Development of Modern Numerical Weather Prediction CC Hennon

Development of Modern Numerical Weather Prediction CC Hennon ATMS 350 UNC Asheville Vilhelm Bjerknes Vision 1901 Wanted to incorporate physics into weather forecasting Start with complete set of initial conditions (3-D) Solve equations using graphical methods Initial state not sufficient for good

forecasts Did not use continuity equation to derive the initial vertical wind component (no direct measurements available) Source: Historical Essays on Meteorology 1919-1995, AMS CC Hennon ATMS 350 UNC Asheville Lewis F. Richardson About same time as Bjerknes; WWI ambulance driver Used continuity equation to obtain initial vertical velocities, as well as the other primitive

equations Failed due to insufficient initial data Solved equations by hand! Time steps were too large would have resulted in computational instability Source: Historical Essays on Meteorology 1919-1995, AMS CC Hennon ATMS 350 UNC Asheville Further Developments Carl-Gustaf Rossby (1939) Showed that atmospheric longwave motion could be explained by vorticity distribution

Wave movement function of wavelength and speed of large-scale zonal flow (Rossby Waves) Jule Charney (1949) Developed first barotropic model Large-scale motions approximately geostrophic and hydrostatic; no vertical motions; no vertical wind shear Numerical prediction now realizable as soon as computers become powerful enough to run the computations Source: Historical Essays on Meteorology 1919-1995, AMS CC Hennon ATMS 350 UNC Asheville

First Numerical Forecast Charney barotropic model run on ENIAC computer (1950) Produced 500 mb height forecast Bad forecast but looked realistic ENIAC Computer Jule Charney Source: Historical Essays on Meteorology 1919-1995, AMS CC Hennon ATMS 350 UNC Asheville Operational Numerical Weather

Prediction May 6, 1955 First regular and continuing NWP forecasts issued for U.S. Early results worse than lab experiments Limitations of assumptions made in the models (Quasi-geostrophic approximation) Many storms missed; public confidence wanes Many early problems due to bad input data from global observation networks Barotropic model (Charney) worked best for several years Barotropic processes mostly controlled large-scale daily motions (e.g. long waves), while baroclinic controlled shortbursts of activity (e.g. mid-latitude cyclones) Source: Historical Essays on Meteorology 1919-1995, AMS CC Hennon ATMS 350

UNC Asheville The Final Major Evolution Successful NWP using full suite of primitive equations occurred in 1966 Original Bjerknes/Richardson vision! Advances in computational power and improvements in input data let to acceptable forecasts Forecasts improved almost 50% in 10years (1955-1965) from subjective forecasts Source: Historical Essays on Meteorology 1919-1995, AMS CC Hennon ATMS 350 UNC Asheville

CC Hennon ATMS 350 UNC Asheville Types of Numerical Models Barotropic Model Barotropic atmosphere (constant density/temperature on pressure surface, no vertical motion) Absolute vorticity conserved d f 0 dt Somewhat skillful at large-scale wave prediction

CC Hennon ATMS 350 UNC Asheville Types of Numerical Models Primitive Equation Models (Dynamical) The primitive equations are fundamental relationships that govern atmospheric motion Equations of motion (3) (cons. of momentum) Continuity equation (cons. of mass)

First law of thermodynamics (cons. of energy) Moisture equation (cons. of moisture) Equation of state First operational baroclinic models Use began in the mid-1960s CC Hennon ATMS 350 UNC Asheville Other Models Statistical Forecasts based off of past events Cannot predict extreme events Knowledge limited to what has occurred

before Statistical-Dynamical (hybrid) Combines some NWP output with historical statistics Example: The SHIPS hurricane intensity model uses model fields as predictors for a statistical regression CC Hennon ATMS 350 UNC Asheville