typhoon forecast,” Pure and Applied Geophysics, vol. 171, no. 7, pp. 1473–1497, 2013. View at: Publisher Site | Google Scholar
  22. W. Skamarock, J. Klemp, J. Dudhia et al., “A description of the advanced research WRF version 3,” NCAR Technical Note, vol. 475, p. 113, 2008. View at: Google Scholar
  23. S.-H. Chen and W.-Y. Sun, “A one-dimensional time dependent cloud model,” Journal of the Meteorological Society of Japan. Ser. II, vol. 80, no. 1, pp. 99–118, 2002. View at: Publisher Site | Google Scholar
  24. M.-D. Chou and M. J. Suarez, “An efficient thermal infrared radiation parameterization for use in general circulation models,” NASA Technical Memorandum No. 104606, vol. 3, p. 85, 1994. View at: Google Scholar
  25. J. Dudhia, “Numerical study of convection observed during the winter monsoon experiment using a mesoscale two-dimensional model,” Journal of the Atmospheric Sciences, vol. 46, no. 20, pp. 3077–3107, 1989. View at: Publisher Site | Google Scholar
  26. S.-Y. Hong, Y. Noh, and J. Dudhia, “A new vertical diffusion package with an explicit treatment of entrainment processes,” Monthly Weather Review, vol. 134, no. 9, pp. 2318–2341, 2006. View at: Publisher Site | Google Scholar
  27. S.-Y. Hong, J. Dudhia, and S.-H. Chen, “A revised approach to ice microphysical processes for the bulk parameterization of clouds and precipitation,” Monthly Weather Review, vol. 132, no. 1, pp. 103–120, 2004. View at: Publisher Site | Google Scholar
  28. Z. I. Janjic, “The Step–Mountain Eta Coordinate Model: further developments of the convection, viscous sublayer, and turbulence closure schemes,” Monthly Weather Review, vol. 122, no. 5, pp. 927–945, 1994. View at: Publisher Site | Google Scholar
  29. J. S. Kain, “The kain-fritsch convective parameterization: an update,” Journal of Applied Meteorology, vol. 43, no. 1, pp. 170–181, 2004. View at: Publisher Site | Google Scholar
  30. E. J. Mlawer, S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A. Clough, “Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave,” Journal of Geophysical Research: Atmospheres, vol. 102, no. D14, pp. 16663–16682, 1997. View at: Publisher Site | Google Scholar
  31. D. S. Wilks, Statistical Methods in the Atmospheric Sciences, Elsevier Academic Press, New York, NY, USA, 2nd edition, 2006.
  32. T. Fujita, D. J. Stensrud, and D. C. Dowell, “Surface data assimilation using an ensemble kalman filter approach with initial condition and model physics uncertainties,” Monthly Weather Review, vol. 135, no. 5, pp. 1846–1868, 2007. View at: Publisher Site | Google Scholar
  33. S.-Y. Hong and J. O. J. Lim, “The WRF single–moment 6–class microphysics scheme (WSM6),” Journal of the Korean Meteorological Society, vol. 42, pp. 129–151, 2006. View at: Google Scholar
  34. Z. Meng and F. Zhang, “Tests of an ensemble kalman filter for mesoscale and regional-scale data assimilation. Part II: imperfect model experiments,” Monthly Weather Review, vol. 135, no. 4, pp. 1403–1423, 2007. View at: Publisher Site | Google Scholar
  35. UCAR, NCAR, 2019.
  36. The NCEP Products Inventory 2019.
  37. The HRM’s Model User Guide 2019.

Copyright © 2019 Tien Du Duc et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

More related articles

 PDF Download Citation Citation
 Download other formatsMore
 Order printed copiesOrder

Related articles

Article of the Year Award: Outstanding research contributions of 2020, as selected by our Chief Editors. Read the winning articles.