Thursday, July 06, 2006

Could NWS use high-tech observations more effectively?

Art Douglas has posted several comments - in one of these he makes what seems a perfectly reasonable suggestion that the NWS consider using areal observations from high-tech observing systems that are routinely used by forecasters and researchers. He suggests considering radar-estimated rainfall and CG lightning strike data to determine when the monsoon has arrived where. The advantages over using a single point observation seem obvious. This, to me, seems a reasonable suggestion, but it also raises a very important question and issues related to how the NWS uses observations.

Consider that:

The new WSR-88D doppler radar system has been operating nationwide for more than a decade. The radar processing systems routinely produce maps of radar-estimated precipitation accumulations.

The CG lightning strike detection system has been operating nationwide for about two decades.

The satellite observing systems have been in place, and continuously improving, since the early 1970s. Satellite rain estimation techniques have been used by researchers and the NWS for more than two and a half decades.

It seems clear that the NWS could draw upon observations from these systems to define better the onset of the monsoon.

But, the larger question is: Why doesn't the NWS use data and products from any of these systems (systems whose cost to the nation has been at least in the tens of billions of dollars) to verify, assess, or crtically evaluate their forecast products?

For example, the SPC could routinely use CG lightning strike data to evaluate the reliability of their thunderstorm outlooks.

Satellite and radar products could be used to evaluate what actually happens relative to point weather observations and etc. etc.

If anyone knows more on these issues, please share your information with us.

Note - now that I've thought about it a bit more, it seems that the NHC might make more quantitative use of the satellite data in assessing their forecasts than other components of the NWS.


  1. From a forecaster's point of view the main drawback to using these systems for critical evaluation of forecast products is more related to the evaluation and verification paradigms. These paradigms are tied directly to point observations of parameters while the radar, satellite and lightning detection network output is more based on pattern recognition.

    Certainly *some* parameters can be quantified to some degree (WSR-88D precipitation accumulations, for example) but at least to this forecaster the real value of these tools still rests in the ability to recognize certain patterns in the images (data) presented.

    This use of the data argues for probablistic measures of forecasts, which, in turn, requires that the forecasts themselves be expressed more in probabilistic terms than in specifics. This is a huge dicotamy in that there appears to be no desire with NWS leadership to move towards probablistic forecasts.

  2. Chuck Doswell2:33 PM

    Not being a blogger, I put this comment in the wrong place ... I sent:

    Thought I'd offer a comment regarding new observations and the monsoon. I certainly think that more intelligent uses of new observational technology could be developed. But ... using them to "define" the monsoon for the purposes of saying on which day the monsooon has "officially" begun seems rather like using lasers and nanotechnology to calculate how many angels are dancing on the head of a pin. The atmosphere surely doesn't know anything about how we humans define a monsoon, so it's hard for me to see much value in this, however it might be done. I find it sadly amusing to think that someone actually cares on what date the NWS decides the monsoon has started. This might be of some pertinence for an office pool about the onset of the monsoon, but it seems pretty pointless for anyone but trivia buffs. Of course, one can define precisely what a monsoon is for some scientific study, but this is virtually certain to lead to something comparable to the usual arguments, both pointless and endless, about how we decide what is El Nino vs. La Nina vs. "neutral". It seems to me that forecasting the weather is substantially more important than defining the specific date on which we say the monsoon has begun/ended.

  3. Steve Burk4:55 PM

    I think the public’s interest is not so much geared to wanting an exact starting date for the summer monsoon. Perhaps the most pressing question is something like this: “Has the local summer rainy season started in earnest or are we likely to return to pre-monsoon conditions for an extended period?” There can be serious reasons behind such a question (besides office pools or angels dancing) such as farmers concerned with their crops, those in construction worried about thunderstorms, etc. For instance, the nearly 4 million people in the greater Phoenix area probably take little comfort in being told that the summer monsoon rains began in early June based on lightning strikes in the Huachuca and Chiricahua mountains --- especially since up until today about all they have seen is an occasional dust storm. Five day mesoscale forecasts are good, of course, but they may not answer the question above. Although meteorologists perhaps would prefer to address some other question, if this is what is of importance to the public (e.g., has the rainy season really started in the Phoenix area or not?), then it seems to me that it is a legitimate question worth wrestling with.

  4. I don’t want to distract too much from the monsoon theme, and your point about smarter usage of more sophisticated datasets remains valid, but I do want to note that the SPC does make extensive use of CG lightning strike data in several applications including statistical predictive guidance and the outlook verification program. As an example, you can view daily verification of the enhanced thunderstorm outlook at the following URL.

    [If you're not familiar with the enhanced thunderstorm outlook, you can find that experimental forecast here. ]

    There is also ongoing development incorporating radar diagnostics in both guidance and verification products, although admittedly, the CG strike data are currently used more extensively than the radar algorithm output.