東華大學圖書館 |

Development of a Model Climatology Tool to Assess Ensemble Variability and Uncertainty Using East Coast Cyclones.

Record Type:	Electronic resources : Monograph/item
Title/Author:	Development of a Model Climatology Tool to Assess Ensemble Variability and Uncertainty Using East Coast Cyclones./
Author:	Mandelbaum, Taylor.
Published:	Ann Arbor : ProQuest Dissertations & Theses, : 2018,
Description:	107 p.
Notes:	Source: Masters Abstracts International, Volume: 80-06.
Contained By:	Masters Abstracts International80-06.
Subject:	Statistical physics. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10846877
ISBN:	9780438694019

Development of a Model Climatology Tool to Assess Ensemble Variability and Uncertainty Using East Coast Cyclones.
Mandelbaum, Taylor.

Development of a Model Climatology Tool to Assess Ensemble Variability and Uncertainty Using East Coast Cyclones. - Ann Arbor : ProQuest Dissertations & Theses, 2018 - 107 p.

Source: Masters Abstracts International, Volume: 80-06.

Thesis (M.S.)--State University of New York at Stony Brook, 2018.

This item must not be sold to any third party vendors.

Probabilistic forecasting is an important tool for both public and private sectors. The use of ensemble models increases the awareness of uncertainty and errors in the output of a model. A common tool to diagnose uncertainty in an ensemble model is ensemble spread, or standard deviation. Ensemble spread is most commonly viewed in a geographic reference frame with implications of the most likely areas of uncertainty, and by extension, errors. Although the use of ensembles has increased, there exist many opportunities for better visualization of ensemble model output, which is a major objective of the Stony Brook University CSTAR project. The Ensemble Situational Awareness Table (ESAT), managed by the National Weather Service and Weather Prediction Center, compares forecasts from the North American Ensemble Forecast System (NAEFS) and Global Ensemble Forecast System (GEFS) to reanalysis (R-Climate) and model reforecast (M-Climate) climatologies. While M-Climate output from the GEFS Reforecast can place the current ensemble mean forecast in context, it does not yet assess the ensemble spread relative to similarly anomalous events. We attempt to take the M-Climate diagnostic a step further by proposing a new index for spread, using a restricted climatology based on ensemble mean anomaly. We also look at the relationship between ensemble spread and the error for East Coast cyclones with the intent of providing context for the tool as well as the performance of EPS's ability to "predict" their own error. Our goal is to output and justify an operational spread anomaly product that will complement the existing ESAT. To justify the tool, an analysis of 90 verified East Coast cyclones from 2007-2014 was performed for the GEFS, CMC, and ECMWF ensembles. Cases restricted to the winter (DJF) timeframe over the contiguous United States were chosen for this project. Mid-latitude synoptic cyclones are the most prevalent high impact events, especially for southern New York and New England. Initial test variables include mean sea-level pressure (SLP), surface temperature, and precipitable water, but SLP will be the focus of this study. By analyzing the ensemble spread for all members of the GEFS, a geographic and forecast cyclone-relative perspective of the magnitude and co-location of the spread and error can be assessed. Spread and mean absolute error (MAE) expand at different rates and are consistently different magnitudes, with maximum values of the GEFS MAE 270% greater than the spread for days 1-3 (short range) and 188% greater for days 5-7 (long range). RMSE values for the GEFS are similar. A method is developed which identifies the centroids of spread and MAE for the upper quartile of cyclone-relative data. The ECMWF does significantly better in both short and long range when looking at the distance between centroids, as well as the percentage of points on the cyclone-relative grid which overlap. The GEFS tends to perform the worst in both metrics. 500 hPa geopotential height patterns are also compared between the best and worst performing cyclone cases with respect to centroid distance and percent overlap where significantly different wave patterns are identified. GEFS Reforecast cyclone forecasts are compared to GEFS operational forecasts to assess the similarity of products, where it is found that there are statistically significant differences in spread and MAE for most lead times. GEFS Reforecast cases at each point within a threshold of comparable standardized anomalies (±5%) to the forecast are utilized to construct a new spread M-Climate. Using this method, a subset standardized spread anomaly (SSA) can be calculated for each point on the domain. The resulting plot, together with the ensemble mean pressure contours of that forecast, displays a filtered, historical reference of uncertainty in the GEFS forecast. The SSA provides context for each lead time while identifying areas of anomalous spread, which shows a correlation with standardized error for short and long ranges with r values ~0.54 and ~0.64, respectively for forecast cyclone relative domain-averaged SSA and standardized MAE.

ISBN: 9780438694019Subjects--Topical Terms:

536281
Statistical physics.

Development of a Model Climatology Tool to Assess Ensemble Variability and Uncertainty Using East Coast Cyclones.
LDR:05289nmm a2200337 4500 001 2263841
005 20200331094341.5
008 220629s2018 ||||||||||||||||| ||eng d
020 $a 9780438694019
035 $a (MiAaPQ)AAI10846877
035 $a (MiAaPQ)grad.sunysb:13842
035 $a AAI10846877
040 $a MiAaPQ $c MiAaPQ
100 1 $a Mandelbaum, Taylor. $3 3540933
245 1 0 $a Development of a Model Climatology Tool to Assess Ensemble Variability and Uncertainty Using East Coast Cyclones.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2018
300 $a 107 p.
500 $a Source: Masters Abstracts International, Volume: 80-06.
500 $a Publisher info.: Dissertation/Thesis.
500 $a Advisor: Colle, Brian.
502 $a Thesis (M.S.)--State University of New York at Stony Brook, 2018.
506 $a This item must not be sold to any third party vendors.
520 $a Probabilistic forecasting is an important tool for both public and private sectors. The use of ensemble models increases the awareness of uncertainty and errors in the output of a model. A common tool to diagnose uncertainty in an ensemble model is ensemble spread, or standard deviation. Ensemble spread is most commonly viewed in a geographic reference frame with implications of the most likely areas of uncertainty, and by extension, errors. Although the use of ensembles has increased, there exist many opportunities for better visualization of ensemble model output, which is a major objective of the Stony Brook University CSTAR project. The Ensemble Situational Awareness Table (ESAT), managed by the National Weather Service and Weather Prediction Center, compares forecasts from the North American Ensemble Forecast System (NAEFS) and Global Ensemble Forecast System (GEFS) to reanalysis (R-Climate) and model reforecast (M-Climate) climatologies. While M-Climate output from the GEFS Reforecast can place the current ensemble mean forecast in context, it does not yet assess the ensemble spread relative to similarly anomalous events. We attempt to take the M-Climate diagnostic a step further by proposing a new index for spread, using a restricted climatology based on ensemble mean anomaly. We also look at the relationship between ensemble spread and the error for East Coast cyclones with the intent of providing context for the tool as well as the performance of EPS's ability to "predict" their own error. Our goal is to output and justify an operational spread anomaly product that will complement the existing ESAT. To justify the tool, an analysis of 90 verified East Coast cyclones from 2007-2014 was performed for the GEFS, CMC, and ECMWF ensembles. Cases restricted to the winter (DJF) timeframe over the contiguous United States were chosen for this project. Mid-latitude synoptic cyclones are the most prevalent high impact events, especially for southern New York and New England. Initial test variables include mean sea-level pressure (SLP), surface temperature, and precipitable water, but SLP will be the focus of this study. By analyzing the ensemble spread for all members of the GEFS, a geographic and forecast cyclone-relative perspective of the magnitude and co-location of the spread and error can be assessed. Spread and mean absolute error (MAE) expand at different rates and are consistently different magnitudes, with maximum values of the GEFS MAE 270% greater than the spread for days 1-3 (short range) and 188% greater for days 5-7 (long range). RMSE values for the GEFS are similar. A method is developed which identifies the centroids of spread and MAE for the upper quartile of cyclone-relative data. The ECMWF does significantly better in both short and long range when looking at the distance between centroids, as well as the percentage of points on the cyclone-relative grid which overlap. The GEFS tends to perform the worst in both metrics. 500 hPa geopotential height patterns are also compared between the best and worst performing cyclone cases with respect to centroid distance and percent overlap where significantly different wave patterns are identified. GEFS Reforecast cyclone forecasts are compared to GEFS operational forecasts to assess the similarity of products, where it is found that there are statistically significant differences in spread and MAE for most lead times. GEFS Reforecast cases at each point within a threshold of comparable standardized anomalies (±5%) to the forecast are utilized to construct a new spread M-Climate. Using this method, a subset standardized spread anomaly (SSA) can be calculated for each point on the domain. The resulting plot, together with the ensemble mean pressure contours of that forecast, displays a filtered, historical reference of uncertainty in the GEFS forecast. The SSA provides context for each lead time while identifying areas of anomalous spread, which shows a correlation with standardized error for short and long ranges with r values ~0.54 and ~0.64, respectively for forecast cyclone relative domain-averaged SSA and standardized MAE.
590 $a School code: 0771.
650 4 $a Statistical physics. $3 536281
650 4 $a Statistics. $3 517247
650 4 $a Atmospheric sciences. $3 3168354
690 $a 0217
690 $a 0463
690 $a 0725
710 2 $a State University of New York at Stony Brook. $b Marine and Atmospheric Science. $3 1683777
773 0 $t Masters Abstracts International $g 80-06.
790 $a 0771
791 $a M.S.
792 $a 2018
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10846877