Parallel implementation of WRF double moment 5-class cloud microphysics scheme on multiple GPUs

Abstract. In the present study, we have used the Weather Research and Forecasting (WRF) model to simulate the features associated with a severe thunderstorm observed over Gadanki (13.5° N, 79.2° E), over southeast India, on 21 May 2008 and examined its sensitivity to four different microphysical (MP) schemes (Thompson, Lin, WSM6 and Morrison). We have used the WRF model with three nested domains with the innermost domain of 2 km grid spacing with explicit convection. The model was integrated for 36 h with the GFS initial conditions of 00:00 UTC, 21 May 2008. For validating simulated features of the thunderstorm, we have considered the vertical wind measurements made by the Indian MST radar installed at Gadanki, reflectivity profiles by the Doppler Weather Radar at Chennai, and automatic weather station data at Gadanki. There are major differences in the simulations of the thunderstorm among the MP schemes, in spite of using the same initial and boundary conditions and model configuration. First of all, all the four schemes simulated severe convection over Gadanki almost an hour before the observed storm. The DWR data suggested passage of two convective cores over Gadanki on 21 May, which was simulated by the model in all the four MP schemes. Comparatively, the Thompson scheme simulated the observed features of the updraft/downdraft cores reasonably well. However, all the four schemes underestimated strength and vertical extend of the updraft cores. The MP schemes also showed problems in simulating the downdrafts associated with the storm. While the Thompson scheme simulated surface rainfall distribution closer to observations, the other three schemes overestimated observed rainfall. However, all the four MP schemes simulated the surface wind variations associated with the thunderstorm reasonably well. The model simulated reflectivity profiles were consistent with the observed reflectivity profile, showing two convective cores. These features are consistent with the simulated condensate profiles, which peaked around 5–6 km. As the results are dependent on initial conditions, in simulations with different initial conditions, different schemes may become closer to observations. The present study suggests not only large sensitivity but also variability of the microphysical schemes in the simulations of the thunderstorm. The study also emphasizes the need for a comprehensive observational campaign using multi-observational platforms to improve the parameterization of the cloud microphysics and land surface processes over the Indian region.

The low-echo centroid (LEC) storm, characterized by the dominance of warm-rain processes and high precipitation efficiency (Vitale & Ryan, 2013), is generally associated with high-intensity rainfall events in tropical and subtropical regions (Hamada et al. 2015). While many double-moment microphysical parameterization schemes are primarily designed for deep convection or cold-rain processes, research on their performance in simulating LEC warm-cloud precipitation systems is limited. In this study, we investigate an extreme rainfall case that occurred on 20 July 2016 in northern China, resulting in over 600 mm of maximum 24-hour accumulated rainfall. According to radar observations, the case is characterized by LEC structure. Using the Weather Research and Forecasting (WRF) model with 4 km grid spacing, we simulate this event employing the Morrison, Thompson, and Milbrandt-Yau double-moment microphysics schemes. Evaluation based on simulated infrared brightness temperature (BT) using a radiative transfer model and simulated reflectivity reveals that while different microphysics schemes generally predict rainfall amount, location, and propagation accurately, they fail to replicate the three-dimensional cloud structures. The simulated convective cores (>35dBZ) are higher than -10 °C, indicating active cold-cloud processes, while the observed LEC suggests the dominance of warm-cloud processes. The model produces an excessive number of upper-level clouds and overshooting clouds, also overpredict the cloud-top height. Sensitivity experiments show that simulated brightness temperatures are primarily influenced by the concentration of cloud ice particles. Morrison and Milbrandt-Yau microphysics schemes produces an overabundance of cloud ice particles in the upper layer, leading to the overproduction of uppe-level cloud and incorrect representation of cloud top height. Warm-rain processes are not fully developed, and the cold-rain processes are not effectively restrained, resulting in unrealistic cloud structure. By adjusting microphysical processes in the schemes, such as increasing cloud water number concentration, the simulated convective cores align more closely with the observed ones. In summary, while the current microphysics schemes effectively simulate rainfall intensity and propagation, there is a clear need for improvement in simulating the cloud particle distribution and vertical structure of LEC storms. Our findings underscore the importance of refining microphysical parameterization schemes for accurate simulation of extreme rainfall events.

The influence of double-moment representation of warm-rain and ice hydrometeors on the numerical simulations of a mesoscale convective system (MCS) over the US Southern Great Plains has been evaluated. The Weather Research and Forecasting (WRF) model is used to simulate the MCS with three different microphysical schemes, including the WRF single-moment 6-class (WSM6), WRF double-moment 6-class (WDM6), and Morrison double-moment (MORR) schemes. It is found that the double-moment schemes outperform the single-moment schemes in terms of the simulated structure, life cycle, cloud coverage, precipitation, and microphysical properties of the MCS. However, compared with UND-Citation observations, collected during the Midlatitude Continental Convective Clouds Experiment (MC3E), the WRF simulated ice hydrometeors with all three schemes do not agree well with the observations. Overall results from this study suggest that uncertainty in microphysical schemes could still be a productive area of future research from perspective of both model improvements and observations.

Abstract. Italy is a territory characterized by complex topography with the Apennines mountain range crossing the entire peninsula and its highest peaks in central Italy. Using the latter as our area of interest and the snow seasons 2018/19, 2019/20 and 2020/21, the goal of this study is to investigate the ability of a simple single-layer and a more sophisticated multi-layer snow cover numerical model to reproduce the observed snow height, snow water equivalent and snow extent in the central Apennines, using for both models the same forecast weather data as meteorological forcing. We here consider two well-known ground surface and soil models: (i) Noah LSM, an Eulerian model which simulates the snowpack as a bulk single layer, and (ii) Alpine3D, a multi-layer Lagrangian model which simulates the snowpack stratification. We adopt the Weather Research and Forecasting (WRF) model to produce the meteorological data to drive both Noah LSM and Alpine3D at a regional scale with a spatial resolution of 3 km. While Noah LSM is already online-coupled with the WRF model, we develop here a dedicated offline coupling between WRF and Alpine3D. We validate the WRF simulations of surface meteorological variables in central Italy using a dense network of automatic weather stations, obtaining correlation coefficients higher than 0.68, except for wind speed, which suffered from the model underestimation of the real elevation. The performances of both WRF–Noah and WRF–Alpine3D are evaluated by comparing simulated and measured snow height, snow height variation and snow water equivalent, provided by a quality-controlled network of automatic and manual snow stations located in the central Apennines. We find that WRF–Alpine3D can predict better than WRF–Noah the snow height and the snow water equivalent, showing a correlation coefficient with the observations of 0.9 for the former and 0.7 for the latter. Both models show similar performances in reproducing the observed daily snow height variation; nevertheless WRF–Noah is slightly better at predicting large positive variations, while WRF–Alpine3D can slightly better simulate large negative variations. Finally we investigate the abilities of the models in simulating the snow cover area fraction, and we show that WRF–Noah and WRF–Alpine3D have almost equal skills, with both models overestimating it. The equal skills are also confirmed by Jaccard and the average symmetric surface distance indices.

Physically based rainfall retrievals from passive microwave sensors often make use of cloud-resolving models (CRMs) to build a priori databases of potential rain structures. Each CRM, however, has its own cloud microphysics assumptions. Hence, approximated microphysics may cause uncertainties in the a priori information resulting in inaccurate rainfall estimates. This study first builds a priori databases by combining the Tropical Rainfall Measuring Mission (TRMM) precipitation radar (PR) observations and simulations from the Weather Research and Forecasting (WRF) model with six different cloud microphysics schemes. The microphysics schemes include the Purdue–Lin (LIN), WRF Single-Moment 6 (WSM6), Goddard Cumulus Ensemble (GCE), Thompson (THOM), WRF Double-Moment 6 (WDM6), and Morrison (MORR) schemes. As expected, the characteristics of the a priori databases are inherited from the individual cloud microphysics schemes. There are several distinct differences in the databases. Particularly, excessive graupel and snow exist with the LIN and THOM schemes, while more rainwater is incorporated into the a priori information with WDM6 than with any of the other schemes. Major results show that convective rainfall regions are not well captured by the LIN and THOM schemes-based retrievals. Rainfall distributions and their quantities retrieved from the WSM6 and WDM6 schemes-based estimations, however, show relatively better agreement with the PR observations. Based on the comparisons of the various microphysics schemes in the retrievals, it appears that differences in the a priori databases considerably affect the properties of rainfall estimations.

This study is intended to simulate and better understand mean and extreme precipitation over the Loess Plateau (LP) in China using the Weather Research and Forecasting (WRF) model. We performed a long-term (1980–2011) simulation with the WRF model at a 10 km horizontal resolution, forced by the ERA-Interim reanalysis. A series of sensitivity experiments were conducted to investigate the influence of resolution and physical parameterization schemes on the simulation ability of the WRF model over the LP. The region has a dominant semi-arid climate and is adversely affected by extreme precipitation. Results show that the WRF model produced better simulations at a 10 km resolution for the LP with complex terrain than it did at coarser resolutions. WRF simulations for precipitation over the LP were most sensitive to cumulus schemes, moderately sensitive to planetary boundary layer schemes, and less sensitive to microphysics schemes. The WRF model not only adequately captured the spatial distribution of precipitation in the LP but also reproduced its magnitude and variability at different time scales. Moreover, the WRF model captured 73% of the observed extreme precipitation events with an intensity of more than 12 mm day− 1, which cause substantial soil erosion in the LP. Although it was unable to capture 27% of the observed events, the WRF model reasonably reproduced ERA-Interim extreme precipitation and its dynamic processes. These analyses strongly suggest that the biased extreme precipitation generated with the WRF model most likely resulted from inaccurate ERA-Interim forcing data. This study provides a better understanding of extreme precipitation simulations as well as insight into predictions of soil erosion in the LP.

<p>In this study, we used the Weather Research and Forecasting (WRF) version 4.2.1 model to simulate the characteristics of a rainfall-induced landslide that occurred on November 28 in Samigaluh, Kulonprogo. In addition, we investigated 22 different microphysics (MP) schemes to see how sensitive they were. The WRF model was employed with three nested domains, the innermost of which had a 1 km grid spacing and explicit convection. The model was run for 73 hours with GFS initial conditions from 00:00 UTC on November 26, 2018. We used reflectivity profiles from the Weather Radar in Yogyakarta and data from rain gauge stations in Kulonprogo to validate the simulated properties of the rainfall. Despite employing identical initial and boundary conditions and model settings, the MP approaches have significant variances in their thunderstorm simulations. To begin with, practically all of the extreme convection simulation methods over Samigaluh had the same pattern as the reported storm. For example, on November 27, radar data indicated the passage of three convective cores above Samigaluh, which the model in most MP schemes simulated. In comparison, the Ferrier_old scheme did an excellent job of simulating the convective cores' observable features. The MP schemes also had difficulties modeling the storm's updrafts. The Ferrier old scheme simulated surface rainfall distributions closer to data than the other three schemes (Goddard GCE, Morrison2, and WDM5). On the other hand, all four MP systems did an excellent job of simulating the convective variations associated with the thunderstorm. The model's generated reflectivity profiles, which showed three convective cores, were similar to the observed reflectivity profile. These characteristics match the simulated convective profiles, which peaked between 10 and 15 kilometers. The current research reveals that the microphysical systems in thunderstorm simulations have a lot of sensitivity and variability. The study also underlines the necessity for a multi-observational program such as Year of Maritime Continent (YMC) to improve the parameterization of cloud microphysics and land surface processes throughout the Indonesian region. </p>

A field campaign to study cold season fog in complex terrain was conducted as a component of the Mountain Terrain Atmospheric Modeling and Observations (MATERHORN) Program from 07 January to 01 February 2015 in Salt Lake City and Heber City, Utah, United States. To support the field campaign, an advanced research version of the Weather Research and Forecasting (WRF) model was used to produce real-time forecasts and model evaluation. This paper summarizes the model performance and preliminary evaluation of the model against the observations. Results indicate that accurately forecasting fog is challenging for the WRF model, which produces large errors in the near-surface variables, such as relative humidity, temperature, and wind fields in the model forecasts. Specifically, compared with observations, the WRF model overpredicted fog events with extended duration in Salt Lake City because it produced higher moisture, lower wind speeds, and colder temperatures near the surface. In contrast, the WRF model missed all fog events in Heber City, as it reproduced lower moisture, higher wind speeds, and warmer temperatures against observations at the near-surface level. The inability of the model to produce proper levels of near-surface atmospheric conditions under fog conditions reflects uncertainties in model physical parameterizations, such as the surface layer, boundary layer, and microphysical schemes.

Dynamical downscaling is a widely used technique to properly capture regional surface heterogeneities that shape the local hydroclimatology. However, in the context of dynamical downscaling, the impacts on simulation fidelity have not been comprehensively evaluated across many user-specified factors, including the refinements of model horizontal resolution, large-scale forcing datasets, and dynamical cores. Two global-to-regional downscaling methods are used to assess these: specifically, the variable-resolution Community Earth System Model (VR-CESM) and the Weather Research and Forecasting (WRF) Model with horizontal resolutions of 28, 14, and 7 km. The modeling strategies are assessed by comparing the VR-CESM and WRF simulations with consistent physical parameterizations and grid domains. Two groups of WRF Models are driven by either the NCEP reanalysis dataset (WRF_NCEP) or VR-CESM7 results (WRF_VRCESM) to evaluate the effects of large-scale forcing datasets. The simulated hydroclimatologies are compared with reference datasets for key properties including total precipitation, snow cover, snow water equivalent (SWE), and surface temperature. The large-scale forcing datasets are critical to the WRF simulations of total precipitation but not surface temperature, controlled by the wind field and atmospheric moisture transport at the ocean boundary. No significant benefit is found in the regional average simulated hydroclimatology by increasing horizontal resolution refinement from 28 to 7 km, probably due to the systematic biases from the diagnostic treatment of rainfall and snowfall in the microphysics scheme. The choice of dynamical core has little impact on total precipitation but significantly determines simulated surface temperature, which is affected by the snow-albedo feedback in winter and soil moisture estimations in summer.

Investigation of riming within mixed-phase stratiform clouds using Weather Research and Forecasting (WRF) model

The evaluation and usefulness of lightning prediction for the Indian subcontinent are demonstrated. Implementation of the lightning parameterizations based on storm parameters, in the Weather Research and Forecasting (WRF) Model, with different microphysics schemes are carried out. With the availability of observed lightning measurements over Maharashtra from the lightning detection network (LDN), lightning cases have been identified during the pre-monsoon season of 2016–18. Lightning parameterization based on cloud top height defined by a reflectivity threshold factor of 20 dBZ is chosen. Initial analysis is carried out for 16 lightning events with four microphysical schemes for the usefulness in lightning prediction. Objective analysis is carried out and quantitative model performance (skill scores) is assessed based on observed data. The skills are evaluated for 10- and 50-km2 boxes from the 1-km domain. There is good POD of 0.86, 0.82, 0.85, and 0.84, and false alarm ratio (FAR) of 0.28, 0.25, 0.29, and 0.26 from WSM6, Thompson, Morrison, and WDM6, respectively. There is an overestimation in lightning flash with a spatial and temporal shift. The fractional skill score is evaluated as a function of spatial scale with neighborhoods from 25 to 250 km. These high skill scores and high degree of correlation between observations and model simulation gives us confidence to use the system for real-time operational forecast over India. The skill for 2019 and 2020 pre-monsoon are calculated to address the predictability of operational lightning prediction over India. Significance Statement A high-resolution model, namely, the Weather Research and Forecasting (WRF) Model, with multiple microphysics parameterization schemes and lightning parameterization is used here. The objective analysis is carried out for the lightning cases over India and the quantitative performance is assessed. The results highlight that there is fairly good probability of detection (POD) of 0.86, 0.82, 0.85, and 0.84 and false alarm ratio (FAR) of 0.28, 0.25, 0.29, and 0.26 from four different microphysical schemes (WSM6, Thompson, Morrison, and WDM6, respectively). These high skill scores and high degree of correlation between observations and model simulation gives us confidence to use the system for real-time operational forecast. The validation of lightning forecast system deployed over India for five pre-monsoon months in real time is carried out, which gives POD of 0.90, FAR of 0.64, hit rate of 0.57, and POFD of 0.50 for the whole Indian region.

The Lake Champlain Basin has socioeconomic and ecological significance for the Northeastern United States and Quebec, Canada. Temperatures and extreme precipitation events have been increasing across this region over the past three decades. Accurate, high-resolution climate simulations are critical to assessing potential climate change risk in the Lake Champlain Basin. We evaluate the performance of a regional climate model, the Weather Research and Forecasting (WRF) model, to downscale ERA-Interim reanalysis data to 4 km for the Lake Champlain Basin. Specifically, we compare an ensemble of five WRF experiments with different physics configurations using a one-way, triple-nested domain (36, 12, and 4 km) over three 5-year periods (1980–1984, 1995–1999, and 2010–2014) to Daymet, a gridded observational dataset. We find that WRF simulations of the Lake Champlain Basin generally reproduce the observed temperature and precipitation seasonal cycles, but have cold and wet biases. The simulation of mean temperature by WRF is most sensitive to the choice of radiation scheme, while the simulation of mean precipitation is most sensitive to the choice of radiation, cumulus, and microphysics scheme. We find that turning the cumulus scheme on improves the simulation of the precipitation seasonal cycle at a 4 km resolution, but also substantially enhances the wet bias. Using a coarser resolution (36 km) produces smaller regionally averaged precipitation biases, but not improved correlations between simulated and observed monthly precipitation. Both spatial resolution and turning the cumulus scheme off have minor effects on simulated temperature.

Atmospheric wet delay caused by Precipitable Water Vapor (PWV) significantly impacts the performance of many geodetic surveying systems such as Global Navigation Satellite System (GNSS). In this study, we use wet delay corrections forecast by the Weather Research and Forecasting (WRF) model to enhance GNSS Precise Point Positioning (PPP) during two observation periods with two different weather conditions, that is, period 1: March 01 to 14, 2020 (average PWV: 23.5 kg/m2) and period 2: June 02 to 15, 2020 (flooding weather with average PWV: 55.6 kg/m2), over the South China. PWV data from 277 to 263 GNSS stations are assimilated into WRF model to enhance the WRF water vapor forecasting capability for period 1 and period 2, respectively. Wet delay corrections from two different WRF configurations, that is, WRF no data assimilation and WRF with assimilation of GNSS PWV, are used to augment the PPP. Totally, eight WRF‐enhanced PPP schemes are tested. The results show that WRF‐enhanced PPP schemes generally have a better positioning performance in the up component than traditional PPP. After using WRF wet delay corrections, for static mode, the vertical positioning accuracy is improved by 14.6% and 33.7% for period 1 and period 2, respectively. The corresponding convergence time are reduced by 41.8% and 25.0% for period 1 and period 2, respectively. For kinematic mode, the positioning accuracy improvements in the up component reach 13.8% and 19.0% for period 1 and period 2, respectively. The kinematic PPP convergence time is reduced by up to 8.2% for period 1.

Assimilation of accurate external satellite-based water vapor data has been demonstrated to be able to improve the Numerical Weather Prediction (NWP) model’s forecasting performance. In this study, two types of near-infrared Precipitable Water Vapor (PWV) data from the Ocean and Land Colour Instrument (OLCI) of the Sentinel-3A and -3B satellites are assimilated into the Weather Research and Forecasting (WRF) model to enhance the weather forecasting performance over the South China area. One type is the raw PWV products obtained directly from the Sentinle-3A and -3B satellites; the other type is the Sentinel-3A and -3B PWV calibrated by PWV from Global Navigation Satellite System (GNSS). The Back Propagation Neural Network (BPNN) model is used in the PWV calibration. Correspondingly, two WRF assimilation schemes are adopted to assimilate the Sentinel-3 raw PWV and Sentinel-3 calibrated PWV. Additionally, the WRF without data assimilation scheme is used as the background run to investigate the impact of data assimilation. Three WRF experimental schemes are performed for two periods, i.e., February 2020 (dry period) and July 2020 (wet period). The forecasting performance of these three WRF schemes are evaluated by observations from GNSS, radiosonde, and surface meteorological stations. The comparison results with GNSS PWV show that assimilation of Sentienl-3 calibrated PWV can improve PWV forecasting accuracy by 5.6% for February period for the first 6 h after data assimilation. Evaluated by radiosonde profiles, the assimilation of Sentinel-3 both raw and calibrated PWV can improve WRF humidity and temperature forecasting accuracy over the July period. The accuracy improvement can be up to 9.2% (at 2 km to 3 km of altitude) for water vapor mixing ratio and 6.0% for temperature (at 4 km to 5 km of altitude). The WRF rainfall evaluation result has been validated by rainfall observations from meteorological stations and it shows that the assimilation of Sentinel-3 PWV can improve the rainfall forecasting performance over the July period. The rainfall forecasting success rates of three schemes of assimilating no data, Sentinel-3 raw PWV, and Sentinel-3 calibrated PWV are 69.4%, 69.6%, and 71.1%, respectively, over the July 2020 in the South China region.

This paper investigates the accuracy of weather research and forecasting by improving coding for solar radiation forecasting for location in Dili Timor Leste. Weather Research and Forecasting (WRF) model version 3.9.1 is used in this study for improvement purposes. The shortwave coding of WRF is used to improve in order to decrease error simulation. The importance of improving WRF coding at a specific region will reduce the bias and root mean square root when comparing to the observed data. This study uses high resolution based on the WRF modeling to stabilize the performance of forecasting. The decrease in error performance will be expected to enhance the value of renewable energy. The results show the root mean square error of the WRF default is 233 W/m2 higher compared to 205 W/m2 from the WRF improvement model. In addition, the Mean Bias Error (MBE) of the WRF default is obtained value 0.06 higher than 0.03 from the WRF improvement in rainy days. Meanwhile, on sunny days, the performance Root Mean Square Error (RMSE) of WRF default is 327 W/m2 higher than 223 W/m2 from the WRF improvement. The MBE of WRF improvement obtained 0.13 lower compared to 0.21 of WRF default coding. Finally, this study concludes that improving the shortwave code under the WRF model can decrease the error performance of the WRF simulation for local weather forecasting.