PROJECT TITLE: Application of Joint Polar Satellite System (JPSS) Imagers and Sounders to Tropical Cyclone Track and Intensity Forecasting
PRINCIPAL INVESTIGATOR: Renate Brummer
RESEARCH TEAM: Galina Chirokova, Robert DeMaria, Jack Dostalek, Kate Musgrave, Andrea Schumacher, Hiro Gosden, Dave Watson, Kevin Micke, Renate Brummer, Kathy Fryer
NOAA TECHNICAL CONTACT: Ingrid Guch (NESDIS/NESDIS) and Phil Hoffman (NOAA/OAR)
NOAA RESEARCH TEAM: Mark DeMaria and John Knaff (NOAA/NESDIS/STAR)
The time scale of tropical cyclone track and intensity changes is on the order of 12 hours, which makes JPSS instruments well suited for the forecasting of these parameters. Two tropical cyclone applications of JPSS data are currently being developed. The first uses the imager and sounder data for improving center location estimates of tropical cyclones, which is the starting point for tropical cyclone forecasts. Methods are being developed to use multi-spectral imagery from VIIRS, in combination with sounder data for this purpose. The second is to use temperature and moisture retrievals from the Advanced Technology Microwave Sounder (ATMS) in the near storm environment to improve intensity analysis and forecasting. This new information is being incorporated into existing intensity estimation techniques and to an operational statistical-dynamical intensity forecast model to improve their performance. The goal is to make these new products available in the satellite Proving Ground to operational forecasters at the National Hurricane Center (NHC) and Joint Typhoon Warning Center (JTWC) for evaluation and feedback.
Tropical Cyclone (TC) forecasts affect risk mitigation activities of industry, public and governmental sectors and therefore supports directly NOAA’s Weather and Water mission goals. Improving forecasts of tropical cyclone track and intensity is a top NOAA/DoD priority.
PROJECT ACCOMPLISHMENTS: Past Fiscal Year by Objective:
1-- Develop database of AMSU MIRS temperature and moisture retrievals and VIIRS data for global tropical cyclones
A large database of VIIRS data is now available at CIRA. An automated procedure for extracting VIIRS and MIIRS data centered on global tropical cyclones has been in place since 2012, and continued to run in 2013 and 2014. The MIRS temperature and moisture retrievals became operational on the NPP Data Exploitation (NDE) system on Feb 6, 2014, however the data are not yet available via DDS feed. A large sample of cases was obtained from NESDIS/STAR using the same algorithm and in the same format as will be available from the NDE system. This sample includes 28 days for 23 TCs from 2012, providing about 200 TC cases from global tropical cyclones. This dataset is being used for the majority of the algorithm testing described below.
2-- Begin development of center fixing routine from AMSU, VIIRS, and GOES.
The center-fixing algorithm was developed and tested with proxy AMSU retrievals. The algorithm was adapted to use ATMS inputs, and a comparison with the AMSU version was made. Figure 1 shows the results of the comparison for the 206 cases from the ATMS MIRS dataset. The blue bars show the first guess center position errors (left bar) and those after the first guess is updated using the quadratic discriminant analysis technique (right bar) for the AMSU cases from 2006-2011 which was used to develop the algorithm. The red bars show the same two sets of errors for the case where the algorithm, trained on the 2006-2011 AMSU cases, was run on the 206 ATMS cases from 2012, and the green bars show the results where the algorithm was trained on the 2012 ATMS cases and run with the 2012 ATMS input. The red and green bars show that when the ATMS input is used, there is a greater improvement over the first guess than when the algorithm is run with AMSU input. The next step is to refine the ATMS center estimates with VIIRS data.
Figure 1. Errors in the center location estimate using the NHC best track positions interpolated to the time of the microwave pass. For each pair of bars (blue, red green), the left bar is the error of the first guess position and the second one is the error after the first guess has been updated using the quadratic discriminant analysis (QDA) technique. Results show that using the ATMS data provides a bigger improvement than the AMSU data, even for the case where the algorithm trained on AMSU data is used with ATMS input. The QDA will then be refined using the much higher resolution VIIRS data.
The utility of an image processing technique called the circular Hough transform algorithm for finding tropical cyclone centers from infrared imagery was evaluated. In the field of computer vision, the circular Hough transform is a commonly used algorithm for identifying circular or nearly circular features within images. An implementation of the algorithm was produced and run on 135 infrared images containing tropical cyclones at various points in their lifetimes. The distance between the center location reported by the algorithm and the best-track location was computed for each image. Additionally, the distance between the real-time location produced by the NHC, extrapolated to the time the IR image was created, and the best-track was used as a baseline comparison. The results can be seen in Figure 2. While the algorithm performed fairly well on images containing an eye, greater error was experienced when no eye was present. However in cases without an eye, the algorithm was able to effectively find the center of the cloud shield. Relating the center of the cloud shield to the rotational center of the storm may be an effective method of performing automated center-fixing and further work will be performed to investigate this possibility. For the weaker systems, additional information will be needed such as spiral cloud lines from visible and day-night band imagery, circulation centers and vertical wind shear estimates from ATMS wind retrievals.
Figure 2. Mean errors reported from using the circular Hough transform on IR images to perform automatic center fixing.
Share with your friends: