MPIPOM-TC Ocean Model

Biju Thomas,
Isaac Ginis

This project is funded by NOAA JHT and HFIP programs.

A brand new ocean model, called MPIPOM-TC, has been developed at URI and coupled to the HWRF atmosphere under support from Round 1 of HFIP. MPIPOM-TC is a state-of-the-art, MPI version of HWRF’s current ocean model, POM-TC, which can run efficiently on multiple processors (Yablonsky et al. 2013). It is essentially a merger of the existing POM-TC with a community-based MPI version of POM, called sbPOM (Jordi and Wang 2012). Figure 1 details the history of POM, POM-TC, sbPOM, and the newly developed MPIPOM-TC.

Figure 1. Schematic detailing the history of POM from its initial development at Princeton in 1977 to the version transferred to URI in 1994, which ultimately led to the 2012 version of POM-TC in HWRF, and the subsequent merging of the 2012 POM-TC with the community-based 2012 sbPOM to form URI’s new MPIPOM-TC.

The original motivation for this major effort was the realization that the existing POM-TC system (Gopalakrishnan et al. 2012), which can only run on one processor, is insufficient for taking full advantage of the new capabilities that a large, high-resolution POM-TC ocean domain can deliver. The original development plan was to initialize the new MPIPOM-TC with Global HYCOM RTOFS (Mehra et al. 2011; Chassignet et al. 2009). Subsequent testing and validation indicated that Global HYCOM RTOFS might not have been mature enough at that point to replace the currently operational feature-based ocean model initialization in the Atlantic Ocean. Hence, after close collaboration and discussion with the HWRF group at EMC, the decision was made to move forward with developing the MPIPOM-TC system using feature-based initialization in the Atlantic Ocean (Yablonsky and Ginis 2008), with the plug-and-play capability to easily replace the feature-based initialization with Global HYCOM RTOFS (or other ocean initialization options, such as Global HYCOM or NCODA) when and if an alternative ocean initialization product begins to show superior performance in the HWRF.

The grid configuration of MPIPOM-TC consists of a single, new, transatlantic ocean domain, which has ~1/12° horizontal grid spacing. This transatlantic domain replaces the two overlapping POM-TC domains in the Atlantic Ocean, each of which have ~1/6° horizontal grid spacing (Fig. 2). Not only does the transatlantic domain alleviate the need for multiple Atlantic domains, it also extends further eastward than the old POM-TC East Atlantic domain, thereby allowing for ocean coupling of tropical cyclones that are closer to the African or European coasts, as well as the offshore islands. A similar ocean domain has also been developed for the East Pacific.

MPIPOM-TC includes 18 years of community-based updates and bug fixes not present in POM-TC. Unlike POM-TC, MPIPOM-TC is a modernized code with netCDF input and output, allowing for improved interoperability with other systems and other users. Since it is an adaptation of the community-based sbPOM, MPIPOM-TC users and developers can take advantage of existing community support for the majority of its basic architecture.

Overlapping United and East Atlantic POM-TC ocean domains (in the operational HWRF and GFDL models), each of which have ~1/6° horizontal grid spacing (left), and the new, transatlantic MPIPOM-TC ocean domain, which has ~1/12° horizontal grid spacing (right).

EMC ran 1883 forecast cycles from the 2010-2012 Atlantic and East Pacific hurricane seasons using (1) the 2013 operational version of HWRF (H213), which is coupled to POM-TC, and (2) a modified version of H213 in which POM-TC is replaced with MPIPOM-TC (H136). The H136 statistics are encouraging relative to H213, as were there statistics using the 2013 HWRF baseline (Fig. 3).

HWRF intensity bias statistics (left), intensity error statistics (center), and track error statistics (right) from 2010-2012 Atlantic test cases run with either the 2013 baseline with POM-TC (H131) or the 2013 baseline with MPIPOM-TC (H134). HWRF with MPIPOM-TC significantly reduces the positive intensity bias compared to HWRF with POM-TC, while the track differences are negligible. An even larger set of test cases (1883 forecast cycles) with the finalized 2013 HWRF atmospheric configuration (H213) is currently being compared against its HWRF/MPIPOM-TC analog (H136).

One of the key advantages of MPIPOM-TC is that the initial condition module is separated from the model code, allowing for plug-and-play initial condition options. Under support from Round 1 of HFIP, URI has begun to investigate the option of initializing MPIPOM-TC with the daily Global HYCOM, Global RTOFS, or NCODA product (e.g. Fig. 4), with or without GFS SST assimilation (which is used to initialize the HWRF atmosphere), instead of the currently operational feature-based initialization procedure.

MPIPOM-TC SST with surface current vectors (upper-left and upper-middle) and 75-m ocean temperature with current vectors (lower-left and lower-right) in the North Atlantic Ocean on 28 August 2012 using Global HYCOM (upper- and lower-left) or Global RTOFS (upper- and lower-middle) for initialization; upper-right and lower-right panels show the Global HYCOM – Global RTOFS difference field.