Alumni Project
Modeling and Analysis of the Earth’s Hydrologic Cycle
PI:
Donald R. Johnson
Space Science and Engineering Center
University of Wisconsin - Madison
Summary
A key aim of this project is to advance the understanding of global water vapor and inert trace constituent transport in relation to climate change through analysis of simulations produced by the global University of Wisconsin (UW) hybrid isentropic coordinate models. Research has established that simulations in isentropic coordinates are remarkably more accurate in the transport of water vapor, inert trace and chemical constituents compared against corresponding sigma coordinate models.
Introduction
The primary objectives of this research are to 1) advance the modeling of climate change by developing a hybrid isentropic model for global and regional climate simulations, 2) advance the understanding of physical processes involving water substances and the transport of trace constituents, and 3) theoretically and diagnostically examine the limits of global and regional climate predictability. Model development has proceeded through a hierarchy of models from a rudimentary channel model to a regional model to the full fledge global climate model in use today [the University of Wisconsin-Madison isentropic-eta (UW q-h) coordinate model]. Model development has entailed studies of a wide variety of Eulerian and semi-Lagrangian numerics and different algorithms for parameterization of boundary layer processes and dry and moist convection.
Significance of Research
This research addresses fundamental issues underlying the understanding and modeling of hydrologic processes. Hydrologic processes, including surface evaporation, long-range transport of water substances, release of latent heating involving evaporation/condensation and attenuation of the radiative flux of energy by water vapor and clouds, are central in the regulation of global and regional climate.
A premise of this work is that accurate simulation of the long-range transport of mass, momentum, energy, water vapor and other trace constituents is crucial for modeling the climate state. Substantial evidence exists which illustrates the inherent difficulties in simulating the long-range transport of water vapor and other trace constituents utilizing conventional sigma coordinate climate models. This limitation stems primarily from problems associated with vertical advection of water vapor and other trace constituents in sigma models. Since the transport of water vapor and trace constituents lies within inclined isentropic layers, the water vapor structure and its transport is primarily two dimensional in isentropic coordinates except in regions of moist convection. The problems associated with vertical advection remain negligible in isentropic coordinates relative to sigma coordinate models over much of the global domain.
The UW Model
The UW q-h model is a grid point model having the option to use either centered differencing, a van Leer type scheme or the piecewise parabolic method (PPM) for transport numerics. The standard model runs with PPM transport numerics and two-level forward time differencing. The model incorporates the full suite of physical parameterizations obtained from the National Center for Atmospheric Research (NCAR) community climate model version 3 (CCM3) including a land-surface model and slab ocean-thermodynamic sea ice model.
Accomplishments
A key aim of this project is to advance the understanding of global water vapor and inert trace constituent transport, in relation to climate change through analysis of simulations produced by the global UW hybrid isentropic coordinate models. During the past year the P.I. has interacted with Professor David Randall of Colorado State University and scientists at NOAA’s National Centers for Environmental Prediction (NCEP) as NCEP’s Special Project Scientist.
Numerous integrations exceeding 14 years have been performed with the UW q-h model at 2.8125° latitude-longitude resolution. Daily 5-day forecasts at 0.7° latitude-longitude resolution (approximately 4,000,000 model grid points) have been ongoing for the last 10 months. Both of these efforts attest to the capability of the UW model for numerical weather prediction (NWP) and climate simulation.
Computational Approach
The UW model can currently use multiple processors on one node of the IBM/SP using OpenMP parallelism. An effort is underway to convert the code to use MPI message passing to enable the use of multiple nodes on the IBM/SP. This capability, which is 70-80% complete, will greatly enhance our research efforts by providing much faster throughput for climate simulations. The use of multiple nodes will allow substantial increases in both horizontal and vertical resolution for climate simulation and NWP. The capability also allows the use of more realistic (although more computationally intense) physical parameterizations. The use of the terascale computing will provide the capability to conduct longer and more complex climate simulations in a reasonable time and greatly facilitate the determination of the impact of different physical parameterizations and different spatial resolutions.
In the MPI version of the model, common blocks of meteorological variables have been converted to modules using allocatable arrays. MPI will be used to transfer data between the nodes for the "ghost" values from adjacent latitudes for PPM transport. The land-sea model will be setup to store and calculate values only for the land points within the latitudes handled by a particular processor.
Future Plans
During the coming year, a primary goal of our research effort will be to complete the conversion of the UW model to use MPI message passing to allow the efficient use of multiple nodes of the IBM/SP. The completion of this effort will greatly facilitate the achievement of the scientific goals of our research effort.
Numerous seasonal and multi-year simulations will be analyzed to further our investigations of regional energy balance and to determine the role of atmospheric hydrologic processes on determining regional climate. This effort will also investigate the relative abilities of isentropic and sigma coordinate models to transport water vapor and explicitly predict cloud substances and diagnose the impact of the explicitly simulated clouds on cloud radiative forcing. The impact of increased horizontal and vertical resolution will also be ascertained.
The increased processing capability will allow assessment of the impact of a more sophisticated representation of moist convection and the role of clouds in determining climate. In this effort a sophisticated moist convective scheme developed by NASA scientists will be implemented in the UW model and compared with the current parameterization. The new scheme is more computationally intensive but may provide for improved representation of atmospheric processes.
The aim of the efforts discussed above is to take advantage of the increased numerical accuracies of long range transport utilizing hybrid isentropic coordinates to improve the simulation of moist reversible processes involving water vapor and cloud water and ice and related hydrological processes. This increased accuracy is essential to increasing the accuracies of other important processes that interact strongly with other hydrologic, chemical and biospheric processes, all of which are critical to accurately simulating the climate state regionally and globally.
Resources Needed
Approximately 200 Gb of HPSS data storage in 4000 files are required for this project. The default 10 Gb of permanent space is sufficient. A 17-year run will use about 50 Gb of scratch space. On average we expect to access the 200 GB of HPSS data storage 3 times over the course of a year. The total amount to be moved would be 600 GB.
The current network throughput will be adequate for FY03.
back to project page
