NAME Modeling and Data Assimilation:

A Strategic Overview

 

 

 


NAME Science Working Group*

 

 

 

June 2005

 

I.              Introduction

II.            Multi-scale Model Development

III.         Multi-tier Synthesis and Data assimilation

IV.         Prediction and Global-scale Linkages

V.      A Roadmap

 

* The NAME Science Working Group:

 

Jorge Amador1, E. Hugo Berbery2 , Rit Carbone3, Miguel Cortéz-Vázquez4, Art Douglas5, Michael Douglas6, Dave Gochis3, Dave Gutzler7, Wayne Higgins8, Richard Johnson9, Dennis Lettenmaier10, René Lobato11, Robert Maddox12, José Meitín18, Kingtse Mo8, Mitchell Moncrieff3, Erik Pytlak13, Francisco Ocampo-Torres14, Chester Ropelewski15, Jae Schemm8, Jim Shuttleworth12, Siegfried Schubert16, David Stensrud6,Chidong Zhang17

 

1University of Costa Rica, San José, Costa Rica

2Dept. Of Meteorology, University of Maryland, College Park, MD

3National Center for Atmospheric Research, Boulder, CO

4Servicio Meteorológico Nacional, México

5Atmospheric Sciences Dept., Creighton University, Omaha, NE

6National Severe Storms Laboratory, NOAA, Norman, OK

7Earth & Planetary Sciences Dept., University of New Mexico, Albuquerque, NM

8Climate Prediction Center, NCEP/NWS/NOAA, Camp Springs, MD

9Colorado State University, Fort Collins, CO

10University Of Washington, Seattle, WA

11Instituto Mexicano de Tecnología del Agua, Jiutepec, Morelos, México

12University of Arizona, Tucson, AZ

13National Weather Service, Tucson, AZ

14 Centro de Investigación Científica y de Educación Superior de Ensenada

Ensenada, Baja California, México

15IRI for Climate Prediction, LDEO of Columbia University, Palisades, NY

16Data Assimilation Office, NASA/GSFC, Greenbelt, MD 

17RSMAS, University of Miami, Miami, FL

18 VAMOS Support Center, UCAR/JOSS, Boulder, CO

 

I.       Introduction

This document presents a strategic overview of modeling and related data analysis and assimilation components of the North American Monsoon Experiment (NAME). Building on the NAME science plan, a strategy is outlined for accelerating progress on the fundamental modeling issues pertaining to NAME science goals. The strategy takes advantage of NAME enhanced observations, and should simultaneously provide model-based guidance to the evolving multi-tiered NAME observing program.

The overarching goal of NAME is to improve predictions of warm season precipitation over North America. Central to achieving this goal are enhanced observations, and improvements in the ability of models to simulate and predict the various components and time scales comprising the weather and climate of the North American Monsoon System (hereafter NAMS). The specific scientific goals outlined in the NAME Science Plan (http://www.eol.ucar.edu/projects/name) are to promote better understanding and prediction of:

In order to accomplish these goals NAME has adopted a multi-scale tiered approach with focused monitoring, diagnostic and modeling activities in the core monsoon region (Tier I), on the regional scale (Tier II) and on the continental scale (Tier III). It should be emphasized that to be successful, NAME modeling activities must maintain a multi-tiered approach in which local processes are embedded in, and are fully coupled with, larger-scale dynamics.

The NAME region represents a unique challenge for climate modeling and data assimilation. It is a region marked by complex terrain and characterized by a wide range of phenomena including, a strong diurnal cycle and associated land-sea breezes, low level moisture surges, low level jets, tropical easterly waves, intense monsoonal circulations, intraseasonal variability, and continental-scale variations that link the different components of the monsoon. In fact, the NAMS exhibits large-scale coherence in the form of several known phenomena that have an important impact on intraseasonal to decadal time scales. Hence there are building blocks to serve as the foundation for climate forecasting. The El Ni–o/ Southern Oscillation (ENSO) phenomenon is the best understood of these phenomena, but previous research on the NAMS has also identified several others, including the Madden-Julian Oscillation (MJO) and the Pacific Decadal Oscillation (PDO). The relative influences of these phenomena on the warm season precipitation regime over North America are not well understood. Conversely, the large scale convective maximum associated with the monsoon affects circulation elsewhere, as shown by the relationship between the strength of deep convection and the amplitude and location of the summer subtropical High to the west. Similarly, intraseasonal and interannual fluctuations of monsoon rainfall in the Tier-1 region are often out-of-phase with summer rainfall across the central United States; at present the mechanisms for this feather remain unclear

Prospects for improved prediction on seasonal-to-interannual time scales hinge on the inherent predictability of the system, and our ability to quantify the initial states and forecast the evolution of the surface forcing variables (e.g. SST and soil moisture). In the NAME Tier 3 region, circulation anomalies are influenced by SSTs in the tropical Pacific associated with ENSO, as well as in the North Pacific and the tropical and North Atlantic. On decadal time scales, both the Pacific Decadal Oscillation and the tropical and North Atlantic SSTs have influences on monsoon rainfall. SSTs in the eastern Pacific influence tropical cyclone development, which in turn influence the frequency and intensity of moisture surges and attendant rainfall. SSTs in the Gulf of California and the Gulf of Mexico also play a role in modulating the low level circulations associated with the monsoon.

In addition to SST influences, the land surface has many memory mechanisms beyond soil moisture, especially over the western US. Snow extends surface moisture memory across winter and spring. Vegetation in semi-arid regions, which shows pronounced seasonal and interannual variability, acts as an atmospheric boundary condition that affects momentum transfer, radiation, heat and moisture fluxes.

Aerosols are an important atmospheric constituent in southwestern North America. Circulation is often weak and anthropogenic sources from urban areas attenuate and reflect shortwave radiation. Fires (both natural and man-made) and their associated particulates have pronounced seasonal and interannual variability. Dust is an important factor in the spring and early summer, when vegetation is sparse and surface winds are strong. Multi-year droughts may be triggered by SST anomalies, but they are likely to be sustained by land-atmosphere feedback processes that can alter the surface conditions for years and have a significant affect on interannual variability.

The NAME 2004 Field Campaign (Table 1) provided a comprehensive short term (one warm season) depiction of precipitation, circulation, and surface conditions in the core monsoon region The EOP included enhanced networks of radiosondes, pilot balloons, raingauges, wind profilers, radars, and lightning detectors, as well as measurements of ocean fluxes, humidity, soil moisture, and vegetation. A principal goal of the EOP was to provide a sufficiently comprehensive data set to help guide the NAME modeling strategy aimed at improved warm season precipitation forecasts. The elements of the NAME modeling strategy include climate forecast system assessments, climate data assimilation, and climate model and forecast system development

The strategy outlined in the following sections recognizes three distinct, but related, roles that observations play in model development and assessment. These are (1) to guide model development by providing constraints on model simulations at the process level (e.g. convection, land/atmosphere and ocean/atmosphere interactions); (2) to help assess the veracity of model simulations or forecasts of the various key NAMS phenomena (e.g. low level jets, land/sea breezes, tropical storms), and the linkages to regional and larger-scale climate variability; and (3) to provide initial and boundary conditions, and verification data for model predictions. Section II discusses the multi-scale model development strategy. Section III describes how data assimilation plays a vital role in addressing the larger-scale NAMS modeling issues, and section IV discusses the role of coupled ocean-atmosphere-land surface models in addressing the global-scale linkages and the NAMS prediction problem. A Roadmap that summarizes ongoing and planned NAME climate model assessment, climate data assimilation and climate forecast system development activities is given in section V.

II.     Multi-scale Model Development

The underlying premise of the NAME modeling strategy is that deficiencies in our ability to model "local" processes are among the leading factors limiting forecast skill in the NAME region. For the most part, the source of the problem is in the simulation of deep convective processes and their organizing and maintaining mechanisms. While this problem is not unique to the NAME region, the presence of land/sea contrasts and complex terrain make this a particularly challenging problem for NAME. The anticipated Tier I observations are geared to addressing this problem with a specific focus on improving the treatment of:

The interactions with the surface provide, among other things, organization and memory to atmospheric convection so that the problems of modeling land/atmosphere and ocean/atmosphere interactions are intertwined with the deep convection problem. Improvements on these "process-level" issues will require both fundamental improvements to the physical parameterizations, and improvements to how we model the interactions between the local processes and regional and larger scale variability in regional and global models. In short, model development efforts must take on a multi-scale approach. As such, we require information about the NAMS and related variability that extends across all Tiers (I, II, III) and beyond to include global scales.

Development efforts are envisioned that simultaneously tackle these issues from both a "bottomÐup" and a "top-down" approach. In the former, process-level modeling is advanced and scaled-up to address parameterization issues in regional and global modeling, while in the latter, regional and global models are scaled-down to address issues of resolution and the breakdown of assumptions that are the underpinnings of the physical parameterizations. The following expands on the specific issues that need to be addressed.

1) Moist convection in the presence of complex terrain and land/sea contrasts

The key issue for organized convection concerns the difficulties that regional and global models encounter with the so-called scale-separation problem: parameterization schemes and grid-scale convection running concurrently. Parameterization schemes are not designed to cope with this multi-scale behavior. The problem has its roots in the principle of scale separation: the dynamical scale of the process being parameterized should be much smaller than the grid-resolution of the numerical model in which it is applied. In climate models, where the grid-resolution is typically hundreds of kilometers, the scale-separation assumption is sensibly valid. However, parameterizations are not designed to represent either the up-scale effects of convection or its organization. In global numerical weather prediction models the validity of the scale-separation assumption is directly challenged by the mesoscale organization of convection. Recent work shows that surrogate organization (aliased grid-scale circulations) can distort, or take over completely, convective parameterization in global models. This is a long-recognized problem in regional mesoscale models wherein organized convection clearly invalidates the scale-separation principle. A resounding message is that organized convection may occur for the wrong reasons in regional and global weather prediction models.

The violation of the scale separation principle is inevitable when model resolution increases while the convective parameterization remains the same. This is a clear indication that parameterization schemes must be designed for specific resolutions.

The approach taken to addressing this problem should involve a hierarchy of models including fully cloud-resolving models, single column land/atmosphere models, regional mesoscale models, and fully coupled global atmosphere-land-ocean models. Numerical models that resolve convection and have computational domains large enough to explicitly represent interaction with the environment (to the extent practicable with present computers) set a path to basic understanding. Such a multi-scale/multi-model approach will necessarily require close collaboration between process (local), mesoscale and global modeling groups, together with the observational analysts.

These multiscale modeling efforts must be interlinked with the observational efforts of NAME. Tier I will provide detailed observational measurements and forcing. When interactively nested domains or high resolution global models are used, the multiscale models are capable of incorporating Tiers I, II, and III.

 

2) Land/atmosphere interactions in the presence of complex terrain and land/sea contrasts

Observational studies have shown that certain monsoonal circulation regimes have strong persistence tendencies: once "monsoon onset" occurs, an identifiable monsoon regime is maintained for up to several months (depending on latitude). Understanding the mechanism(s) for this persistence is important for making credible seasonal forecasts of monsoon-related climate variables.

A positive surface moisture feedback process has been postulated as playing a critical role in maintaining the monsoon. Vegetation "greens up" dramatically and rapidly across a large portion of the monsoon region following the initiation of rainfall in early summer, providing (via evapotranspiration) a quick pathway for returning precipitation to the atmosphere. The spatial and temporal variations of the water and energy fluxes involved in this feedback are currently poorly measured and simulated. NAME modeling and diagnostic studies should strive to quantify this feedback. Recent improvements to coupled models with advanced land surface models and proper initialization may enhance the ability to simulate these feedbacks.

The persistence of the monsoon is presumably also dependent on large-scale heat sources and sinks, as in the Asian summer monsoon, where onset and maintenance depends on surface sensible heat fluxes over elevated terrain (the Tibetan Plateau). The timing of the monsoon onset and demise is an important issue for the NAMS, as it is for the Asian monsoon. Modeling studies are required to explore the mechanisms controlling the timing of the onset and the demise.

In addition to their theoretical importance, there will be immediate and important social and economic applications of better information on the variability of land surface and hydrologic characteristics in the NAME region. Improvements in understanding the mechanisms that control vegetation, groundwater infiltration and surface runoff are critically important to both U.S. and Mexican efforts to foster sustainable economies and ecosystems in the arid Southwest.

3)    Ocean/atmosphere interactions in coastal regions with complex terrain

 

Monsoons are air-sea-land interaction phenomena, but it is not straightforward to separate out a purely air-sea interaction component because the air-sea interactions are strongly mediated by air-land processes. The predictability of intraseasonal variability of the NAM is heavily tied up in predictions of sea surface temperature (SST) and land soil/vegetation moisture reservoirs - both extreme challenges for todayÕs models. Field campaigns (e.g., PACJET) on the US West Coast have shown that the dynamical influence of steep coastal terrain extends as much as 100 km out to sea for mid-latitude fronts. Monsoons, however, are primarily powered by the land-sea temperature contrast: land has a strong seasonal and diurnal variation while the ocean has a moderate seasonal variation and a relatively small diurnal variation. Furthermore, CAPE is very sensitive to PBL moisture and a 1¡C change in SST produces 2.5 times the moisture in the Gulf of California than off the coast of California. Deep convection over the steep terrain has a strong effect on fluxes over the Tier-I ocean, which in turn influences the evolution of SST. At night, land convection tends to propagate westward over the eastern Pacific. Because the convective clouds block incoming solar radiation and produce cold pool outflows with resulting enhancement of mixing processes, the diurnal timing is critical in determining their cooling effect on the ocean. Models must produce the strength and timing correctly or feedbacks from the nearby oceans will further deteriorate the forecasts. For climate models this is a formidable problem because the land-sea interface is often poorly resolved.

           

            The primary Tier-I air-sea interaction issues for NAME are:

 

To assess and improve model treatment of these issues, NAME 2004 gathered ship-based observations near the mouth of and along the Gulf of California. The mix of observations gathered by the Mexican Navy Research Vessel Altair, the University of Mexico El Puma and the CICESE Research Vessel Ulloa (see the NAME data catalog at http://www.eol.ucar.edu/projects/name) include 4-times daily rawinsonde launches with enhanced launch frequency during IOPs, high resolution wind and precipitation profiler to provide continuous measurements of the wind and drop size distribution characteristics, direct turbulent and radiative flux measurements, and near-surface bulk variables. The principal strategy of this suite of observations is to provide detailed time series of surface fluxes and bulk meteorology combined with profiles of wind, thermodynamics, and precipitation structure for direct comparisons with model output. The air sea interaction observations also allow single-column investigations of model physics, and provide critical validation for convective parameterization schemes. The ships also provide the only continuous high-quality soundings over the local ocean, which are important in defining the time series of mesoscale forcing.

Direct observations and turbulent and radiative fluxes will allow us to examine the efficacy of standard bulk flux routines in a coast zone. Present operational routines are tuned to the open ocean where there is typically an approximate balance between the wind and wave fields. Coastal regions (and particularly semi-enclosed bodies such as the Gulf) are hypothesized to have higher fluxes because the wave field is suppressed. Observations of covariance fluxes, bulk variables, and ocean surface wave fields will allow us to examine this issue (relevant to local sources of moisture and the acceleration of Gulf surges over the water). Finally, very accurate measurements of the surface heat budget and the SST will shed light on the relative roles of surface forcing and oceanic processes (advection and entrainment) in the predictability of SST in the Gulf of California.

These products will be useful for verification and calibration of models and can be used to enhance our understanding of physical processes.

 

4)    The diurnal cycle as a prototype multi-scale problem

A key focus of the NAME modeling effort will be on improving the representation of the diurnal cycle. The diurnal cycle is important to the NAME region for the following reasons:

Current models have difficulty simulating the diurnal cycle so it is an important problem for multi-scale modeling. The NAME 2004 enhanced observations and NAME "Value Added" Products (including mixed layer depth and its diurnal cycle; diurnal evolution of specific humidity, winds, and surface fluxes) can be used to calibrate the satellite measurements, verify model output and increase the general understanding of the diurnal cycle.

As we move beyond Tier I, we need to consider the large regional differences within the broader-scale NAM region including differences in terrain, land surface conditions, and the basic climatology. In particular, efforts should be geared to understanding and improved modeling of the differences between the representation of organized convection in the coastal terrain of NAME, its representation over the Great Plains in the presence of a strong low-level jet, and its representation over the relatively wet land surface conditions of the eastern United States. Here too the diurnal cycle will likely play a central role, especially in terms of its interaction with topography, the land surface, and with the large-scale flow.

Over NAME Tier II the influence of soil moisture and surface fluxes on circulation patterns and precipitation is particularly strong. Models with appropriate land-surface subcomponents and proper initialization are critical. Addressing and verifying such large-scale interactions and regional differences will require that the NAME Tier I observations are put in the context of other in situ and remote observations. This is best accomplished through data assimilation (including land data assimilation) as discussed in the next section.

III.    Multi-tier Synthesis and Data assimilation

The observations obtained from the NAME 2004 field campaign should provide valuable new insights into the mechanisms and phenomena of the monsoon in the Tier I region and, as outlined in the previous section, help to improve the representation of key physical processes in models. Nevertheless, in order to pursue a true multi-scale modeling strategy, we require information about the monsoon that extends well beyond the Tier I region. In this section, we discuss the role of data assimilation in enhancing the value and extending the impact of the Tier I observations to allow addressing issues of model quality and monsoon variability on scales that extend across the greater NAM region. In addition, data assimilation can provide an important framework for quantifying the impact of observations, and for assessing and understanding model deficiencies.

The basic goal is the creation of the best possible research quality assimilated data sets for studying the NAM region and its interactions with the large-scale environment. It is expected that this effort will rely primarily on regional data assimilation systems with some limited work done with global systems. The former have the potential to provide high resolution, and spatially and temporally coherent (compared with the Tier I observations alone) estimates of the various NAMS phenomena such as Gulf surges, low level jets, and tropical easterly waves, while the latter provide information (at a somewhat lower resolution) about linkages between the greater NAMS and global-scale climate variability and the role of remote boundary forcing, and will be discussed more in section IV. Additionally, we anticipate that off-line land data assimilation systems, as well as, simplified 1-dimensional land/atmosphere and ocean/atmosphere data assimilation systems will provide invaluable "controlled" environments for addressing issues of land-atmosphere and ocean-atmosphere interactions and model errors.

Specific examples of data sets to be generated include a series of assimilations for North America covering the EOP both with and without the NAME 2004 enhanced observations. Since the data assimilation products are model dependent and coarse resolution models will not resolve the Gulf of California, the data impact studies will be performed using high resolution global models and regional models. If observations are found to improve monsoon forecasts or simulations, recommendations will be made to continue such data collection beyond 2004. Parallel simulations, obtained by nesting regional models in the analyses both with and without the NAME 2004 enhanced observations will also be performed. Here efforts should also take advantage of existing operational and special reanalysis data assimilation products.

 

Some specific objectives of these studies are:

 

1) To better understand and simulate the various components of the NAM and their interactions:

 

These components are crucial to our understanding of the seasonal evolution of the monsoon. They may also help explain the out of phase relationship between precipitation in SW North America and that in the US Great Plain.

 

2) To quantify the impact of the NAME observations

Specifically, the aim is to assess the impact of the NAME observations on the quality of the analyses. The impact on predictions is one measure of quality, though that will be addressed in the next section. Improvements to the analyses can come about directly through the assimilation of the observations, or indirectly through improvements in the models used in the assimilation systems. As such, it is also important to understand the extent to which model errors impact the quality of the assimilated data.

To a large extent quantifying improvements will require comparisons with independent data that has not been assimilated. For this purpose, the NAME community has compiled additional precipitation products and satellite products that cover the period of the NAME 2004 field campaign. All of the products are in a useful format for quantitative applications (as opposed to .gif files, pointers, or raw measurements). Many of the products are subsets for the NAME domain at high spatial and temporal resolution. The precipitation products include: