A new approach for mesoscale surface analysis: the space-time mesocale analysis system

In recent years the number of surface observations over the U.S. with spatially dense time and space coverage has grown rapidly, resulting in an essentially national mesonetwork offering the possibility of performing frequent monitoring of mesoscale features that may generate significant local weather, as well as hazards to aviation. Data acquisition systems to collect, store, and offer easy access to the data have arisen (e.g. MADIS). Despite these advances, the density of the mesonet data remains highly variable across the country, literally being composed of oases and deserts of data. Traditional approaches to objective analysis (successive correction methods, optimal interpolation, etc.) have great difficulty properly representing details in the oasis regions without introducing undesirable noise in the analyzed fields in the desert regions. At the same time, the National Digital Forecast Database (NDFD) has a nominal grid spacing of 5 km across the United States, necessitating the development of a high-quality, very detailed, and robust surface mesoanalysis system.

Given these needs, we have recently been developing a spatial-temporal objective surface mesoanalysis scheme offering highly detailed analyses at frequent intervals (5-15 min) where the data support such detail, while avoiding noise in other parts of the analysis domain where only coarser-scale features can be resolved. Our Space-Time Mesoscale Analysis Scheme (STMAS) is designed to fully exploit spatial variability in data density and reporting frequency allowing small-scale features such as thunderstorm gust fronts and lake breezes to be revealed and tracked in a time consistent manner. STMAS is designed to be compatible with current workstation display capabilities that allow compositing of fields and looping. The need for this rapid updating feature within the limits imposed by the current WFO computer environment meant that the scheme had to be efficient and robust.

STMAS has three components: data quality control, analysis processing, and product generation. The data quality control is based on a Kalman filter approach operating in observation space. The Kalman filter individually models each observational site based upon self-trend, buddy trends, and external forcing. The net result is that each observation in the domain has a unique projection “engine” that provides a one data-cycle “forecast” value useful for quality control. The analysis engine has two options currently: a space-time recursive filter or a spectral wavelet approach, both applied to the observational network spatially and temporally. Both schemes use iterative procedures to sequentially add more detail. The first pass defines the large-scale structure. This structure is retained and residual differences between the observations and the first pass become the input to the analysis in the subsequent pass. Although this part of the procedure is similar to a standard successive corrections scheme, STMAS then iterates across the grid in a variational attempt to minimize a global penalty function that includes terms for optimum least square matching of the observations and smoothness, until the residuals are close to the range of observation error. The wavelet scheme uses a set of local basis functions to fit the observations, but here the approach is to discretize the analysis domain into subregions of varying size depending on the data density in time and space. As in the case of the recursive approach smoothness, constraints may be applied. Such an approach should allow better analysis in situations where data is inhomogeneous and meteorological systems are on a variety of spatial and time scales.

The scheme, run on a 5-km grid at 15-min intervals, is demonstrated for severe weather events in the Northeastern US for the Federal Aviation Administration’s Corridor Integrated Weather System (CIWS). We will compare the two approaches against a standard successive corrections method (Barnes scheme) currently available on AWIPS, and assess the capability of STMAS to reveal important mesoscale features that lead to hazardous local weather. The primary goal of this research is to exploit the huge number of surface observations now available for mesoscale diagnosis and nowcasting. This will be the emphasis of the results to be presented.