Mr. Reynolds began his talk by explaining that the Monterey Office supplies the forecasts that are the foundation of many weather reports. Not just the ones used by broadcast media, but also pinpoint forecasts used by tug boats, airports, and other special applications. Groups like firefighters sometimes ask questions like "is it safe enough to drop flame retardant on this wildfire?", which can be nontrivial to answer.

The NWS computer system puts all of the current weather information into a grid inside their computer system. Current weather information includes measurements from observation posts (airports, volunteers, weather buoys, etc.), satellite data, and interpolations for data points between measurements. This is then crunched with weather models to predict the conditions for the next seven days. People can access these forecasts in a variety of ways, including over the internet from cell phones and laptops.

Mr. Reynolds spent quite a bit of time talking about weather forecasts for San Francisco Airport (SFO). Because of the fog, we have 97% of our delays caused by weather. Only Denver has a higher percentage, 98%. A team including the NWS, San Jose State, NASA and others worked together to develop a better weather predicting model. They developed one good enough that they could schedule airplane arrivals for just after the weather clears. This required as much as five hours or more of accurate lead time, because they take off from distant cities. Delays cost airlines at least $74/minute that airplanes stay airborne. Since going live in 2004 the new system has saved $4.2 million per year.

The NWS has two interesting special forecasts, space forecasts and tsunami warnings. Space forecasts are valuable to the national security and communications industries. Solar flare activity is currently low compared to most of the previous decade. Sun spots are expected to peak again in 2014. Tsunami warning systems have gotten more integrated around the Pacific since that Indonesian one that killed a lot of people a couple of years ago. NWS is looking to improve their coverage in the Caribbean.

The NWS likes to improve data collection by partnering with other institutions. KPIX (CBS 5) has a weather radar that has significantly improved the ability of the NWS to predict flooding in the Russian River watershed. They also partner with individuals, who can maintain weather stations in their back yards and report the information for inclusion in the data grid that is used to make forecasts. This can improve the accuracy of the results.

Another tool for improving forecasts is NEXRAD Dual Polarization radar, which produces better quality images. The NWS plans to retrofit this into all 122 of their weather radar stations over the next four years.

The NWS has a very sophisticated computer system, including "Two redundant IBM Power6-based systems containing 156 nodes (32 processors per node), 4 GB memory/physical processor and IBM's AIX-based high performance computing parallel environment including GPFS. Each system is capable of delivering 73TF as measured by Linpack benchmark code. Each operational system supports a 320 TB disk storage subsystem providing sustained performance of not less than 9 GB/sec for write and 10 GB/s for read." These are primarily used for weather forecasting. Compute time is scheduled far into the future for things like daily forecasts and climate model testing.

Looking forward, the NWS has to deal with changes in the demands of office holders, new technology, budget constraints, and the marketplace. Their goal is to continue to provide an accurate and timely forecast so that increased industry reliance on the forecast can just be cost effective planning.

It takes the NWS ten years to adopt a new technology. That's way down from what it used to take.

The highest risk of flooding in the USA is in New Orleans. The second highest risk of flooding is in Sacramento. California's worst recorded floods to date happened in 1862.

This year we expect a dryish winter. It's a La Nina which will push the precipitation to the north of us.

Models don't predict the behavior of boundary layers very well. Too much turbulence there.