In meteorology, we've delved into decision science. How do we get people to respond appropriately to threats. This is is a rather significant issue because of the number of deaths from various weather events, most of which would be avoidable if people were to take action when faced with National Weather Service warnings.
All too often, we've heard, "The tornado came out of nowhere..." when in fact, a severe weather event had been foreseen and information published to local resources 24 hours (or more) before, and the NWS and local media had been talking about the event for nearly as long. Instead of moving to safety when a warning is issued, too many people grab a cellphone, rush outside, and look for the storm, to take photos. Or ignore it completely, because they feel it can't happen to them.
Warning fatigue has been hypothesized as an issue. NWS took action to reduce the number of warnings for large areas, instituting "Storm Based Warnings" in the early 2000s, and reducing the area warned by nearly 80%, thus making warnings more specific. There may have been a brief improvement in response, but not one we could actually measure. New tools and large programs have ensued, trying to increase warning times, specificity and locality, but it seems the only people who are paying any attention are those already deeply involved in decision support processes and who understand the risk. The public is able to watch the devastation (think of the storms in the last week!) but in 10 days, when another round of storms pushes through the central US, they'll have forgotten, because it didn't happen to them, so they're immune, or, it was too far away and can't happen in my city/town, or, they've literally forgotten it could happen to someone.
I've been working on this problem with NWS for close to 26 years (external participant), and just realized this is a similar phenomenon, thanks to this discussion.
Great commentary Gerry! Decision Science is a great field! I sometimes feel that more and better data to share with the public isn't a solution. It may be part of the problem. People may feel swamped with data and develop a species of resistance to it. If we look at a computational model of consciousness, then we might say in terms of decisions people rely on data, memory, and powerful algorithms in concert. Many of the latter are involved with pre-conscious Bayesian probability calculations. There is a local weatherman who while using the various forecasting models, extols the GOLU model- "go out, look up" approach. That's not as simplistic advice as it sounds. It really our own internal ensemble process. We all use it, but when it comes to strictly meteorologicsl forecasting were well sdvised to rely on the experts
I just got done with a little exercise doing operational weather forecasting for a flying exercise for 2 weeks. Almost every day, I warned, in the briefing, that the models were not in good agreement beyond about 24 hours (Sort of disturbing: We usually can get 7-10 days of reasonable data out of several of them) and that I'd be consulting the Weather Rock for real time updates. Said Rock allowed me to get 'em flying several days when the forecasts did not verify (weather was better than models, and even humans, thought).
In the forecasting world, you have to possess an innate knowledge of the way weather develops in an area. To that end, a new forecaster should probably work with an experienced one for some period of time... 6-12 months... to grasp how things work on the local starship.
NOAA and NWS are employing a number of social scientists now to try to improve the decision support landscape around severe weather events... severe thunderstorms, rain and flooding, tornadoes, hurricanes, and of late, increased emphasis on winter storms. That said, I need to see if they've an epidemiologist involved in severe weather, or if we'll simply continue to call that "climatology"?
In the work we did on Storm Based Warnings, we attempted to refine the area where a severe storm (thunderstorm or tornado) was likely to have significant impact. This involved a human in the loop, drawing a polygon defining where the storm was expected to proceed in the next 15-45 minutes. The polygon approach reduced the warned area by some 80%, when compared to the older way, where we warned entire counties even when only a small portion was likely to be affected. Part of this related to how warnings were disseminated. Storm Based Warnings were the first to be disseminated as web-based products with a probability associated with them. Newer research is using both high-frequency, high-resolution ensemble weather models and algorithmic parsing to define a threat track that can be shown with time hacks for arrival. The models mentioned are a part of a 10 year program, now well into its 14th year, to create a numerical modeling system that could produce time-critical weather warnings based on model result, rather than waiting for an observation-based warning, either directly observed or remotely sensed (radar). This effort has seen some significant advances in modeling, but we're reaching the conclusion that, providing perhaps an hour to the warning timeframe might lead the public to take less head of the need to take cover and remain safe. Based on several studies, we're now targeting the output of this program, as well as the algorithmic threat tracker, to emergency managers who are more likely to understand the need for timely action. NWS continues to depend on private sector partners, as well as tools such as web pages and NOAA Weather (All Hazards) Radio to disseminate warnings. This includes the media and services that send warning data to cellphones and mobile devices.
The Bayesian decision process is more complicated than most appreciate, in my estimate. For that matter, most of us intuitively do utilize a Bayesian approach to a number of daily decisions, but subconsciously. And, some people can make stellar decisions based on sparse or incomplete data while others cannot make a viable decision even with complete data. So much of their decision process is, as you note, based on memories: I don't need to evacuate because I rode out a Cat 3 hurricane as a kid and it wasn't that bad... That the parameters, if not the specific metrics of a hurricane have changed (larger wind fields, more likelihood of significant storm surge) while the deterministic parameters of classification have not is often lost on the public. As an experienced tropical cyclone researcher, I couldn't believe the models' precipitation estimates for Hurricane Harvey in 2017 (nor the wind field fetch size of Katrina in 2005) based on my own experience. I've reset my "normals", much as I had to do routinely during the pandemic as more data became available.
Awesome expertise on display in your remarks. I live in the northern Willamette Valley and forecasting for NWS is difficult to say the least. What with continental and oceanic airmasses colliding, the Cascade Range doing its adabiatic thing with downward Westerly flow, the Gorge acting as a conduit for high winds, microclimates galore, the urban heat island in the pdx/Vancouver area... We've had major winter storm warnings for storms that never occurred and on the flip side, off shore clipper lows that stalled out and brought heavy rain/snow for days! This winter one forecast had us getting a dusting, but instead we got eight inches of snow over ice!
Just recently, here in Colorado Springs, we were forecast for a dusting overnight. We got 5.25 inches (measured), but it was all gone, some early sublimation, and later, melting, by nightfall.
I've threatened to trademark the term, "Winter weather forcasting is HARD."
It's such a small world! I used to live in C-Springs fifty years ago- 4th INF Div. stationed at Ft Carson! I lived in the east side of town on San Miguel. I imagine the town has grown up a lot since then. I had BOQ off base so it was like a commuter job. Beautiful place. I used to take my sports car up Pikes Peak on a curvy dirt road that I imagine is all paved by now. Good memories.
My wife's a retired Army Nurse. We go to Carson's Commissary because going on an Army base is more familiar than an Air Force (or Space Force) base... and to be honest, the Carson Commissary's laid out better and has more stock than Peterson SFB's does (the Academy Commissary is a cute little place, perfect for the Academy).
The road up Pikes Peak is, indeed, paved, but still a slow drive because of the twists and turns. One of my goals in retirement is to start doing 14ers like the locals, but I've got a little bit more physical training to do first!
The Cambridge Elements series has a nice series of small books on Decision Science for any interested in this fascinating field.
In meteorology, we've delved into decision science. How do we get people to respond appropriately to threats. This is is a rather significant issue because of the number of deaths from various weather events, most of which would be avoidable if people were to take action when faced with National Weather Service warnings.
All too often, we've heard, "The tornado came out of nowhere..." when in fact, a severe weather event had been foreseen and information published to local resources 24 hours (or more) before, and the NWS and local media had been talking about the event for nearly as long. Instead of moving to safety when a warning is issued, too many people grab a cellphone, rush outside, and look for the storm, to take photos. Or ignore it completely, because they feel it can't happen to them.
Warning fatigue has been hypothesized as an issue. NWS took action to reduce the number of warnings for large areas, instituting "Storm Based Warnings" in the early 2000s, and reducing the area warned by nearly 80%, thus making warnings more specific. There may have been a brief improvement in response, but not one we could actually measure. New tools and large programs have ensued, trying to increase warning times, specificity and locality, but it seems the only people who are paying any attention are those already deeply involved in decision support processes and who understand the risk. The public is able to watch the devastation (think of the storms in the last week!) but in 10 days, when another round of storms pushes through the central US, they'll have forgotten, because it didn't happen to them, so they're immune, or, it was too far away and can't happen in my city/town, or, they've literally forgotten it could happen to someone.
I've been working on this problem with NWS for close to 26 years (external participant), and just realized this is a similar phenomenon, thanks to this discussion.
Great commentary Gerry! Decision Science is a great field! I sometimes feel that more and better data to share with the public isn't a solution. It may be part of the problem. People may feel swamped with data and develop a species of resistance to it. If we look at a computational model of consciousness, then we might say in terms of decisions people rely on data, memory, and powerful algorithms in concert. Many of the latter are involved with pre-conscious Bayesian probability calculations. There is a local weatherman who while using the various forecasting models, extols the GOLU model- "go out, look up" approach. That's not as simplistic advice as it sounds. It really our own internal ensemble process. We all use it, but when it comes to strictly meteorologicsl forecasting were well sdvised to rely on the experts
I just got done with a little exercise doing operational weather forecasting for a flying exercise for 2 weeks. Almost every day, I warned, in the briefing, that the models were not in good agreement beyond about 24 hours (Sort of disturbing: We usually can get 7-10 days of reasonable data out of several of them) and that I'd be consulting the Weather Rock for real time updates. Said Rock allowed me to get 'em flying several days when the forecasts did not verify (weather was better than models, and even humans, thought).
In the forecasting world, you have to possess an innate knowledge of the way weather develops in an area. To that end, a new forecaster should probably work with an experienced one for some period of time... 6-12 months... to grasp how things work on the local starship.
NOAA and NWS are employing a number of social scientists now to try to improve the decision support landscape around severe weather events... severe thunderstorms, rain and flooding, tornadoes, hurricanes, and of late, increased emphasis on winter storms. That said, I need to see if they've an epidemiologist involved in severe weather, or if we'll simply continue to call that "climatology"?
In the work we did on Storm Based Warnings, we attempted to refine the area where a severe storm (thunderstorm or tornado) was likely to have significant impact. This involved a human in the loop, drawing a polygon defining where the storm was expected to proceed in the next 15-45 minutes. The polygon approach reduced the warned area by some 80%, when compared to the older way, where we warned entire counties even when only a small portion was likely to be affected. Part of this related to how warnings were disseminated. Storm Based Warnings were the first to be disseminated as web-based products with a probability associated with them. Newer research is using both high-frequency, high-resolution ensemble weather models and algorithmic parsing to define a threat track that can be shown with time hacks for arrival. The models mentioned are a part of a 10 year program, now well into its 14th year, to create a numerical modeling system that could produce time-critical weather warnings based on model result, rather than waiting for an observation-based warning, either directly observed or remotely sensed (radar). This effort has seen some significant advances in modeling, but we're reaching the conclusion that, providing perhaps an hour to the warning timeframe might lead the public to take less head of the need to take cover and remain safe. Based on several studies, we're now targeting the output of this program, as well as the algorithmic threat tracker, to emergency managers who are more likely to understand the need for timely action. NWS continues to depend on private sector partners, as well as tools such as web pages and NOAA Weather (All Hazards) Radio to disseminate warnings. This includes the media and services that send warning data to cellphones and mobile devices.
The Bayesian decision process is more complicated than most appreciate, in my estimate. For that matter, most of us intuitively do utilize a Bayesian approach to a number of daily decisions, but subconsciously. And, some people can make stellar decisions based on sparse or incomplete data while others cannot make a viable decision even with complete data. So much of their decision process is, as you note, based on memories: I don't need to evacuate because I rode out a Cat 3 hurricane as a kid and it wasn't that bad... That the parameters, if not the specific metrics of a hurricane have changed (larger wind fields, more likelihood of significant storm surge) while the deterministic parameters of classification have not is often lost on the public. As an experienced tropical cyclone researcher, I couldn't believe the models' precipitation estimates for Hurricane Harvey in 2017 (nor the wind field fetch size of Katrina in 2005) based on my own experience. I've reset my "normals", much as I had to do routinely during the pandemic as more data became available.
Awesome expertise on display in your remarks. I live in the northern Willamette Valley and forecasting for NWS is difficult to say the least. What with continental and oceanic airmasses colliding, the Cascade Range doing its adabiatic thing with downward Westerly flow, the Gorge acting as a conduit for high winds, microclimates galore, the urban heat island in the pdx/Vancouver area... We've had major winter storm warnings for storms that never occurred and on the flip side, off shore clipper lows that stalled out and brought heavy rain/snow for days! This winter one forecast had us getting a dusting, but instead we got eight inches of snow over ice!
Just recently, here in Colorado Springs, we were forecast for a dusting overnight. We got 5.25 inches (measured), but it was all gone, some early sublimation, and later, melting, by nightfall.
I've threatened to trademark the term, "Winter weather forcasting is HARD."
It's such a small world! I used to live in C-Springs fifty years ago- 4th INF Div. stationed at Ft Carson! I lived in the east side of town on San Miguel. I imagine the town has grown up a lot since then. I had BOQ off base so it was like a commuter job. Beautiful place. I used to take my sports car up Pikes Peak on a curvy dirt road that I imagine is all paved by now. Good memories.
My wife's a retired Army Nurse. We go to Carson's Commissary because going on an Army base is more familiar than an Air Force (or Space Force) base... and to be honest, the Carson Commissary's laid out better and has more stock than Peterson SFB's does (the Academy Commissary is a cute little place, perfect for the Academy).
The road up Pikes Peak is, indeed, paved, but still a slow drive because of the twists and turns. One of my goals in retirement is to start doing 14ers like the locals, but I've got a little bit more physical training to do first!