CURRENT RESEARCH

THORPEX: Predicting the effects of routine observations on 3-7 day forecasts

AOSN-II Ocean Observing Experiment

Targeted Observations to Improve Hurricane and Typhoon Forecasts

Predictability experiments using a barotropic model

Research Overview

	Adaptive Observations The bulk of my research in recent years has focused on identifying locations in which extra "adaptive" observations should be taken, in order to improve a particular numerical forecast. These locations change from day to day, depending on the flow. The most common application of adaptive observing strategies to date has been to deploy aircraft laden with GPS dropwindsondes over the Pacific Ocean, in order to improve 1-4 day forecasts of winter weather over North America. These Winter Storm Reconnaissance Programs have been operational at the National Weather Service since 2001. I am currently working on adaptive observing strategies for: (1) forecasting hurricanes, and (2) coastal ocean prediction (see below). A general WEBSITE on adaptive observations is being developed (slowly)! Funding by the NOAA THORPEX program is gratefully acknowledged.
Tropical Cyclones involved in improving hurricane forecasts. One of my current projects is on refining an adaptive observing strategy, to deploy the best sequence of observations from the synoptic surveillance aircraft (NOAA G-IV above), in an attempt to improve the forecast of a landfalling hurricane. The data (winds, temperature, pressure, humidity) collected by the hurricane hunters are assimilated into the operational numerical forecast models along with all the regular data (from satellites etc).   Funding by the NOAA Joint Hurricane Testbed initiative, in collaboration with NOAA Hurricane Research Division and Environmental Modeling Center, is gratefully acknowledged. Satellite data are making a massive contribution to weather forecasting. Another area of interest for me is to investigate the types of satellite data that are most beneficial for hurricane analysis and prediction, and the methods by which these data ought to be assimilated into the models.
	Coastal Ocean Prediction (My AOSN Page) New underwater observations are being deployed in coastal regions.  In parallel, numerical modeling and data assimilation techniques are improving all the time. Through an integrated capability of these aspects, a better understanding and prediction of previously little-known physical processes can be realized. One experiment that focuses on these goals is the Autonomous Ocean Sampling Network II (AOSN-II), which is taking place in Monterey Bay between August-September 2003. One major goal of AOSN-II is to improve the understanding and 1-3 day numerical forecasts of upwelling plumes in the Bay. Physical, chemical and biological variables are observed by a multitude of instruments (aircraft, underwater gliders and floats, radar, AXBTs etc). My own interests are in (1) adaptive sampling of underwater gliders, and (2) designing ensemble forecasts to capture the variability of the ocean's behaviour, particularly with respect to wind stresses induced at the ocean surface. Funding by the Office of Naval Research, via the Monterey Bay Aquarium Research Institute, is gratefully acknowledged.
Ensemble Forecasting There is much benefit to be derived from producing an ensemble of many forecasts, compared with a single forecast. In principle, the ensemble can give estimates of the probability that a particular event is going to happen (>6 inches of snow, a Cat-4 hurricane strike, an upwelling plume). Moreover, information in ensembles may be very fruitful in specifying error statistics for data assimilation. In practice, ensemble forecasts are expensive, and there are sometimes large errors in the initial conditions, the forecast model, boundary conditions, and external forcings (atmospheric forcing of coastal ocean currents, or ENSO). Methods to account for these error sources are being developed for a wide variety of atmospheric and oceanic situations: I am investigating a few of these.
Pa = Pf- PfHT( HPfHT +  R )-1H Pf	Data Assimilation With the explosion of both computing power and geophysical data, there is now a pressing need to develop clever methods to produce the most out of information contained in the data and numerical models. New techniques such as ensemble-based Kalman filters (using an ensemble to specify Pf on the left) are showing great promise in a hierarchy of models. As with ensemble forecasting, data assimilation is a common theme that runs through my research. Since I get asked so often about how ensemble-based data assimilation works, I keep planning to write a short primer to distribute to interested people. All I have managed so far is a bibliography.