Loading...
Please wait, while we are loading the content...
Similar Documents
J 4 . 3 an Information-theoretic Approach to Quantifying the Uncertainty in Operational Tropical Cyclone Intensity Predictions , with Application to Forecast Verification
| Content Provider | Semantic Scholar |
|---|---|
| Author | Moskaitis, Jonathan R. |
| Copyright Year | 2008 |
| Abstract | Quantification of the uncertainty inherent in predictions of tropical cyclone (TC) intensity is not only of scientific interest, but also is of relevance to users who must involve TC intensity forecasts in decisionmaking processes. Here, we describe an approach to quantify the uncertainty in (deterministic) operational TC intensity forecasts, based solely on a set of such forecasts and the corresponding set of observed intensity values. The aforementioned data sample is used to estimate an unconditional probability distribution of the observations, as well as a set of conditional probability distributions of the observations given the forecast. Qualitatively, if these conditional distributions are sharper than the unconditional distribution, then knowledge of the forecast value serves to reduce uncertainty about the value of the observation, relative to knowledge of the unconditional distribution alone. The average reduction in uncertainty about the observation due to knowledge of the forecast is quantified via calculation of the mutual information between the forecasts and observations, a concept borrowed from information theory. Further details concerning mutual information and its interpretation are contained in Sec. 2. The data samples used in calculating the mutual information between various operational TC intensity forecasts and the observations are described in Sec. 3, followed by a demonstration of the calculation process for a particular data sample in Sec. 4. Sec. 5 then shows the results. It is seen that the mutual information between the various operational TC intensity forecasts and the observations is positive for all lead times (0 to 5 days), meaning that even the longer lead forecasts reduce uncertainty about the value of the observation relative to that inherent in the unconditional distribution of the |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://ams.confex.com/ams/pdfpapers/133164.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
