Main Page

From MapModeling

The Modeling, Analysis, and Prediction Program Community

Here is the cheatsheet for fast starts. Cheatsheet for Fast Editting.

Here is a link to mapmodeling.org help and the more complete information from wikimedia: [mapmodeling.org help]

Contents 1 Introduction 1.1 Here is a skeleton for contributions to the page. 1.2 Who is the audience? 2 Development of MAP Community Science Directions 2.1 Hindcasts and Predictions 2.2 Polar Environments 2.2.1 Science questions and Research Directions 2.3 Weather and Climate 2.4 The Current and Future Role of Dynamical Cores in Model Development 2.4.1 Background 2.4.2 Ongoing activities 2.4.2.1 Strategic partnership between NASA, NCAR and GFDL: The Finite volume dynamical core 2.4.2.2 Evaluation of dynamical cores 2.4.3 Proposed future research directions 2.4.3.1 Building bridges between global and regional climate modeling activities: Unified dynamical cores 2.4.3.2 Tackling the scientific and computational challenges 2.4.3.3 Unified approach to model evaluations: Standard test series and evaluation criteria 2.4.4 Importance to NASA’s Modeling, Analysis and Prediction Program (MAP) 3 MAP Projects 3.1 Standing Projects 3.1.1 GMI: Global Modeling Initiative 3.1.2 MERRA: Modern Era Retrospective-analysis for Research and Applications 3.1.3 CMAI: Cloud Modeling and Analysis Initiative 3.1.3.1 Background 3.1.3.2 Ongoing activities 3.1.3.2.1 a. The MAP Climatology of Mid-latitude Storminess (MCMS) 3.1.3.3 Importance to MAP 3.1.4 Ent Dynamic Terrestrial Ecosystem Model 3.2 Closed-ended Projects 3.2.1 MAP05 3.2.2 MAP06 3.2.3 Summer 2007 3.2.3.1 SNAPA and FastJX 4 New Science Directions 4.1 Modeling atmospheric layers 4.2 Lightning 4.2.1 A short description 4.2.2 Existing activities 4.2.3 Why this is important to MAP 4.2.4 What other subject areas are relevant? 4.2.5 Research Directions 4.3 Why Model Validation is important to MAP 4.4 Next Generation Model Parameterizations 4.5 OSSEs and Optimal Observing System Design 4.6 Understanding Processes that Control Tropospheric Composition Above 5 km 4.6.1 Background 4.6.2 Ongoing activities 4.6.3 Importance to MAP and other fields 4.6.4 Directions for Development 4.6.5 Summary 4.7 Global Change in Atmospheric Chemistry, Composition, and Climate: Hindcasting Trends and Variability since 1980. 4.7.1 Goals 4.7.2 Experimental Approach 4.7.3 Analysis & Diagnostics 4.7.4 Additional notes 5 Information Technology and Computational Infrastructure 6 Community-Supported GEOS5 Users' Guide & Tutorial 7 Help Using the Wiki Software

[edit] Introduction

Welcome to the MAP Modeling Community. This wiki will be used to develop documents that represent the community and to organize the community.

MAP: The Modeling, Analysis, and Prediction (MAP) Community is organized around the investigators in NASA's program in Earth system modeling. This effort is focused primarily on modeling the global environment, and has at its foundation evaluation and use of observations from NASA's Earth-observing satellites. The effort includes climate and weather modeling, and therefore, modeling of all of the major components of the Earth system - atmosphere, ocean, land, cryosphere, chemistry, aerosols.

NASA satellites take many measurements of the Earth. These data are essential for understanding weather, climate, and climate change. The MAP program is a critical element of this observing program. The models and observations work synergistically. The observations are used in model evaluation, and model forecasts can also be used in quality assessment of new observations. A unique focus of MAP is research in data assimilation, which formally melds information from observations and models.

More detailed information about the MAP Community is found in the Community Portal.

[edit] Here is a skeleton for contributions to the page.

For each subject area:

1. A short description of the subject area.
2. An inventory of existing activities, with links to primary sources of information.
3. Why is this important to MAP, or to modeling in general?
4. What other subject ares are relevant?
5. What are the important directions for development?
6. A paragraph, perhaps an abstract, that summarizes the research directions.

[edit] Who is the audience?

1. Each other
2. To provide information for program managers who want to know what is in MAP - both inside and outside of NASA.
3. To provide continuity of program from one program manager to the next.
4. Suppose you were asked to provide or find a paragraph of important research directions, this would be a good place to come.

[edit] Development of MAP Community Science Directions

The March 2007 MAP Program Meeting was held in College Park, Maryland. There were three science focus areas. Associated with each of these focus areas, MAP investigators presented some recent results and led discussions to help define future research directions. This wiki is the organize and extend those discussions.

[edit] Hindcasts and Predictions

MAP Meeting Presentation (Prather/Logan) Past as Prologue

[edit] Polar Environments

MAP Meeting Presentation (vanderVeen) Community Modeling

MAP Meeting Presentation (Bromwich) Arctic and Antarctic Modeling

MAP Meeting Presentation (Curry) Atmospheric GCM's at High Latitudes

[edit] Science questions and Research Directions

1. Polar environments.

1.1 One of the major issues that arose from IPCC AR4 was sea-level rise related to melting of Greenland and the West Antarctic Ice Sheet.

1.1.1 Are we seeing natural variability or are we seeing more rapid melting of the ice sheets than expected? 1.1.2 Have we identified and can we represent the dynamics of melting of large ice sheets and glaciers? 1.1.3 At what level must land-ice and ice-sheet models be interactive with climate models to make useful predictions? 1.1.4 What is the role of enhanced snow build up in the interior of Greenland and East Antarctica, etc. in the polar ice balance? 1.1.4.1 What are the fundamental time scales?

1.2 Sea ice and sea ice changes

1.2.1 Are we seeing natural variability or are we seeing more rapid melting of the sea ice than expected? 1.2.1.1 What is the difference between the northern and southern hemisphere? 1.2.2 What are the changes in ocean circulation associated with the melting of sea ice? 1.2.3 What are the changes in ocean ecosystems associated with the melting of sea ice? 1.2.4 What are the changes in terrestrial ecosystems associated with the melting of sea ice? 1.2.4.1 Is there relevant information from the paleo record when there was an ice free Arctic? 1.2.5 What are the changes in the radiative balance and the global and regional consequences?

1.3 High-fidelity modeling of the North Atlantic: The North Atlantic ocean is one of the most critical regions to model correctly for the climate problem. The fidelity of the simulation is highly sensitive to resolution, bottom topography and viscosity. The freshening of the ocean water from the possible melting of the Greenland ice sheets impacts the fluid dynamics and the ecosystems. The formation of bottom water, the moderation of European temperatures, and the integrity of sea ice simulations are all related to the circulation in the North Atlantic. The observational-foundation and the simulation-foundation need to be a research focus.

Key Points for Studies of the Cryosphere

Future NASA MAP studies encompassing the cryosphere should emphasize the synthesis of multiple data sources and algorithms including numerical modeling to achieve a comprehensive picture of polar climate, especially the accelerating change in the Arctic, to understand and quantify the roles of anthropogenic and natural components. The Antarctic region, where dramatic local changes are accompanied by large regions of relative climate stability, stands a useful contrast to the Arctic. These studies should incorporate the plethora of modern observations from NASA polar-orbiting satellites, especially since the start of the EOS era. Furthermore, space-borne data should be enhanced with ground-truth from field studies, including surface-based investigations and airborne campaigns. Such field campaigns can include platforms such as lidar, radar, sodar and microwave sounders. The satellite-based observations can be expanded by the addition of new GPS-occultation based techniques. Polar regional reanalyses of the coupled atmosphere-ocean-sea ice-land system provide a high resolution synthesis (in space and time) for separation of anthropogenically-forced change from natural variability, and need further development.

Studies should target the hydrologic cycle, including clouds, precipitation, atmospheric water vapor, ice-crystal structure, blowing snow, and atmospheric moisture transport. Improved detection of phenomena such as clear sky precipitation (aka diamond dust) is desirable. Furthermore, cloud-clearing algorithms need refinement in the polar regions. These refinements are needed to enhance studies of the surface mass balance of the Greenland and Antarctic ice sheets (as well as other ice reservoirs) and their contribution to global sea level rise.

Numerical modeling studies are an important component of NASA MAP research; studies that combine the various observations through data assimilation offer the promise of eliminating the oft-cited polar data void. The assimilation algorithms need to be improved and optimized for the polar oceans and polar land surfaces. Observing System Sensitivity Experiments (OSSE) are a natural component to this work. Improved land surface data assimilation (LSDA) is especially important for the polar regions. Remote sensing techniques should be applied to yield better descriptions of sea ice, such as snow cover, surface albedo, and meltpond coverage. Snow cover, surface albedo, surface temperature and surface moisture should be assimilated for polar land areas. Modeling studies can also diagnose the local and global teleconnections associated with polar climate variability and change. Furthermore, remote sensing observations should be increasingly applied to model verification.

[edit] Weather and Climate

MAP Meeting Presentation (Schubert) Overview of Data Assimilation for Weather and Climate

MAP Meeting Presentation (Bosilovich) Reanalysis Data Sets

MAP Meeting Presentation (Gelaro) Basic Challenges for Climate Data Assimilation

MAP Meeting Presentation (Riishoijgaard) Observing System Simulation Experiments

[edit] The Current and Future Role of Dynamical Cores in Model Development

[edit] Background

The dynamical core of an Atmospheric General Circulation Model (GCM) is the central component of every climate and weather prediction model. It encompasses the numerical methods used to solve the equations of motion on the resolved scales. Research in dynamical cores faces many scientific and computational challenges. First, there is a high demand for numerical accuracy, consistency and built-in physical conservation laws. Examples are the conservation of mass, energy and enstrophy as well as physics-based approaches to the subgrid-scale diffusion and mixing mechanisms. These aspects have become increasingly important for future-generation Earth System Models that now include atmospheric processes such as the online transport and reactions of chemical constituents and the carbon cycle. Second, the anticipated increases in grid resolution put strong demands on the computational efficiency of the numerical algorithms. This is especially true on massively parallel computer architectures that now incorporate hundreds of thousands of processors. The trend in parallel computing architectures necessitates a paradigm shift in the dynamical core designs. This includes highly scalable numerical methods on non-traditional computational grids on the sphere. Examples of alternative grids are icosahedral and cubed-sphere meshes, that are both free of grid singularities, as well as Adaptive Mesh Refinement (AMR) techniques. The latter bridges the gap between global-scale and regional/local-scale climate assessments. In addition, the next generation dynamical cores will need to foster switches in the dynamics equation sets. This offers choices between hydrostatic and non-hydrostatic modeling approaches and unifies the dynamical core research across scales.

[edit] Ongoing activities

[edit] Strategic partnership between NASA, NCAR and GFDL: The Finite volume dynamical core

There is a strong partnership between NASA’s Goddard Space Flight Center (GSFC), the National Center for Atmospheric Research (NCAR) and the Geophysical Fluid Dynamics Laboratory (GFDL). The partnership fosters the maintenance and development of the state-of-the-art hydrostatic Finite Volume (FV) dynamical core. The hydrostatic FV dynamical core was originally developed at NASA & GFDL (Lin, S.-J., 'A "Vertically-Integrated" Finite Volume Dynamical Core for Global Models', MWR 2004) and is now being employed operationally at all three research institutions. Recently, a non-hydrostatic shallow-atmosphere FV dynamical core has been developed at GFDL, and both NASA and GFDL currently support the FV research on cubed-sphere computational meshes. In addition, there is a strong collaboration between NASA and NCAR. The goal is to closely coordinate the dynamical core modeling efforts in NASA’s GEOS-5 model and NCAR’s Community Atmosphere Model CAM. The application areas of the FV dynamical core include weather prediction, climate modeling, and the transport of trace constituents. In summary, the current FV research activities 1) investigate the characteristics of the present implementations, 2) develop advanced versions of the dynamical core, and 3) explore implementations on new computational grids.

[edit] Evaluation of dynamical cores

The objective evaluation of dynamical cores is not only a key component for future model developments but also for the decision processes at research institutions. Over the last few years, active research has been pursued by the University of Michigan and NCAR. The research targets new dynamical core test cases and objective evaluation criteria that not only assesses the correctness and accuracy of solutions but also takes the uncertainty of reference solutions into account.

[edit] Proposed future research directions

[edit] Building bridges between global and regional climate modeling activities: Unified dynamical cores

Traditionally, the development of weather and climate models had been considered two separate disciplines. The distinction was primarily motivated by the large differences in the target grid resolutions and science missions that necessitated very different physical parameterizations and numerical approximations for the equations of motion. This distinction is no longer obvious. With the ever increasing computing resources the difference in the resolutions is rapidly narrowing, bringing the local-scale and global-scale modeling communities closer together than before. Therefore, many climate research and weather forecasting institutions worldwide are considering or have already initiated the development of unified weather and climate models. The goal is to develop dynamical cores that can be used across all scales for cloud-resolving-, meso-, regional- and global-scale applications.

[edit] Tackling the scientific and computational challenges

1. Fully compressible, non-hydrostatic equation sets: Future application areas of dynamical cores will span both large-scale climate regimes as well as meso-scale, or even cloud-scale, processes. The latter require spatial resolutions of a few kilometers or smaller. Such resolution demands invalidate the hydrostatic approximation. Therefore, future dynamical cores need to be built upon the non-hydrostatic equation set, most desirably with switches for shallow atmosphere, deep atmosphere and hydrostatic atmosphere configurations.
2. Local grid refinement capability: A dynamical core with built-in adaptive mesh refinement (AMR) techniques offers an attractive solution to regional-scale modeling questions. It enables the modeler to consistently embed high-resolution areas in the global domain. Such an approach avoids boundary data inconsistencies, present in traditional nested models, and allows the user to flexibly focus the model resolutions over pre-defined areas or dynamic features of interest. Examples of high-resolution domains are mountainous terrain or land-sea boundaries.
3. Horizontal grid uniformity: A dynamical core with a quasi-uniform computational grid is highly desirable. It not only avoids grid singularities at the pole points, but also removes filtering operations closely linked to the convergence of the meridians in traditional latitude-longitude meshes. If local, grid-point-based numerical methods are chosen, the uniform grid furthermore accelerates the scalability on parallel computer architectures. Examples of quasi-uniform grids are icosahedral, cubed-sphere, Yin-Yang and Fibonacci grids. We suggest focusing on cubed-sphere grid configurations with finite-volume based numerical methods. The research should be done in close collaboration with NASA and GFDL who recently adopted the cubed-sphere geometry in a variant of the FV dynamical core.
4. Shape-preserving, mass-conservative advection and subgrid-scale mixing: Advection schemes are not only used as a fundamental building block in dynamical cores but also for tracer transport processes. They demand a consistent treatment of the advection process for both air and tracer constituents. Mass conservation, low diffusivity and monotonicity (shape) preservation are the key aspects for accurate advection algorithms. The latter two are for example paramount for water substances that exhibit sharp gradients in their spatial distributions. Shape preservation suppresses spurious numerical noise, such as overshoots and undershoots in the distribution, but on the other hand adds numerical diffusion. Such mixing processes need to be critically reviewed. The goal is to motivate mixing processes by truly physical phenomena and avoid empirically tuned approaches to subgrid-scale diffusion mechanisms. Perhaps also greater emphasis on Lagrangian accuracy is needed, while not sacrificing the Eulerian conservation.
5. Conservation properties: The continuous, adiabatic and inviscid equations of motion satisfy an infinite number of conservation laws. They include the conservation of mass, total energy, tracer variance, enstrophy and potential vorticity. From a physical viewpoint, a dynamical core should possess the discrete analogues of the continuous conservation properties. However, the discrete equations can only strictly enforce a subset of them and only in certain combinations. Questions about the suitable combinations of invariants arise that are most beneficial for climate applications.
6. Grid staggering, vertical coordinate and computational modes: There are many options for the vertical and horizontal positions of the prognostic variables, but no choice is optimal for all possible motions. Divergent motions, such as gravity waves, are best supported in a C-grid configuration which is, on the other hand, unfavorable for rotation-dominated processes. The latter are best represented by a D-grid arrangement. In addition, time-stepping algorithms can have computational modes that need to be controlled. This triggers questions about the choice and impact of suitable grid staggerings in multi-scale dynamical cores. Furthermore, the options for the vertical coordinate need to be reviewed. This addresses their characteristics and suitability for stratospheric and mesospheric applications.

[edit] Unified approach to model evaluations: Standard test series and evaluation criteria

Tests of atmospheric GCMs and, in particular, tests of their dynamical cores are important steps during all phases of the model development. They reveal the influence of an individual model design on atmospheric simulations and indicate whether the circulation is described representatively by the numerical approach and computational grid. However, testing a global 3D atmospheric model is not straightforward. In the absence of non-trivial analytic solutions, the model evaluations most commonly rely on intuition, experience and model intercomparisons. Standard test cases for model evaluations are a valuable tool that enables the modeling community to objectively compare the dynamical cores. To date, very few standard test cases for 3D dynamical cores on the sphere exist. Therefore, we propose building an idealized hierarchical test suite that furthermore addresses quantitative measures for accuracy and computational efficiency. Building such a test suite will be a community effort.

[edit] Importance to NASA’s Modeling, Analysis and Prediction Program (MAP)

The MAP program can play a lead role as an integrator and facilitator that coordinates and supports the Finite Volume modeling efforts at NASA, NCAR and GFDL. It thereby prepares NASA and the community for the next-generation dynamical cores that are likely based on non-traditional computational meshes with local grid-point-based numerical schemes. These will be highly scalable on parallel computer architectures. MAP’s support will also be fundamental in the development of the standard test suite for 3D dynamical cores. It will help build the community effort and initiate a dynamical core intercomparison project.

Christiane Jablonowski (University of Michigan, cjablono@umich.edu) and Peter H. Lauritzen (NCAR, pel@ucar.edu), October/30/2007

[edit] MAP Projects

The MAP Program periodically sponsors “projects” that include integration across models, super-computing, organizations, and NASA and non-NASA systems infrastructure. In both 2005 and 2006 the MAP Program participated in assessing how the GEOS-5 model performed for hurricane tracking. In addition, in 2006 the MAP '06 Project participated in testing an interface between the GEOS-5 global model and the WRF regional model,and provided that information to the NAMMA field experiment.

The MAP program has two types of Projects - Standing Projects and Closed-ended Projects.

[edit] Standing Projects

The MAP Program funded a number of activities to deliver specific program products, e.g. assessment of the impact of trace gases on the atmosphere. In addition there are a number of community coordinating activities that have been encouraged and sponsored by the program office.

[edit] GMI: Global Modeling Initiative

Link to GMI Homepage GLOBAL MODELING INITIATIVE (GMI)

1. Background

The Global Modeling Initiative was originally constituted to address the need for three-dimensional assessment of the impact of supersonic and subsonic aircraft. Its philosophy highlights two basic aspects: a) integration of a "modular" chemistry-transport model that can incorporate different input and/or algorithms to examine the reasons for divergence in model results and contribute to our understanding of uncertainty in assessments; b) institution of a Science Team to contribute different inputs, algorithms and analysis tools, and to endorse the model results. The current GMI model is maintained by the Software Integration and Visualization Office (SIVO) at NASA/Goddard Space Flight Center. The model has been driven with several different meteorological fields, both from free-running GCMs and analysis, and various degrees of complexity: tropospheric chemistry, stratospheric chemistry, or combined (Combo) tropospheric/stratospheric chemistry. Modules to carry out simulations of aerosol transport and dynamics, as well as aerosol cloud activation have also been included.

2. Ongoing activities

a, The "Combo" version of GMI has proven to be an excellent modeling tool for the analysis of satellite and aircraft data, particularly in the region of the Upper Troposphere/Lower Stratosphere (UT/LS). Very good agreement has been found with results from A-Train satellites in this region, and performance in the troposphere is being evaluated. Thus, the GMI Combo model provides a validation tool for the chemical mechanisms, emissions, and other processes that would form part of coupled chemistry-climate models.

b. The components of the GMI Combo model have been made ESMF compliant as part of an effort to incorporate the GMI chemistry, emissions and deposition parameterizations into the GEOS-5 system for use in coupled chemistry-climate studies.

c. Simulations are being carried out for the AURA period using the GEOS-4 assimilations with a new lightning parameterization and up-to-date emission inventories. These simulations will be made available to the community through a data portal.

d. Simulations have also been carried out for the Hemispheric Transport of Air Pollutants comparison, both for aerosols and gas-phase species. Results of these simulations are available, and have also been submitted to HTAP. Additional aerosol simulations were also carried out for the AEROCOM project.

e. Simulations are also being carried out to test the sensitivity of calculated aerosol mass loadings to different parameterizations of emissions and particle growth. Such simulations are a first step to understand the sensitivity of the calculated aerosol loading in the troposphere to different processes.

f. The cloud activation module has been utilized to explore the sensitivity of the first indirect effect to adopted meteorological fields and cloud parameterizations.

g. The Finite Volume advection algorithm utilized in the GMI model is being utilized in CO2 simulations with increasing spatial resolution, and compared to results using the Second Order Moments in order to ascertain whether both algorithms converge to the same solution.

More details on the above activities can be obtained from the GMI website, or from Jose M. Rodriguez@nasa.gov, Susan.E.Strahan@nasa.gov.

3. Importance to MAP and other fields

a. The GMI model can work synergistically with other elements of the MAP modeling environment, particularly the Chemistry-Climate and GMAO models, to test model input and processes against a wide range of satellite, aircraft and ground-based observations. Because of its flexibility and lack of stochastic noise, CTMs will continue to be the preferred tool for model validation against composition measurements.

b. The GMI model can be driven by a variety of meteorological fields from different GCMs and assimilation models, thus evaluating their performance when compared to measurements of atmospheric composition. In addition, the GMI model can utilize tools being developed by the MAP Modeling environment to a) facilitate the proper gridding of different meteorological fields, and b) explore the impact of different parameterizations used by the different GCMs/assimilation systems, in a common computational framework. This will contribute to understanding the processes controlling the calculated atmospheric composition, and quantifying the uncertainties in atmospheric model predictions.

c. Additional couplings can be introduced to other elements of the Earth System model, such as land/ocean processes. The model can then be a valuable tool to study some of the important MAP scientific questions, such as the hindcast of atmospheric composition.

d. GMI will continue to be the assessment tool for MAP. In addition, GMI is a prototype for developing a "CTM framework" for use in the MAP modeling environment. Finally, GMI can also contribute to other NASA programs, such as ACMAP, Assessment of Effects of Aviation, etc.

4. Directions for Development a. Keep an ongoing simulation of the stratosphere and troposphere using the GEOS products, and make it available to the community on a routine basis. Continue incorporation of the GMI modules into the GEOS-5 system.

b. Incorporate meteorological fields from different modeling groups: GMAO, CAM, ECMWF, GFDL, NCEP to evaluate their performance in driving tracer transport and chemistry simulations as compared to NASA observations. Utilize tools in the MAP Modeling Environment to reproduce column physics in "replay" mode for these different models, to understand the processes driving the spread in model results.

c. Serve as a "core facility" of the MAP modeling environment available for use by different investigators to address specific problems of interest to the MAP community.

d. Lead the effort to develop a "CTM framework" as a tool of the MAP Modeling Environment to streamline use of specific algorithms, computational platforms, and facilitate research by different modeling groups.

5. Summary

The CTM from the Global Modeling Initiative will be used synergistically with other atmospheric components of MAP to test the different parameterizations and inputs used in those models against observations by NASA satellites and other platforms. An on-going simulation of the A-Train period using GEOS-4 and 5 systems will provide the basis for this testing. In addition, the model will incorporate meteorological fields from other modeling and assimilation systems, to carry out a similar testing and evaluate uncertainty in model results. Utilization of the MAP Modeling Environment capabilities to "play back" the column physics from different models will allow understanding the reason and processes for differences in model results. This work will improve NASA's assessment capabilities, and will lead the effort towards establishing a CTM framework to be used by MAP researchers. --Jrodriguezfl 14:56, 2 October 2007 (EDT)

---

---

[edit] MERRA: Modern Era Retrospective-analysis for Research and Applications

Retrospective-analyses (or reanalyses) have been a critical tool in studying weather and climate variability for the last 15 years. Reanalyses blend the continuity and breadth of output data of a numerical model with the constraint of vast quantities of observational data. The result is a long-term continuous data record. The Modern Era Retrospective-analysis for Research and Applications was developed to support NASA’s Earth science objectives, by applying the state-of-the-art GMAO data assimilation system that includes many modern observing systems (such as EOS) in a climate framework.

The MERRA time period will cover the modern era of remotely sensed data, from 1979 through the present, and the special focus of the atmospheric assimilation will be the hydrological cycle. Previous long-term reanalyses of the Earth's climate had high levels of uncertainty in precipitation and inter-annual variability. The GEOS5 data assimilation system to be used for MERRA implements Incremental Analysis Updates (IAU) to slowly adjust the model states toward the observed state. The water cycle benefits as unrealistic spin down is minimized. In addition, the model physical parameterizations have been tested and evaluated with data assimilation present which also reduces the shock of adjusting the model system. Along with the Catchment hydrology land surface model, MERRA is anticipated to improve upon the representation of the water cycle in reanalyses.

In addition, the 72 vertical levels in GEOS5 extend through the stratosphere. A special data product, developed in conjunction with the chemistry community will support chemistry transport modeling. The disseminated data products will provide much improved initial conditions for predicting weather and other subseasonal variability that is strongly linked to tropical moisture. Studies of climate variability rely on reanalysis data sets. Limited domain models use reanalyses to provide the boundary forcing and initial conditions for mesoscale and regional climate simulations.

MERRA output data will resemble existing reanalyses, with several key advances. The use of IAU also allows output at higher frequencies than the 6 hourly analyses. Two dimensional diagnostics (surface fluxes, single level meteorology, vertical integrals and land states) will be produced at 1 hour intervals. These data products and the 6 hourly three dimensional atmospheric analyses are also available at the full spatial resolution (1/2 degrees Latitude × 2/3 degrees Longitude). Extensive three dimensional 3 hourly atmospheric diagnostics on 42 pressure levels will also be available, but at the coarse (1.25 degree) resolution. The Goddard Earth Sciences Data Information Services Center (GES DISC) is developing the utilities for users to access and subset the MERRA data products.

Production status and recent news will be regularly updated on the MERRA WWW page.

Link to the MERRA WWW Page

[edit] CMAI: Cloud Modeling and Analysis Initiative

Link to more CMAI information

[edit] Background

In the 2007 MAP meeting the CMAI breakout session decided to form three groups to examine cloud variability in three major cloud forming regions: the tropical-subtropical transition, the midlatitude storm regions, and the polar regions. The tropical-subtropical group and the polar group decided to base their study on geographic cross-sections that encompass the major cloud transitions in those regions. The midlatitude group decided instead to first define the major regimes of variability related to midlatitude storms and then to study cloud variability in the regime transitions. The group, that includes both modeling and observational participants, is working to examine cloud, radiation, and precipitation changes with midlatitude dynamic regime both in observational analyses and in climate model output, with the objective to understand the mechanisms of cloud-dynamics interactions and to examine the reasons behind potential problems in model representation of midlatitude clouds. The first part of the project that includes the derivation of a dataset that maps the major properties of the midlatitude dynamic field is presented in this section.

[edit] Ongoing activities

[edit] a. The MAP Climatology of Mid-latitude Storminess (MCMS)

Overview: Baroclinic cyclones are the primary weather-makers in extra-tropical regimes. The specific goal of the MAP Climatology of Mid-latitude Storminess (MCMS) project is to provide a detailed 50 year climatology of the areas that come under cyclone influence at a given point in time. This task requires that we solve specific problems such as how to find cyclones in space and time, and then once this is done, how to demarcate the area of influence around them. These are unsettled issues, the second, more so than the first. Indeed, there are many proposed methods for locating cyclones (e.g, Benstad, R. & Chen, D. (2006); Simmonds & Keay (2000); Sinclair (1994); Wernli and Schwierz (2006)). We use the most popular method; locating cyclones as depressions in the sea level pressure (SLP) field. Delineating cyclone influence is a lesser explored topic and we currently lack well established criteria for doing it. A commonly used, but blunt, solution to this problem is to simply place a fixed-size box around a found cyclone and extract a composite (e.g., Bauer & Del Genio (2006); Field & Wood (2007); Lau & Crane (1997); Sinclair & Revell (2000)).

This one-size-fits-all method has a major limitation for our purposes given that cyclones vary widely in terms of size and shape. In principle we’d like to treat each cyclone individually and adaptively. Following the ideas in Wernli and Schwierz (2006) we derived such a method based on the idea that a cyclone’s area of influence or “storminess” is bound by the unique set of concentric SLP contours surrounding that cyclone. Essentially, we treat storminess as being confined to the bowl of the cyclone’s SLP depression. There are some caveats with this method too of course. For example, cyclone features sometimes extend beyond the confines of the closed SLP contours surrounding the cyclone. In other cases, cyclones simply lack closed SLP contours altogether or they depart from the ideal of simple isolated features and exist in complex cyclone families. That said, our method works well most times and we’re working continuously to make it better.

So who might find MCMS useful? Anyone who wants to contextualize their data (observed or modeled) by the presence or absence of a nearby cyclone is the general answer. An obvious research avenue is to use MCMS to examine cyclones themselves. For example, model makers can use MCMS to improve and validate the dynamic and parameterized model response to baroclinic waves (e.g., Bauer & Del Genio (2006)). Observations can likewise be sifted and organized with MCMS to extend our basic understanding of extratropical climate and weather (e.g., Field & Wood (2007)). Less obviously, MCMS can be used as a weather sensitive filter for any sort of data or model component. Ecological and oceanographic studies come to mind as do aerosol and pollution studies.

Documentation: For a detailed account of our procedure, including how to read the data, and examples of what can be done with the data see our documentation. Below is a short overview.

Procedure: we use a three step process

• Step 1: Identify candidate cyclones (a.k.a. 'centers') in the sea level pressure (SLP) field of the NCEP/NCAR Reanalysis (6 hourly). Conditions for being a 'center' beyond the basic requirement of being a local SLP minima are designed to ensure that a center is a significant, persistent and mobile disturbance in the SLP field.
• Step 2: Refine the list of potential cyclones by connecting 'current' and 'past' centers via nearest neighbor and other similarity arguments to form a 'track.'
• Step 3: Define a region of influence for each potential cyclone using a technique akin to that taken by Wernli and Schwierz (2006). In essence, this means finding the largest set of closed SLP contours enclosing each center and no other.

Using this scheme each grid on every SLP map falls into one of four main categories; 1) an attributed grid, 2) a "stormy" grid, and 3) the remaining non-stormy grids which are assumed to have no association with cyclones. The fourth category, for "empty" centers, accounts for situations where a center (always a very shallow one) lacks even a single attributed grid. This happens because that center's very first contour is shared with another center(s) and so all that center's grids get classified as 'stormy.'

Caveats and Limitations:

• Center identification at high latitudes (> 85°) suffers from the contraction of grid spacing with latitude because our scheme works on the native grid-space of the SLP source. In particular, it is often the case that near polar centers fall on the same longitude because the polar cap's uniform SLP results in the average center location falling at 0° longitude.
• Center identification is suspended over high topography (> 1500 m) because errors due to SLP reduction, and the real noise in the surface pressure field over high topography, precludes reliable center identification. As an aid to the researcher we therefore flag all high topography grids that are associated with cyclone activity as being "problematic." That is, we caution that an undetected center may be present but unaccounted for in our results.
• Center identification is discards potential centers if they are not local SLP minima which means open-wave cyclones are likely to be under-represented in our results.

Products and Tools:

• Data Sets Available Here: We provide the results for each of the above steps in a formated ASCII text file. Aside from the when and where of each center, these files provide a detailed assessment of each center including measures of its relative intensity, size, and analogous properties. Note that these products are subject to minor revisions and fixes as they are identified and suggested (each data set as a version number).

1. Centers Attribution Database: A sorted list of centers and their area of influence (~300 mb per decade, 70 mb compressed).
2. Tracks Attribution Database: A subset of the centers attribution database with just the cyclones that can be tracked.
3. Discards: A sorted list of found centers that were subsequently discarded from our analysis along with the reason why.

• Source Code: We provide the full source code for our algorithm, which is written in the Python programming language, and should be executable on most systems and readily understandable by people familiar with C-like dynamic languages including Perl and Ruby [Coming Soon: Complete with a case study of how to adapt to new SLP sources, in this case GISS ModelE].
• Utilities Available Here: We provide utilities (written in Python) for working with our data sets. This list will grow, especially with contributions by other researchers (hint/plea). Current facilities include the ability to:

1. Extract center/track information by time (e.g., by specific year, month, day, hour, range of date/times, or predefined seasons), by place (e.g., by hemisphere, latitude-longitude pairs, or a land/sea mask) and by property (e.g., by cyclone intensity). These criteria can be combined to refine your selection with high granularity. For example, all DJF cyclones whose center or any attribution grid occupied the NCEP Reanalysis grid containing New York City.
2. Extract minimal center/track information (e.g., location in place and time) or the complete database entry for each center/track.
3. Reorganize the database by tracks rather than by the default chronological ordering.
4. Create fixed frame and adaptive frame composites. A fixed frame composite use a fixed size window to collect cyclone centric statistics whereas an adaptive frame composite uses the variable number of attributed grids found around each center. [Coming Soon]

[edit] Importance to MAP

One of the primary objectives of the MAP CMAI program is to quantify the scale dependence of the coupling of atmospheric dynamics with radiation and precipitation by clouds necessary to generalize findings so that they can be accurately and effectively represented in climate (Earth System) and weather models. The present study attempts to quantify this coupling in the midlatitude storm regions, an area where atmospheric dynamics play the primary role in defining the character and strength of the energy and water cycles. The first step in this process, undertaken by NASA/GISS, was to produce a dataset that, for the reanalysis time period, locates all the low pressure centers, tags whether they are part of a storm track, and then defines their area of influence through the use of a closed contour methodology. This way, all grid points related to a midlatitude storm system are identified and related to the storm center that influences them. The dataset presented here will then be provided to the observational groups that participate in the study in order to composite cloud, radiation, and precipitation datasets in ways that will reveal how those fields vary with varying properties of the dynamic regime. In addition, the code that locates and tracks the storm centers and defines their area of influence will be provided to the participating modeling groups in order to produce similar model composites and compare them to the observed ones. The objective is to not only evaluate model midlatitude cloud fields but to also identify the dynamic regimes and conditions where model errors occur and thus provide information that could lead to model parameterization improvements. The active participation in the group of the GISS and GMAO modeling groups makes it probable that the results of the composite analysis will be used to improve model representation of midlatitude cloud systems.

[edit] Ent Dynamic Terrestrial Ecosystem Model

The Ent model is a dynamic global terrestrial ecosystem model (DGTEM) being developed specifically for coupling with atmospheric general circulation models (GCMs). Ent will be capable of predicting the fast time scale fluxes of water, carbon, nitrogen and energy between the land-surface and the atmosphere and the resulting diurnal surface fluxes, seasonal and inter-annual vegetation growth, and decadal to century scale alterations in vegetation structure and soil carbon and nitrogen.

Special features that Ent introduces to improve representation of vegetation and ecosystem dynamics in global modeling include: canopy radiative transfer and competition in mixed vegetation communities, dynamic nitrogen allocation according to photosynthetic activity, and a deep soil layer to capture deep soil carbon storage. The Ent modules are being tested rigorously on Fluxnet site data to check the ability of the model to calculate vertical profiles of radiation, and to simulate water vapor and CO2 fluxes, seasonal growth, and equilibrium soil carbon pools. MODIS data are being used to test albedo and phenology algorithms.

The finished Ent product will be a standalone set of modules for the climate modeling community to use to couple with land surface models and atmospheric GCMs for studies on seasonal weather evolution, vegetation phenology, the carbon budget, climate variability, paleoclimate, global change scenarios, vegetation-climate feedbacks, and astronomical biosignatures. Ent is envisioned as a tool for understanding the conditions and signatures of habitability of the Earth, ancient, modern, and future.

(Link forthcoming)

[edit] Closed-ended Projects

The MAP Program periodically sponsors “projects” that include integration across models, super-computing, organizations, and NASA and non-NASA systems infrastructure. In both 2005 and 2006 the MAP Program participated in assessing how the GEOS-5 model performed for hurricane tracking. In addition, in 2006 the MAP '06 Project participated in testing an interface between the GEOS-5 global model and the WRF regional model,and provided that information to the NAMMA field experiment.

[edit] MAP05

During the summer and fall of 2005 NASA's Modeling, Analysis, and Prediction (MAP) program funded the "MAP '05" project, a major effort designed to implement three versions of Goddard’s flagship global atmospheric model. The goal was to enable “real time” prediction of Atlantic tropical cyclones and other meteorological events. By executing the research products in an operational environment in which side-by-side comparisons with other types of models could be analyzed, the strengths and weaknesses of the Goddard models could be better understood, particularly under extraordinary atmospheric conditions. For the first time, Goddard’s numerical models were introduced as an offline member of the Florida State University (FSU) “Superensemble,” a novel approach to forecasting tropical cyclones that combines the output from multiple models and quantifies the biases of each such that an optimal forecast of storm track and intensity can be determined. Results from the Superensemble were made available to forecasters at the National Hurricane Center. Although prediction of tropical cyclones was a significant outcome of this effort, MAP ’05 is designed to support multiple scientific investigations.

Link to more MAP05 information

[edit] MAP06

During the summer of 2006 The Earth-Sun Exploration Division of Goddard Space Flight Center(GSFC) and the Science and Mission Systems Office at Marshall Space Flight Center will bring together resources from NASA and from corporate partners to study tropical cyclones. The “MAP ’06 Project,” so named for its affiliation with NASA’s Modeling, Analysis, and Prediction (MAP) program, will apply NASA’s advanced satellite remote sensing technologies and earth system modeling capabilities to improve our understanding of tropical cyclones that develop in and move across the Atlantic basin. MAP '06 will implement the most recent version of the Goddard Earth Observing System (GEOS) fifth-generation global atmospheric model and the Grid point Statistical Interpolation (GSI) analysis system under development as a collaboration between NOAA's National Centers for Environmental Prediction (NCEP) and the Global Modeling and Assimilation Office (GMAO) at GSFC. In addition, the capability to initialize the Weather Research and Forecast (WRF) regional model using GEOS-5 will be developed and implemented. The project will begin in the early portion of the 2006 hurricane season and continue through late autumn.

Link to more MAP06 information

[edit] Summer 2007

What's going on this year?

[edit] SNAPA and FastJX

SNAPA stands for the Swift Neurological Acceleration of Atmospheric Photochemistry and Aerosol Calculations. This year effort has been focussed on using neural networks to add new capabilities to FastJX (v6.1), namely the real time calculation of scattering parameters instead of pre-tabulations. This will facilitate the coupled aerosol chemistry work in GMI and beyond.

A fortran90 version of FastJX (v6.1) has been created, which in our tests gave identical results to the Fortran77 version, and are now adding the capability to do real time calculation of the scattering properties (starting with the sulfate aerosols). Currently the scattering parameters are pre-tabulated as the calculations are so expensive. By using the neural networks we aim to remove this limitation and have real time estimation of the scattering parameters.

Satisfying progress has been made using genetic algorithms to objectively design and choose the best neural network configurations. This has been a major breakthrough.

Starting with the latest version of Michael Mishchenko's Lorenz-Mie theory code (updated on 08/06/2005 and based on the book: Mishchenko, M. I., L. D. Travis, and A. A. Lacis (2002), Scattering, Absorption, and Emission of Light by Small Particles. Cambridge University Press, Cambridge), we are currently considering a massive range of log normal distributions, refractive indices (real and imaginary), and wavelengths that encompass all the variants we be likely to encounter, this constitutes many thousands of individual cases. Each of these individual cases take quite a while to compute. This constitutes our neural network training dataset.

We hope to latter be able to extend this to use the T-matrix approach for computations of light scattering by nonspherical particles (applicable to spheroids, Chebyshev particles, and finite circular cylinders) as described in: M. I. Mishchenko, L. D. Travis, and D. W. Mackowski, T-matrix computations of light scattering by nonspherical particles: a review, J. Quant. Spectrosc. Radiat. Transfer, vol. 55, 535-575 (1996) and M. I. Mishchenko and L. D. Travis, Capabilities and limitations of a current FORTRAN implementation of the T-matrix method for randomly oriented, rotationally symmetric scatterers, J. Quant. Spectrosc. Radiat. Transfer, vol. 60, 309-324 (1998).

[edit] New Science Directions

[edit] Modeling atmospheric layers

Long-range transport in the troposphere often takes place in vertical layers typically ~1 km thick preserving their structures over thousands of km. These layers spread horizontally and interweave, as seen for example in the juxtaposition of polluted and stratospheric air in observations of continental outflow. Global Eulerian models are incapable of reproducing such structures. Vertical gradients dissipate far too rapidly, even in models with relatively high vertical resolution. This failure of models seriously compromises their ability to describe the global-scale chemical evolution, radiative properties, and transport pathways of pollution plumes. It also represents a major limitation for the assimilation of satellite vertical profiles into atmospheric models.

MAP is the ideal vehicle for addressing this problem, for three reasons: (1) it is a problem with a broad range of implications for atmospheric chemistry and climate research, (2) it is critical for the quantitative interpretation of satellite observations, (3) there are various possible approaches for solving it and it is not clear which is best. Possible approaches might include for example adaptive grids, embedded Lagrangian plumes, or isentropic coordinates. Projects to investigate these different approaches should be encouraged.

[edit] Lightning

[edit] A short description

Lightning is an atmospheric discharge of electricity, which typically occurs during thunderstorms, and sometimes during volcanic eruptions or dust storms. Lightning can also occur within the ash clouds from volcanic eruptions, or can be caused by violent forest fires which generate sufficient dust to create a static charge. About 77 million lightning bolts annually strike the U.S. Measurements before and after lightning strikes have confirmed the generation of nitrogen oxides in the atmosphere. A bolt of lightning can travel at a speed of 45 km/s (100,000 mph, 160,000 km/h). It can reach temperatures approaching 28,000 °C (50,000 °F), hot enough to fuse soil or sand into glass channels.

An average bolt of lightning carries a negative electric current of 40 kA, although some bolts can be up to 120 kA, and transfers a charge of five coulombs and 500 MJ, or enough energy to power a 100 watt lightbulb for just under two months. The voltage depends on the length of the bolt: with the dielectric breakdown of air being three million volts per meter, this works out at about one billion (thousand million) volts for a 300m (1,000 foot) lightning bolt. With an electric current of 100 kA, this gives a power of 100 trillion (million million) watts.

Lightning produces large scale ionization and a shock wave. Both the ionization and high temperatures are significant for atmospheric chemistry, and the full implications are usually completely overlooked with attention be paid almost exclusively to NOx. Ionization produced by cosmic rays and precipitating particles is well known to produce NOx and HOx (Brasseur, 1987). The ionization associated with lightning is between six and fifteen orders of magnitude greater than that associated with cosmic rays.

[edit] Existing activities

[edit] Why this is important to MAP

Electrified convection provides a unique setting for atmospheric chemistry. While the chemistry of lightning is sometimes thought of as predominantly the chemistry of the hot (≈ 30,000K) lightning channel, the large radial electric fields (> breakdown strength of air) surrounding the lightning channel results in high ion and electron production rates in the "corona sheath". Additionally, the high-temperature lightning channel releases large amounts of short-wavelength, ionizing, UV radiation (λ_max ≈ 100nm) that is absorbed in the surrounding region. The relative volumes of the hot lightning channel and the surrounding corona sheath are also noteworthy. If a typical lightning channel has a radius of a few centimeters and the surrounding corona sheath is a few tens of meters then the ratio of the volume of the coronal sheath to that of the lightning channel is 10⁶:1.

The fate of the highly reactive, charged products formed outside of the hot channel determine, in part, the net chemical effect of electrified convection. These products can dominate over the hot-channel chemistry and alter the local concentrations of all the major chemical families considered in standard photochemical studies. Additionally, it is of interest to know whether electron-capturing gases with very long tropospheric lifetimes, such as SF₆ may be removed by in-cloud ion chemistry at a rate sufficient to materially alter their total atmospheric lifetime.

The simulation of convection, lightning and subsequent NOx emissions with global atmospheric chemistry models is associated with large uncertainties since these processes are heavily parameterized. Each parameterization by itself has deficiencies while the combination substantially increases the uncertainties from the individual parameterizations.

In addition to producing NOx, lightning is almost certainly producing HOx and HCN. This has to date been largely ignored. This is significant as the HOx abundance is crucial for the oxidizing capacity of the atmosphere, and HCN is a stable, long-lived, sparingly soluble molecule with a long residence time against rain-out. Unlike NOx, HCN may act as a relatively inert 'smoking gun marker' of lightning activity, and may possibly be useful as a proxy for the total amount of lightning activity in the atmosphere. The vertical structure of HCN observed during thunderstorms may also provide a good test for the hypothesis that HCN is produced by lightning.

[edit] What other subject areas are relevant?

Ion Chemistry. HCN can be produced through the ion chemistry associated with lightning. Electrons and EUV photons produce N₂⁺, N⁺, O₂⁺, and O⁺ ions, this can then lead to the production of HCN by reaction sequences such as:

N⁺ + CH₄ → HCN⁺ + H₂ + H

HCN⁺ + H₂ → H₂2CN⁺ + H

H₂CN⁺ + e^- → HCN + H

We know that in the terrestrial atmosphere N₂⁺, N⁺, O₂⁺, and O⁺ ions can be produced by photons and cosmic rays. The production rate due to cosmic rays is typically of the order of 1 to 100 ion pairs cm^-3 s^-1 and this is known to produce both NOx and HOx. These same ions are also produced, but in much larger numbers by lightning, typically of the order of 10⁸ to 10¹⁷ ion pairs cm^-3 s^-1 (Boldi, 1992), i.e. an ion production of between 6 and 15 orders of magnitude more than that produced by cosmic rays. So it is very likely indeed that the massive ionization produced by lightning will produce not only NOx but significant amount of HOx, and HCN also. This is borne out by HCN observations, the observations of elevated HOx associated with elevated NOx injected by deep convection and lightning (Jaegle 1999), and the detailed calculations of Boldi (1992). Boldi (1992) found that the NOx mixing ratio was increased by about 3 orders of magnitude from 20 pptv to 100 ppbv in the presence of lightning, but the maximum OH mixing ratio was increased by 5 orders of magnitude within the region of lightning activity. Averaged over a larger domain this corresponded to a lower increase of OH of an order of magnitude. This can readily explain the association of elevated HOx with elevated NOx injected by deep convection and lightning.

However, unlike HCN, and NOx, the HOx species are more soluble and may be incorporated in the aqueous phase and rained out. Ion chemistry is not the only source of HCN and HOx that is associated with lightning activity.

It is timely to revisit the interaction of neutral and ion chemistry and develop the modeling capability in this area, in particular to have a more complete lightning simulation capability.

• Robert A. Boldi, 1992, Report 23, MIT Center for Global Change Science. A Model of the Ion Chemistry of Electrified Convection

[edit] Research Directions

Development of lightning ion chemistry schemes suitable for global models linked to improved multivariate non-linear parameterizations of lightning flash rates/locations. This ion chemistry may lead to better model simulation of HNO3/NOx ratios.

[edit] Why Model Validation is important to MAP

5.3.1 For models to be credible in policy assessments they need to be validated. The wide range of satellite and other constituent data available can be used to validate models in the same way that satellite observations are validated. This can build on the work already done for long lived tracers such as N2O (Strahan) and be extended to more reactive species such as HNO3, HCl, NO2.

Combining multiple observational datasets is also of use in this regard. For example, knowledge of the distribution of inorganic chlorine Cly in the stratosphere is needed to attribute changes in stratospheric ozone to changes in halogens, and to assess the realism of chemistry-climate models (Eyring et al., 2006,2007). However, there are limited direct observations of Cly . Simultaneous measurements of the major inorganic chlorine species are rare (Zander, 1992; Gunson, 1994; Bonne, 2000; Nassar, 2006), although in the upper stratosphere, Cly can be inferred from HCl alone (e.g., Anderson (2000), Froidevaux (2006)).

Chapter 6 of the recent WMO assessment on stratospheric ozone depletion showed a range of predictions for ozone recovery. These models contained a wide range of total inorganic chlorine, Cly. Funding from the last MAP program facilitated the production of a consistent time series of stratospheric inorganic chlorine Cly from 1991 to present using space-borne observations together with neural networks. A neural network was first used to account for inter-instrument biases in HCl observations. A second neural network was used to learn the abundance of Cly as a function of HCl and CH4, and to form a time series using available HCl and CH4 measurements. The estimates of Cly were broadly consistent with calculations based on tracer fractional releases and previous estimates of stratospheric age of air. These new estimates of Cly provide a critical test for global models, which exhibit significant differences in predicted Cly and ozone recovery. Such studies should be extended to systematically test the models currently available.

[edit] Next Generation Model Parameterizations

Many complex parts of global models are parameterized, for example, lightning, convection, etc. These parameterizations are necessary because fully representing all the important processes in a model would be computationally prohibitively expensive. Since these parameterizations are critical to a successful model, an effort is needed to explore new approaches. A continuing effort is also needed to upgrade and refine important model parameterizations. For instance model parameterizations are often created and tuned for a particular resolution of the model in which they are embedded. As resolution increases, these parameterizations need to be revisited and revalidated.

One issue with parameterizations is often choosing a suitable functional form and being able to account for the non-linearities. Both of these tricky issues can be addressed using neural networks. Neural networks are non-linear non-parametric learning algorithms that are universal approximators. They have proved useful in a variety of applications (Lary et al., 2004, 2007a,b), from the acceleration of expensive code elements to learning the cross-calibration between large earth observing datasets including atmospheric composition, aerosol optical depth, and vegetation indices. For these tasks a variety of neural networks have been used including feed-forward multi-layer perceptron networks trained with the Levenberg-Marquardt algorithm, and neuro-fuzzy networks.

The success of the neural networks largely depends on two factors. First, having a training dataset that adequately spans the parameter space. Second, including the variables that explain the variance in the dataset. If these two criteria are met then the neural networks give excellent results as they are universal approximators. It is timely to extend their use in global models. Two obvious uses are for lightning parameterizations and emissions.

• Lary, D. J., Muller, M. D. & Mussa, H. Y. (2004) Using neural networks to describe tracer correlations. Atmospheric Chemistry and Physics, 4, 143-146.
• Lary D.J. et al. (2007a), Variations in Stratospheric Inorganic Chlorine Between 1991 and 2006 (using neural networks), GRL, in press.
• Lary, D.J., Aulov, O. (2007b), Space-based measurements of HCl: Intercomparison and Historical Context (cross-calibration using neural networks), JGR, in press.

It should be noted that neural net-based parameterizations are members of a class of parameterizations that do not attempt to reproduce in a physical way (though simplified) the processes they represent. Rather, they only attempt to represent the effects of the parameterizaed process in a model. For instance, in the case of lightning, a neural net-based approach would map a set of model variables such as precipitation rates or vertical mass fluxes to a NOx production rate. Such parameterizations may prove very useful, but they provide little insight into the nature of the physical processes they represent.

[edit] OSSEs and Optimal Observing System Design

Models can play an invaluable role in designing the next generation of observing system via Observation Simulation Sensitivity Experiments (OSSEs). An OSSE can be subdivided into four basic steps:

• Generate a "nature" atmosphere;
• Compute synthetic observations;
• Assimilate the synthetic observations; and
• Assess the impact on the resulting forecast.

Each step is performed with the goal of minimizing any external influences, which may compromise the value of the synthesized observations, the assimilation process, or the results of the numerical forecasts.

This can be taken one step further using genetic algorithms to optimize the observing system design (e.g. orbit, weight, power, view, etc). Genetic algorithms have already been used in the MAP SNAPA study to objectively design neural networks, so they could also be used in new mission design. The current mission design carried out at places such as the GSFC Integrated Mission Design Center (IMDC), is not focussed on this level of optimization. Given current budgetary constraints it is timely to extend our current software infrastructure to perform joint optimization for the science objectives and the practical space craft issues such as power, orbit, view, etc in an objective manner via techniques such as genetic algorithms.

[edit] Understanding Processes that Control Tropospheric Composition Above 5 km

[edit] Background

Our current understanding of the climate system indicates that the atmospheric composition above 5 km plays a crucial role in the coupling between chemistry and climate. Several processes contribute to this coupling: a) radiatively and chemically active species, such as O3, dust and black carbon at these altitudes play a major role in the radiative balance of the atmosphere, due to the low temperatures in the upper troposphere; b) the long range transport of radiatively active species and their precursors occurs primarily in this region; c) the composition of this region may affect and be affected by processes in the lower stratosphere; d) the composition in this region, particularly aerosols, may play an important role in determining formation of high-level clouds important to the climate system, such as cirrus. At the same time, our understanding of the processes controlling the composition in this region is still fraught with uncertainties.

[edit] Ongoing activities

.Results from the near 20 models participating in AeroCOM exhibit large differences in their calculated aerosol loadings in this region. A large variability in results for gaseous constituents was also observed in the intercomparison/validation carried out under the TRADEOFF program and the variability in ozone burdens and budgets is also influenced by process in the middle-upper troposphere. Due to the complex interactions between physical, chemical and transport processes that determine the composition, we are missing a systematic approach to unraveling the relative importance of the different processes.

Under the sponsorship of SPARC, the CCMval activity has made important contributions to understanding the processes controlling the chemistry and climate response in the stratosphere. A similar activity for the upper troposphere/lower stratosphere is highly desirable, but would be complicated by the multitude of processes to be considered. Such processes would include: a) vertical transport by large-scale winds and convection; b) interaction between vertical transport and long-range horizontal transport; c) removal of aerosols and gaseous species by wet and dry depositon; d) in-situ production of ozone precursors (NOx) by lightning and aircraft; e) stratospheric-tropospheric exchange. The ultimate goal of improving our understanding will be best served by concentrating on the above processes in an incremental fashion. We submit that we first address the representation and role of convection and wet deposition in addressing the problem, in the belief that these processes may control a large portion of the observed inter-model variability for aerosols. These processes have not been examined in any systematic way since the intercomparisons summarized by Jacob et al. (1997) and Rasch et al. (2000). Subsequently, the role of in-situ ozone production, lightning NOx production, and stratospheric-tropospheric exchange can be explored utilizing existing efforts and insights gained from the convection/wet deposition effort.

The Atmospheric Chemistry and Climate initiative of IGAC has recently identified four focus activities of interest for the international modeling community. The problem discussed here constitutes Activity 2. Past meetings and discussions on this subject indicate a strong interest in the international modeling community in addressing this issue.

[edit] Importance to MAP and other fields

As the Modeling, Analysis and Prediction program, the predictive capability of existing models is crucial for ultimate attainment of the program's goals. The large variability in results from current simulations can be due to real uncertainties in our representation of atmospheric processes, and/or faulty representation in some models. The Models and Mesurements effort sponsored by NASA for the stratosphere in the 1990's pointed out both of the above problems in the models at the time, and contributed to reducing the uncertainty in model predictions. The complexity of tropospheric processes makes an effort of that magnitude much more difficult, but a better approach may be to try to understand different processes in an incremental fashion.

In addition to its importance to MAP, understanding the processes in this region of the troposphere is also important for air quality studies, in particular those concerned with the long-range transport of atmospheric pollutants (HTAP). This long-range transport is very sensitive to the convective characteristics of different models, and the large-scale winds in the middle-upper troposphere.

[edit] Directions for Development

Advances in our understanding of this problem should include the following elements:

• Community workshop(s) to review the state of knowledge and agree on specific strategies. • Identification and collection of a set of atmospheric measurements that can specifically test the different process representations in current models. • Identification of a set of simulations for different models that will isolate how changes in the representation of a particular process impact model results. Ideally, such simulations would be most useful if all model components (for example, emissions, large-scale winds, others?), except for the one being studied, could be fixed to an agreed-upon representation. • Coordination with other ongoing international efforts, such as CCMval and HTAP.

[edit] Summary

As part of an effort to improve the predictability of atmospheric composition in a region important to climate, the MAP program will sponsor efforts towards a systematic examination of processes that control the gas-phase and aerosol composition in the troposphere above 5 km. We encourage an incremental approach to the problem, focusing first on convection and wet deposition, which are important to determine aerosol mass loading. The strategy should include plans for: a) systematic model studies and intercomparisons; b) utilization of existing atmospheric measurements to reduce uncertainty; c) coordination with other ongoing international efforts.

--Jrodriguezfl 10:57, 4 October 2007 (EDT)

[edit] Global Change in Atmospheric Chemistry, Composition, and Climate: Hindcasting Trends and Variability since 1980.

Overview. Understanding how the chemistry and composition of the atmosphere may change over the 21st Century is essential in preparing adaptive responses or establishing mitigation strategies. The abundance of aerosols and many greenhouse gases is controlled by atmospheric chemistry and physics integrated over global scales. A changing atmosphere not only drives climate change but also directly threatens human health, agricultural productivity, and natural ecosystems. Projections of future climate change are coupled with changes in atmospheric composition whose impacts extend to air quality: e.g., the projected increases in near-surface ozone levels threaten air quality standards for much of the world's population (Prather et al., 2003; Stevenson et al., 2006). This activity seeks to test and diagnose global chemistry-transport models (CTMs), using the past few decades of observations to quantify and reduce uncertainties when these models are used in climate system models to project the 21st century.

[edit] Goals

The primary objective of this IGAC Atmospheric Chemistry and Climate Activity #1 (AC&C1) is to evaluate the capability of current atmospheric chemistry models to integrate over the variations and trends in circulation and climate, in emissions, and in chemical feedbacks that control atmospheric composition. The composition of the atmosphere changes in response to altered forcings (i.e., emissions, land-surface change, solar cycle) and climate variations (winds, convection, precipitation, clouds). An ideal test of the models used to project future atmospheric chemistry and climate is provided by the past few decades for which we have observations of trends and variability in atmospheric composition. This effort focuses on space (1000+ km) and time scales (multi-year to decadal) that are essential in projecting 21st century change, and that effectively integrate over many atmospheric processes. We propose to develop a set of hindcast experiments to test and grade CTM performance (e.g., Douglass et al., 1999). Through this process, we expect to derive more objective measures of uncertainty in modeling atmospheric chemistry and transport, and this in projecting future composition. As has been demonstrated in almost all such model and measurement efforts, CTM development is advanced, and we learn more about the atmosphere.

AC&C1 takes an integrative approach and will not focus on process validation, which is being examined by others (AC&C2, CCMVal). In certain tests, however, the primary cause of variation or trend may be identified through model studies with a specific atmospheric process, and thus this activity will augment those studies.

As with most all such projects, this remains primarily a volunteer effort. We anticipate that many research groups may be able to persuade their funding sources to directly support AC&C1 activities. Nevertheless, to gain participation, we rely on developing a program that provides scientific curiosity, excitement, and opportunity for researchers trying to understand the coupling of atmospheric chemistry and climate.

[edit] Experimental Approach

AC&C1 needs to establish a core group to oversee the project. It is expected that each member of this group, as well as others participating in the process, will define and agree to work on specific hindcast experiments. By an experiment we mean the clear definition of (i) a multi-year measurement time series (post-1980) with clear grading criteria for any model simulation; (ii) all prescribed forcings needed to predict those measurements; and (iii) minimum guidelines on the types of CTMs and meteorological fields that should participate. Hopefully, many of these experiments will have already been "tried out" by one research group.

The obvious big question is: Why is there not a single hindcast from 1980, one that includes gases and aerosols, troposphere and stratosphere, all natural and anthropogenic forcings? After all, we expect global atmospheric chemistry to couple most all species. This monolithic 1980-hindcast would be an interesting challenge, but, more practically, our measurement and forcing data sets may span different periods (e.g., the long-term record of CFCs and N2O from the AGAGE network vs. the more recent global satellite mapping of aerosols with MODIS) and thus lead to different experiment design. Equally important, current CTMs have very different properties and capabilities: simulated or prescribed stratospheres, embedded in climate models or free-running with any prescribed meteorological fields, with or without aerosol-cloud interactions. We need to develop a series of experiments that can be run with several CTMs, that is as inclusive as possible of the different forms of these models.

As one example, consider the trends and variability of the nearly-inert CFCs and N2O as recorded by five surface stations of the ALE/GAGE network. These have been used extensively in CTMs to test and calibrate global transport and mixing (e.g., Prather et al. 1987). It was suggested that the year-to-year variations in Samoa represented large changes in the inter-hemispheric mixing as opposed to merely a shift in the southern ITCZ. Recent modeling analysis (Nevison et al., 2007) supports the latter view, but leaves open uncertainties on stratosphere-troposphere exchange as a source of variability. Extending the Nevison et al work to stratospheric models and/or using CCMVal studies to impose an upper boundary forcing for the stratospheric flux might allow resolution of this. Extending the measurements to the more geographically extensive NOAA network may provide a classic data set for testing CTM trace-gas transport.

Other examples will be more complex, far more interesting, and need careful definition: CH4 trends and variability; global distributions of aerosols from recent satellite data. Perhaps the most central test of tropospheric chemistry is understanding O3 trends and variability (Fusco and Logan, 2003). In addition to the obvious forcing from ozone precursors, this experiment involves aerosols, methane, the stratosphere, and climate change. Definition of the tropospheric O3 experiment will be one of the most difficult tasks. Other important issues involve the type of met fields: either analyzed, forecast, or AMIP-forced GCM, with/without reasonable stratospheric transport. A more general experiment might address the chemical response to the documented changes in climate (T, H2O, circulation) over the last twenty-five years.

[edit] Analysis & Diagnostics

Ready access to, and standardization of the data sets -- for forcings, measurements, and model results -- are essential to the success of AC&C1. Whether a single data base is maintained by one participant, or distributed data sets by many, it is critical to have a central web site and knowledge base for AC&C1.

In addition to the grading of a CTM's ability to hindcast a given set of observations, we propose a process-oriented set of diagnostics, and in some cases sensitivity studies, to evaluate the cause of variability in each model. For example, detailed analysis of model output in the AEROCOM study finds that CTMs often arrive at similar results (e.g. aerosol total optical depth) through different mechanisms. Thus, AC&C1 will contribute to those studies by evaluating with models the important of specific processes within the fully coupled system.

[edit] Additional notes

Participants. AC&C1 will be an open process, however, participants will have the responsibility to: contribute the forcing and measurement datasets that define the hindcast experiments, and/or perform the model simulations and delivering standard diagnostics, and/or analyze and write up the particular experiments.

Schedule. A January or February workshop is necessary to build community consensus, to establish the central site, and to get specific recommendations for the hindcasts, including specific assignments/deadlines for delivery of data.

Coordination with other projects. A review of the RETRO, CCMVal , HTAP, and AEROCOM activities (past and present) is needed to collect currently evaluated data sets for observations and forcings. Continued coordination with those ongoing projects as well as with the other AC&C activities must also be maintained.

Douglass, A.R., et al. (1999), Selecting the best meteorology for the global modeling initiative's assessment of stratospheric aircraft, J. Geophys. Res., 104, 27545-27564. Fusco, A.C. and J.A. Logan (2003), Analysis of 1970-1995 trends in tropospheric ozone at Northern Hemisphere midlatitudes with the GEOS-CHEM model, J. Geophys. Res., 108 (D15), 4449. Nevison, C. D., N. M. Mahowald, R. F. Weiss, and R. G. Prinn (2007), Interannual and seasonal variability in atmospheric N2O, Global Biogeochem. Cycles, 21, GB3017. Prather, M., M. McElroy, S. Wofsy, G. Russell and D. Rind (1987), Chemistry of the global troposphere: fluorocarbons as tracers of air motion, J. Geophys. Res. 92, 6579-6613. Prather, M., et al. (2003), Fresh air in the 21st century?, Geophys. Res. Lett., 30(2), 1100. Stevenson, D. S., et al. (2006), Multimodel ensemble simulations of present-day and near-future tropospheric ozone, J. Geophys. Res., 111, D08301,.

--[[User:MPrather|] 11:38, 21 October 2007 (PDT)

[edit] Information Technology and Computational Infrastructure

In the MAP NASA Research Announcement it was stated that MAP would support the development of an infrastructure to support the program. This was the foundation for the formation of the Software Integration and Visualization Office (SIVO). Infrastructure projects are currently underway, to find the latest status of each project go to infrastructureprojects.

From the NRA:

"As guidance, in the third year of the current proposal cycle the activities residing in the multi-investigator proposals will converge into a shared MAP modeling environment. This environment will be ESMF compliant and investigators will be able to configure model and assimilation systems to support specific applications. There will be planned releases of major development cycles in support of MAP priorities. The Core Integration Team, which will receive software deliveries from the development activities and will have prime responsibility for managing the MAP modeling environment. Successful research teams will include resources to provide an effective interface with the Core Integration Team. Development and application teams will assume primary responsibility for evaluation and certification of major development cycles. The MAP program will evolve the structure and management over the first two years of this proposal cycle."

To date, the MAP Modeling Environment infrastructure effort has been focused on supporting development of “Earth System Science Models”. Earth System Science Models couple differing discipline components across multiple teams of science investigators. The infrastructure effort does not duplicate the science, but instead is focused across modeling efforts to try to make development of the software systems less burdensome to the science developers. These on-going efforts include:

• Computer scientist support for using ESMF to couple models Earth System Modeling Framework (ESMF)
• Computer scientist support for model optimization
• Computer scientist support for hosting models on NASA and other super-computing facilities
• Web page development providing a central view of the MAP model research efforts and software (map.nasa.gov)
• Blog and wiki development by the MAP project scientist to enable new forms of inter-scientist communication
• Work with NASA technology transfer offices to release NASA models under open source
• Systems development for a Knowledge Base for sharing lessons learned among model developers (coming soon)
• Systems development for a Workflow tool enabling researchers to run models from a web page, archive, restore experiments, and monitor and control experiments while executing (coming soon)
• Systems development for a model portal with visualization tools supporting web based sharing of results among distributed researchers (http://map.nasa.gov/tools.html)
• Analysis of a technical capabilities roadmap enabling key support capabilities [1]

All of the efforts to date have focused on the achieving support for the following key ‘outcomes’:

• Increasing the number of science experiments by reducing set up and execution times
• Enabling multiple discipline investigators collaboration

It is anticipated that the next phase of the MAP program will continue to couple models and will need continued infrastructure support as above, albeit with less focus on development and more focus on operational support in expanded super-computing environments (e.g. initial tools are available at NCCS/GSFC, but should also be available at NAS/ARC, and other select investigator facilities).

However, it is also expected that the MAP Modeling Environment infrastructure may need to consider additional mechanisms particularly focused on increased support for analysis capabilities in order to respond to evolving community needs as described by Dr. Eric Barron to the House science committee:

“…. Evaluation and assessment of model capability will increasingly be the focus of future measurement activities. Demonstrating model capability is likely to be a driver for developing and evolving observation systems and field campaigns. …. The demand to understand the connection between climate and specific impacts on natural and human systems will require a more comprehensive approach to environmental observation and modeling in order to integrate the multiple stresses that influence human and natural systems (i.e. climate, land use, and other human stressors such as pollutants)….” (Dr. Eric Barron, house science committee testimony)

The scientific community may consider new infrastructure tools that aid and enhance integrating the analysis of models, observations, and impacts on human and natural systems. Some ideas are as follows, and are stated as ‘straw man concepts’ to encourage community feedback, with endorsement, rejection, or addition all being valuable inputs.

• Development tools enabling direct comparisons between observations and differing models
• Model component archiving, compatibility checking, and automated assembly
• User Interfaces enabled by service-oriented architectures targeted on model-specific assessment efforts (e.g. ensemble efforts, OSSE analysis, specific application needs)
• Expanded host platforms beyond the traditional NCCS and NAS super-computers (e.g. newer cell processing architectures, distributed grids of collaborating models akin to a model-web, etc.)
• Links to standard assessment tools (e.g. Taylor diagrams, reference observations)
• Assessment and development of a community model collaboration approach integrating NASA remote sensing
• Project management approaches to recommend model-based short term activity support (e.g. field campaign support)

[edit] Community-Supported GEOS5 Users' Guide & Tutorial

NASA's Software Integration & Visualization Office (SIVO) has the responsibility of deploying the GEOS5 atmospheric model for use by researchers at universities and Government laboratories. The services include hosting of the modeling software and documentation from a central repository, assistance with porting the model to different hardware platforms, and help with coupling the model to external components through the use of the Earth System Modeling Framework (ESMF).

In October 2007 SIVO will begin work on a GEOS5 Users' Guide to assist researchers with new deployments of the model. Although SIVO will write the baseline document the community is encourage to contribute any knowledge or suggestions related to using the model, as well as quick references to solutions to problems. The SIVO Knowledge Base, known as the "Modeling Guru", will be used to generate and distribute the live document. The web site (not functional at the time of this writing) is [2] .

The community is encouraged to participate in this effort, and comments and suggestions are welcomed.

--207.172.83.193 06:25, 8 August 2007 (EDT)Michael Seablom, SIVO Head

[edit] Help Using the Wiki Software

Here is the cheatsheet for fast starts. Cheatsheet for Fast Editting.

Here is a link to mapmodeling.org help and the more complete information from wikimedia: [mapmodeling.org help]

Consult the User's Guide for information on using the wiki software.

simple instructions for starting a new page

mediawiki instructions for starting a new page.

• Configuration settings list
• MediaWiki FAQ
• MediaWiki release mailing list