Lecture 24 Weather Forecasting What should I wear today How should the Forest Service fight a summer wildfire Should Jack go skiing this weekend Should Ace Hardware stock more umbrellas this fall When should Paul Bunyan plant his wheat How many snowplows does Seattle need For what rate should Allstate insure homes on the North Carolina coast Do we need to worry about global warming require weather climate forecasts for today a few days months decades How are forecasts made How good are they How good can they be We focus mainly on midlatitude weather forecasts out to a week ahead where most progress has been made How are forecasts made 1 Current weather conditions observed around the globe balloons satellites surface aircraft etc 2 Observations sent over Global Telecommunications System to weather centers In US National Centers for Environmental Prediction or NCEP Camp Springs MD 3 Center quality controls and analyzes observations 4 Computer model at center makes global or national weather forecast maps 5 Center disseminates observations and forecast information to public and private agencies such as regional National Weather Service forecast offices e g at Sand Point in Seattle 6 Agencies use computer model maps and knowledge of local weather patterns to issue local and regional forecasts 7 News media broadcast these forecasts Time is of the essence In a miracle of technology and communication 1 3 day forecasts are typically broadcast in US within 6 hours of the observations on which they are based Weather Observations Weather is global and requires global observations to forecast Through the cooperation of 130 member countries the UN World Meteorological Organization ensures uniform observation procedures and transmission protocols and free access to all available data Data see ECMWF link on 101 weather web page for current example Radiosondes Wind press temp humidity profiles Pilot Balloons Wind profiles only Satellites Geostationary cloud track winds cloud heights Polar orbiters temperature profiles 1 2 km resolution humidity profiles 3 5 km resolution surface wind over ocean Surface Observations Weather reporting stations over land Buoys Commercial ships and aircraft Numerical Forecast Model Atmospheric model uses mathematical equations to predict changes in temperature humidity pressure winds etc due to airflow precipitation radiation turbulence friction etc These equations are approximations that represent to the best of our scientific knowledge the net effects of smallscale atmospheric motions on the larger scale motions that we can accurately observe The model is formulated in terms of the atmospheric state at a set of grid points typically at 15 50 height levels at up to 20 000 horizontal positions Using the observations with the prior forecast to fill in gaps the model starts with its best guess at the current atmospheric state at the grid points The model equations are solved stepping the forecast atmospheric state 520 mins forward at a time Billions of computations on the world s most powerful supercomputers go into a several day forecast Forecast maps 500 mb height 850 mb temp precipitation etc printed each 12 hours into forecast Forecast Model Grid Holton fig 3 7 Horizontal position Gridpoints are at box corners Grid follows terrain Some Current Forecast Models NCEP Global Forecast System GFS model global run 4x daily out to 16 days Grid points spaced at 50 km distance from Seattle to Tacoma ETA model Covers North America and adjacent regions run out to 84 hours every 6 hours Grid points spaced at 12 km distance from Seattle Bellevue Runs in 2 hours UW Seattle NWS Mesoscale Model MM5 Covers E Pacific Western N America run out to 72 hours every 12 hours Grid points every 12 km across Pacific NW 4 km over western WA ECMWF European Centre for Medium Range Weather Forecasts England Like GFS model 25 km resolution run out to 20 days using data from 12 GMT daily most skillful in the world Pacific NW topography for 36 km grid spacing www atmos washington edu under Weather Climate Info MM5 Olympic Mts Puget Sound are smoothed over 12 km resolution is required for these features which are quite important to our local weather How does forecaster turn maps into a forecast Local geography may be smeared out in model Model often has small biases too low temperatures too much rainfall W of Cascades but not enough E of Cascades Model forecast not perfect radar satellite info may suggest improvement The human forecaster has an important job Model predicted surface temperatures must be adjusted and local rainfall patterns and probability of precipitation predicted using experience and statistical models Small scale processes such as thunderstorms valley fog or low cloud formation must be inferred from the maps based on forecaster experience Forecaster must make forecasts weather warnings intelligible to the public for many zones across Washington state If forecast appears systematically off forecaster must use updated current weather information from satellites radar surface obs to correct it