| Do you think better weather forecasting requires | | | | the statistics. Check the forecast highs and lows |
| a degree in meteorology? Maybe a degree in | | | | for the last 365 days, and check the actual |
| statistical analysis would be more helpful. Consider | | | | temperatures for those days. See what the |
| this following new way to forecast the weather | | | | predicted probabilities of rain or snow were, and |
| with more accuracy and less knowledge. | | | | what actually happened. |
| Friday, February 2, 2007 - Canon City, Colorado. I | | | | Suppose that of the last 24 times a forecaster |
| brought in my Canon City Daily Record from the | | | | predicted a 50% chance of rain, it actually rained |
| porch when it arrived, at about 3 in the afternoon. | | | | 18 times. He may have the best data, but he |
| I opened the newspaper to the page with the | | | | may be too conservative in how he uses it. |
| weather forecast, wondering how cold it would be | | | | Suppose this was not a fluke - which can be |
| the following day. | | | | determined by doing more statistical analysis. You |
| The paper forecast a high temperature was 13 | | | | could know nothing about weather forecasting |
| degrees Fahrenheit. I knew this was way too low. | | | | and provide a more accurate forecast simply by |
| Forecasts on television and on the internet said | | | | saying "A 75% chance of rain tomorrow" every |
| that we would reach 23 or 27 degrees the | | | | time he said there was a 50% chance, right? |
| following day. I knew they were also too low, and | | | | That's the basis for this new forecasting model. |
| I told my wife it would be in the 30s at least. The | | | | First gather the statistical information on the |
| actual high temperature the next day was 53 | | | | forecasts of several weather forecasting services |
| degrees Fahrenheit. | | | | or meteorologists. Compare this to the actual |
| No, that's not a typo. The various weather | | | | weather that happened, and look for any |
| forecasting "experts" were off by as much as 40 | | | | consistencies in the inaccuracies. Then you create |
| degrees - and that was for a simple 24-hour | | | | a computer program. As you enter each of these |
| forecast. How could they be so far off? And how | | | | forecasts into it, they are adjusted for known |
| could I be better than them at forecasting the | | | | tendencies. The result is a more accurate |
| weather? | | | | forecast. |
| I can't answer the first question. The weather | | | | If Forecaster A has managed over the last year |
| here is more unpredictable than in most places. | | | | to forecast a high that averages 4 degrees over |
| Also, perhaps the meteorologists follow there | | | | the actual high, the computer adjusts for that. |
| computer models too slavishly, even when their | | | | More sophisticated analysis might show that |
| experience and intuition tell them to adjust the | | | | Forecaster B is consistently predicting a higher |
| forecast. | | | | probability of rain than there is in the fall, but a |
| However, I can answer the second question. My | | | | lower probability of rain than there actually is in |
| guess was closer because the forecasters were | | | | the spring. The computer can take this into |
| so consistent in the way they made their errors. | | | | account. Finally, it may work best if the adjusted |
| Around this time, I remember counting something | | | | forecasts of three or more sources are then |
| like 15 out of 20 days when all the various | | | | averaged. |
| weather forecasts predicted a high temperature | | | | There really is no need to know anything at all |
| that was 5 degrees or more too low. All I needed | | | | about weather forecasting. This is based on the |
| to do was take the highest temperature forecast | | | | idea that even when experts have all the best |
| and add five degrees. | | | | knowledge and data, they sometimes apply it |
| A New Weather Forecasting Model | | | | incorrectly, and do so in consistent ways. Don't be |
| Consistency in their errors was the key to my | | | | surprised if some television stations get rid of |
| success. They weren't forecasting too high one | | | | their meteorologists and take advantage of this |
| day and too low the next. They were wrong in | | | | new weather forecasting idea. |
| the same ways over and over. | | | | "Now your electronic weather forecast, from our |
| Are the errors as consistent in other parts of the | | | | Statistical Analysis Weather Machine. |
| country? That could be determined by looking at | | | | |