Close Window

Univ. of Hawaii: "Application of the NCEP nested regional spectral model to improve weather forecasting in Hawaii"

Final Report


In a collaborative effort among the University of Hawaii, NCEP and NWSFO HNL, the hydrostatic version of the Regional Spectral Model (RSM) with a 10 km resolution was implemented in early 1997 into the operational run stream for the State of Hawaii by NCEP. Nevertheless, the mountainous terrain of the Hawaiian Islands causes large spatial variabilities in local weather which cannot be adequately resolved by the 10 km grid of the RSM.

Kodoma and Juang (1999) evaluated the performance of the RSM forecasts during the 1997 warm (May-September) season against the AVN for four surface sites (Hilo, Lihue, Kahului, and Honolulu). Verification statistics show that the RSM outperformed the AVN at Hilo on the island of Hawaii and Lihue on the island of Kauai but not Kahului on the island of Maui and Honolulu on the island of Oahu. The disappointing results at Kahului and Honolulu are likely caused by the poor representation of the local topography for these two sites. it appears that the mountainous terrain of the Hawaiian islands causes large spatial variabilities in local weather which cannot be adequately resolved by the 10 km grid of the RSM.

The primary objective of this project is to apply the NCEP high resolution (< 3 km), nonhydrostatic mesoscale spectral model (MSM) to improve weather forecasting, especially for high winds and heavy rainfall events, over the Hawaiian Islands. The MSM was first installed on the Dec Alpha 2100A workstation at the University of Hawaii for research purposes and model testing. The MSM is set up for three regions of the State of Hawaii: 3 km resolution for the Hawaii Maui Molokai domain, 1.5 km for the Oahu domain and 1.5 km for the Kauai domain (Chen et a]. 1999). The RSM/MSM was installed on our DecAlpha ES40 with two 667 X111Z CPUs during the summer of 2000. The purchase of ES40 was funded by the NWS Pacific Region. We started the daily experimental forecast for the nested MSM Oahu domain using DecAlpha ES40 since December 2000.


A. Case studies of heavy rainfall and high wind events

We have evaluated the performance of the nested domains against the RSM predictions for a few high wind and heavy rainfall events during the cool season. The results show that high resolution models arc needed to simulate the strong winds on the leeside slopes and orographically induced localized rainfall maxima. it is apparent that our MSM domains have provided better model guidance than the 10 km operational RSM run because the terrain and orographic effects are better resolved in the nested high resolution MSM domains. A few cases are briefly described in this report.

High wind events

During late March of 1998, strong trade winds were observed over the Hawaiian Islands because of a large meridional pressure gradient. During this period, a strong high pressure cell moved to the north of the Hawaiian island chain from the west. We conducted a modeling study of this high wind event on 30 March 1998 for all three model domains. One of the most interesting features in the MSM simulations is that in addition to strong winds around the comers of the islands and within the channels between islands, the orographically enhanced strong winds were simulated on the lee slopes of the mountain ridges with peaks beneath the trade wind inversion.

During 14 15 February 2001, a high pressure cell behind a midlatitude cold front moved to the north of the island chain and generated high winds that knocked down trees and cut power over all major islands of our state. The National Weather Service issued a high wind advisory for all islands in the afternoon of 14 February. Gale warnings were issued for ocean channels between major islands as well. The Alenuihaha Channel between Maui and Hawaii was particularly windy. From the newspaper reports, most of the power failures and damages occurred in the lee sides. Oahu residents reported power failures throughout the, day. The largest area affected was in Kalihi near downtown Honolulu where 6,500 customers were left in the dark after 3:55 pm. The operational 10 km RSM forecast did show strong winds along the northern and southern tips on the island of Hawaii and the Alenuihaha Channel between Maui and Hawaii. Nevertheless, the winds from the 10 kin RSM for the island of Oahu were rather uniform. Similar to other high wind cases, the nested high resolution (1.5 tan) Oahu MSM experimental forecast called for amplification of trade wind flow (>15 m/s) in the lee sides of Koolau and Waianae mountain ranges in agreement with the damage reports. In addition to high winds, strong trade winds also brought in trade wind rain showers especially on the. windward sides and drifted toward the lee sides. The 10 km RSM run predicted no rainfall for Oahu. Nevertheless, appreciable rainfall was predicted along the windward side of the nested high resolution Oahu domain in better agreement with the observed local weather.

Heavy rainfall events

We studied a few heavy rainfall cases including a Kona low that occurred during 3 4 November 1995. Wang et al. (1998) showed considerable improvement of the 10 km RSM run over the global AVN run especially on the mesoscale and synoptic flow patterns. Nevertheless, there were large discrepancies between the observed and predicted rainfall distributions over Kauai because the terrain is not adequately resolved by the 10 km RSM. The maximum 24 h rainfall predicted by the 10 km RSM was about one third of that observed (70 mm vs.250 mm). In addition, the area of heaviest rainfall was simulated over southern Kauai whereas observations showed the maximum over northern Kauai (Wang et al. 1998). Our nested MSM over Kauai with a 1.5 km horizontal resolution simulated maximum 24 h rainfall exceeding 160 mm/day near the ridge tops with appreciable rainfall over northern Kauai in better agreement with observations, It is apparent that the orographic lifting and the localized convergence between the orographically enhanced winds coming down from the ridge top and the airflow moving around the island were not well simulated by the 10 krn RSM.

The recent Hilo flood during 1 2 November 2000 is the most devastating rain storm to hit the area in two decades. Rainfall measurements showed 24 h maxima ranging from 750 940 mm. The AVN run didn't pick up the circulation patterns associated with the remnants of Tropical Storm Paul that provided the low level moisture. The excessive localized rainfall was also completely missed. As it stands, our 3 km nested Maui Hawaii MSM did quite well for a model prediction. The simulated 24 h rainfall accumulation has maxima exceeding 450 mm in the Hilo and Volcanoes Park area in good agreement with the observed rainfall. For this case, in addition to favorable large scale conditions, orographic effects are very important for the development of excessive localized heavy rain showers.

A rare warm season heavy rainfall event occurred in the southeast portion of the island of Oahu between 0800 HST 2000 HST 5 June 2001, with a maximum amount of 3 inches near Manoa Lyon. GOES 10 infrared image at 1530 HST 5 June 2001 shows a mass of deep convective clouds over south Oahu and its adjacent waters. This rare summer thunderstorm filled Oahu's skies with thunder and lightning with heavy rain showers. The 10 km RSM and the 1.5 km MSM simulations for this case were initialized at 1200 UTC 5 June 2001, about 12 hours prior to the heaviest rainfall. The 10 km RSM predicted only about 0.5 to 1.5 inches of rainfall for the island of Oahu between 0800 HST and 2000 HST 5 June 2001 with the maximum rainfall located in the western part of the island. In comparison, the MSM produced 3.3 inches of rainfall along the southeast side of the island in better agreement with observations.

B. RSM/MSM website for experimental daily forecast for the nested Oahu domain

We started the high resolution (~1.5 km) experimental forecast for the island of Oahu in December 2000. The daily 36 h high resolution experimental forecast is available at our website: rsm. We revised our website based on the suggestions from the NWS forecasters.

C. Coupling of the RSN/MSM with an advanced land surface model (LSM) with improved lower boundary conditions

Our tests of the nested MSM domains show that the model performs reasonably well in capturing orographic effects under significant synoptic forcing. Nevertheless, the local circulations and weather under normal trade wind conditions are not well simulated. The surface variables predicted by the RSM agree well with observations in the adjacent oceans. However, over land, the surface air temperature has a cold bias especially in the afternoon hours. As a result, the model consistently underpredicts the local trade wind showers on the slopes in the afternoon hours. The surface winds over land predicted from the MSM are also consistently higher (10-15%) than observed. The f97 version of the nonhydrostatic MSM has only one vegetation type with a constant vegetation fraction of 70%. The land surface is specified as sandy clay loam soil. The albedo and roughness length are interpolated from the global data to the nested domains. It appears that improved surface boundary conditions are needed for the nested MSMs.

A new version of the land surface model (LSM) is implemented to better represent the heterogeneous soil characteristics and vegetation cover over the Hawaiian Islands. Our preliminary results from the coupled LSM/MSM show improved model results within the nested MSM domains over land as compared to the f97 version. For weak trade wind days, the surface temperature over land and the adjacent oceans had a cold bias for the f97 version. It was caused by the fact that the horizontal diffusion terms in temperature and moisture for the f97 version were computed along the sigma surface in the original model. As a result, colder and drier air over the higher terrain within the island interior spread out from the mountains to the adjacent areas under weak wind conditions. This problem has also been corrected by computing the diffusion terms on the height surface and then transforming to the sigma surface.

D. Lessons Learned

i) High resolution (< 3 km) numerical guidance is needed to improve the high wind and heavy rainfall forecasts over the Hawaiian Islands because these events are highly localized in nature due to the presence of complex terrain. Most of these events over the Hawaiian Islands are results of a combination of synoptic forcing and orographic effects.

ii) For summer trade wind showers, it requires accurate representation of the thermal forcing from the surface related to the diurnal heating cycle. The island circulations during the summer months are strongly modulated by the diurnal heating cycle. The summer trade wind rainfall is related to the interactions between the trade wind flow and the island induced circulations. In addition, the model predicted daily maximum and minimum temperatures are sensitive to the land surface properties given in the model. Simulating the full diurnal cycle requires great care in the treatment of surface boundary conditions because of large variations in local microclimate ranging from humid tropical climate on the windward sides to hot desert over bare lava soils with different vegetation cover.

iii) There is a need to perform sensitively tests of precipitation schemes on the simulated rainfall pattern within the high resolution (< 3 km) island scale MSMs with the presence of steep terrain. Our results show that without allowing the condensates to drift with the winds, the model tends to overpredict rainfall on the windward side and underpredict rainfall in the lee side. Some of the condensates generated on the windward side due to orographic lifting may not fall there. They could drift with the winds and grow, resulting in rainfall on the lee side slopes.


Two graduates students (Zhang and Esteban) and Chen attended the RSM short course during the 2nd RSM workshop. The short course presented by NCEP modelers (Juang, Kanamitsu, Hong) is very useful. UH researchers have close research collaboration with NCEP, especially with Drs. Juang and Hong. Dr Juang visited us several times during the grant period and is serving on Mr. Zhang's PhD thesis committee. He attended Mr. Zhang's Ph.D. Comprehensive Examination in March 2002. Dr. Hong visited us in January 2002 to implement the LSM.

Chen participated in the NWS aerosonde experiment and worked closely with NWS and In Situ Group in May 1999. NWS provided working space for the aerosonde project and the NWS assisted in the preparing forecasts for the aerosonde project and determining the flight paths. NWS staff participated in pre project planning with Aerosonde staff and FAA to get flight approval. The data collected were used to compare with model simulations.

Our students also attend the daily weather briefing and gain valuable experience on weather forecasting. One of Chen's graduate students, Ms. Mary Ann Esteban, worked at NWSFO HNL as a student intern. Chen also gave a training session (November 19, 2001) to the forecasters of the National Weather Service on applications of high resolution mesoscale model output to improve weather forecasts.

The primary benefit to the university is the exposure of our students to operational forecasting. Also, the project provides us the opportunity to conduct research related to problems that are important to operational forecasting. Mr. Zhang has the opportunity to interact with the model developers directly because Of OUT collaboration with NCEP. He is working on this project as part of his PhD dissertation. Our students also attend the daily weather briefing and gain valuable experience.

The NWS uses the MSM Web site to view the model output. The NWS has provided feedback to the University on both the model performance and the Web site contents.



Chen, Y. L., Y. Zhang, H. H. M. Juang, S. Y. Hong and K. Kodama, 2002: Application of the coupled land surface model and Mesoscale Spectral Model (RSM) for the Hawaiian Islands. The 4th RSM Workshop. 31 July 2 August, 2002, Los Alamos, New Mexico.

Zhang, Y., Y. L. Chen, I I. M. Juang, 2001: An observational and modeling study of the 3 March 1998 leeside windstorm in the Hawaiian islands. The 3rd RSM Workshop, Central Weather Bureau, Taipei, Taiwan, 23 26 July.

Chen, Y. L., Y. Zhang, H. M. H. Juang, S. Y. Hong, K. Kodama and P. Jendrowski, 2001:

Application of the NCEP nested regional Spectral Model to improve weather forecasting in Hawaii. The 3rd RSM Workshop, Central Weather Bureau, Taipei, Taiwan, 23 26 July.

Zhang, Y. X., Y. L. Chen, H. M. Juang, S. Y. Hong, K. Kodama, and R. Farrell, 2000: Validation and sensitivity tests of Mesoscale Spectral Model simulations over the Hawaiian islands. The 2nd RSN4 Workshop, 17 21 July, MHPCC, Maui, HI.

Kodama, K. and H. M. H. Juang, 1999: An assessment of regional spectral model forecasts for the Hawaii islands. The 1st Regional Spectral Model Workshop, Maui, Hawaii, 9 13 August, 1999.

Chen, Y._L., J. Li, H. M. H. Juang, P. Jendrowski and K. Kodama, 1999: Application of the NCEP nonhydrostatic spectral model to improve weather forecasting in Hawaii. The 1st Regional Spectral Model Workshop, Maui, Hawaii, 9 13 August, 1999.


Frye, J. L. and Y. L. Chen, 2001: Evolution of downslope flow under strong opposing trade winds and frequent trade wind rainshowers over the island of Hawaii. Mon. Wea. Rev., 129, 956 977.

Chen Y. L. and J. Feng, 2001: Numerical simulations of airflow and cloud distributions over the windward side of the island of Hawaii. Part I: The effects of trade wind inversion. Afon. Wea. Rev., 129, 1117 1134.

Feng J. and Y. L. Chen, 2001: Numerical simulations of airflow and cloud distributions over the windward side of the island of Hawaii. Part II: Nocturnal flow regime. Afon, Wea. Rev., 129, 1135 1147,

Li, J. and Y. L. Chen, 1999: A case study of nocturnal rainshowers over the windward coastal region of the island of Hawaii. Afon. Wea. Rev, 11, 2674 2692.

Wang, J. J., H. M. Juang, K. Kodama, S. Businger, Y. L. Chen, J. Partain, 1998: Application of the NCEP regional spectral model to improve weather forecasts in Hawaii. Wea. Forecasting, 13, 560 575.

Wang, J. J., and Y. L. Chen 1998: A case study of Hawaiian trade wind rainbands and their interaction with the island induced airflow. Mon. Wea. Rev., 126, 409 423.


During the first year, the progress of this project was affected by the turnover of the personnel supported by this project. Initially, Dr. Jun Li worked as a postdoc fellow with 50% support from this project. He left in January 1999 to work in the private sector. Mr. Yeh was supported by this project during May June 1999 after he passed his PhD defense to perform RSM/MSM simulations. He accepted a faculty position in Taiwan and departed at the end of June. Mr. Zhang entered our Ph.D. program and joined this project in August 1999.

The progress of this project was also affected by frequent unexplained failures of the new Dee Alpha ES40 server during 2000. It crashed 28 times since it was delivered in early May 2000. At first, the local Compaq service team had difficulties repairing it. Finally, Compaq brought in 2nd and 3rd level trouble shooters from the mainland in October 2000 and replaced our CPUs. The problem was finally fixed in November 2000.

The other problem is the unexpected departure of Dr. S. Y. Hong of NCEP. Dr. Hong is the main person working on the model physics for RSM and MSM. He accepted a faculty position offer from Yonsei University, Korea and left NCEP shortly after the 2nd RSM workshop in the summer of 2000. He was unable to find time to visit us until January 2002 to implement the land surface model on OUT DecAlpha ES40.