The Joint Typhoon Warning Center (JTWC) at Pearl Harbor, Hawaii, provides tropical cyclone (TC) monitoring, warning, and forecast support for the U.S. Department of Defense and other U.S. government agencies across the Pacific and Indian Ocean basins—from the east coast of Africa to the west coast of the Americas. Due to the lack of routine airborne reconnaissance and overall data-sparse nature of this geographically large area of responsibility (AOR), JTWC relies heavily on space-based environmental monitoring (SBEM) methods to provide the requisite data to characterize TC position, intensity, and structure analysis (wind radii, radius of maximum winds, etc.). During the 2019 TC year official poststorm best track (BT) review, JTWC began incorporating new SBEM methods and data into their analysis practice. This summary attempts to communicate the impact of incorporating these new methods into practice through the 2019 JTWC Best Track intensity reanalysis and subsequent real-time storm analysis by presenting comparisons of SBEM-derived intensity estimates to the JTWC 2019 TC BT data. Throughout this paper, we use knots as the unit of measure for wind speed (kt; 1 kt = 0.514 m s−1), as this is the unit utilized operationally at JTWC and refers to 1-min average winds.
JTWC relies heavily on the Dvorak technique as its primary method to estimate tropical cyclone intensity due to the widespread availability and high temporal resolution of visible and infrared geostationary sensors (Dvorak 1975). Satellite analysts at JTWC issue Dvorak-based current intensity “fixes,” derived from the subjective Dvorak technique, under the WMO agency identifier PGTW every 3 h. These fixes contain position and intensity estimates, with current intensity (CI) estimates corresponding to maximum wind speed as detailed in Dvorak (1975). Despite its recognized biases and challenges, studies have shown that Dvorak CI estimates correlate well with observed maximum 10-m sea surface winds (SSW) (Kossin and Velden 2004; Olander and Velden 2007; Knaff et al. 2010). Complementing these subjective Dvorak fixes, the National Environmental Satellite, Data, and Information Service (NESDIS) and Cooperative Institute for Meteorological Satellite Studies (CIMSS) at the University of Wisconsin, provide objective intensity, position, and wind radii estimates via the advanced Dvorak technique (ADT) algorithm, which also processes IR and visible satellite imagery (Olander and Velden 2019).
JTWC also complements Dvorak-based position and intensity estimates with data from microwave-frequency polar-orbiting sensors. Despite arriving late in many instances, microwave data can be used to establish (when timely) or validate (when late) BT position, intensity, and wind radii. These data aid forecasters in establishing confidence in storm position and intensity both during analysis and poststorm review, especially when the spread of available subjective and automated Dvorak-based methods is high. When available, JTWC forecasters use data from microwave sensors to develop storm positions via detection of low-level banding features and regions of heavier precipitation that are often obscured in infrared and visible imagery. Forecasters also use microwave derived SSWs based on ocean brightness temperature differences where available (Hihara et al. 2015). Furthermore, algorithms that exploit the polarization of backscattered microwave radiation, such as that from the ASCAT sensors on board the MetOp-B and MetOp-C satellites, enable estimation of SSWs when rain attenuation is minimal (Portabella et al. 2012). SMOS and SMAP radiometer data are largely unaffected by rain due to their operation in the L-band and therefore complement scatterometry data at higher SSW ranges (Knaff et al. 2021; Reul et al. 2017; Meissner et al. 2016). Additionally, the high spatial resolution of SAR sensors such as Sentinel-1A/B and RadarSat-2 provide the most complete picture of storm intensity and structure. Recent validation studies reveal that SAR-derived winds (SAR sends and receives polarized signals in both horizontal and vertical polarizations, enabling high wind speed estimates) are accurate at much higher SSWs than data from other radiometers and scatterometers (Mouche et al. 2019; Li et al. 2013; Katsaros et al. 2000; Combot et al. 2020). These new capabilities for estimating SSW are described in detail in Knaff et al. (2021). In addition to these microwave data, CIMSS combines their ADT intensity estimates with data from several microwave sounders to calculate “satellite consensus intensity estimates” (SATCON) (Velden and Herndon 2020).
During the poststorm review process for the 2019 season, the Naval Research Laboratory (NRL) and Cooperative Institute for Research in the Atmosphere (CIRA) at Colorado State University provided JTWC access to archived SMAP, SMOS, AMSR2, and SAR derived SSW data for TCs. Because these SSW data were not available to JTWC in real time during 2019, they functioned as an independent data source for the 2019 poststorm reviews. Poststorm reviewers adjusted the BT intensity based on these data in numerous cases and analysis (this paper) found that adjustments were small (±5 to 10 kt) in the majority of cases. The following sections detail and attempt to communicate to the TC community how the SSW data from the SMAP, SMOS, AMSR, and SAR resulted in these adjustments, and the extent these adjustments impacted the final BT.
Data
Of the available data, SAR collections were the least frequent during the 2019 season (38) due to satellite tasking requirements. SMAP and SMOS together had the most frequent TC overpasses (201), while AMSR2 overpasses fell between with 75 passes. SMAP, SMOS, and AMSR2 sampled TCs across the intensity spectrum due to consistent collection along their orbital paths and SAR retrievals captured higher intensity systems in general due to the need for scheduled tasking, which was often coordinated to focus on higher intensity TC’s.
Discussion
During the 2019 storm review, poststorm intensity estimates derived from SAR, SMAP, SMOS, and AMSR2 were utilized alongside agency Dvorak and CIMSS objective ADT and SATCON intensity estimates to refine the final BT intensity. Because these late data were unavailable to the forecaster during real-time analysis, their effect on the entire 2019 storm BT dataset may be analyzed independently. Only satellite data falling within ±3 h of the BT time are chosen for comparison. There are a few sources of potential uncertainty and biases in both the satellite-based intensity estimates, and comparisons to BT data. First, the satellite-based intensity estimates are based on the highest representative SSW evident in any given image. For SAR and AMSR2, localized patches of anomalously high SSW with unrealistic spatial gradients are dismissed, as they are considered suspect and often an indication of rain-induced biases that can occur in these sensor measurements (Hihara et al. 2015; Combot et al. 2020). SAR methodologies also may result in higher intensities due to the high resolution of the sensor resolving small TC eyes, which are known to cause negative biases in Dvorak intensity estimates (Knaff et al. 2010). Furthermore, occasional rain-induced biases can occur in SAR data about 14% of the time (Combot et al. 2020), potentially introducing artifacts in the SSW estimates. AMSR2 is also known to have both positive and negative TC SSW biases depending on environmental conditions such as precipitation rate. For example, in areas of high precipitation there is a known negative bias due to scattering and/or absorption of the outbound radiation (Meissner et al. 2021). It is also possible that low-resolution sensors (e.g., SMAP and SMOS) may underestimate maximum SSW, particularly for small TC’s, a source of potential low bias in the estimates in some cases. Ultimately, derived intensity estimates are subjective and based on a qualified forecaster’s assessment of the data’s reliability and consistency. Last, the comparison of the SBEM intensity estimates to the BT estimates at the nearest synoptic time adds a small degree of possible intensity error due to time mismatches of up to 3 h. However, given the temporal smoothing present in best track analyses (Landsea and Franklin 2012), this is not expected to significantly impact these results.
Despite these challenges, SMAP, SMOS, SAR, and AMSR2-based intensity estimates correlate with the operational real-time BT intensities. The correlation coefficients between each dataset and the original best track intensities are determined to be statistically significant using the linear least squares method. The AMSR2 intensity estimates provide the lowest correlations, with an R2 value of 0.5328. This spread likely results from the removal of large areas of “rain flagged” data during AMSR2 post processing, which is common due to the high precipitation rates associated with TCs. Due to this low correlation, AMSR2 data were rarely considered in poststorm analysis unless another source of microwave data were also available (<30% of AMSR2 passes). In only 1.6% of the data were AMSR2 used without another microwave intensity estimate (not including SATCON) to refine the BT intensity. Correlation was highest between SMAP intensity estimates and the original best track intensity and CIMSS SATCON CI values, with R2 = 0.8124 and R2 = 0.8125, respectively. Additional comparison of the datasets is complicated by differing resolutions, frequencies, and sample sizes among the sensors. Figure 1 shows the relationship between SAR, SMAP, SMOS, and AMSR2 sensor intensity estimates and the unadjusted 2019 BT data.
Scatterplot comparing the original JTWC best track intensity to the intensity derived from SMAP/SMOS (blue), SAR (red), and AMSR2 (gold) sensors. Linear best fit (solid lines) and R2 linear least squares correlation coefficient values (dashed lines) are also included.
Citation: Bulletin of the American Meteorological Society 103, 10; 10.1175/BAMS-D-21-0180.1
Regression analysis also indicates the SAR, SMAP, SMOS, and AMSR2 derived intensity estimates correlate well with JTWC Dvorak, CIMSS ADT, and CIMSS SATCON intensity estimates. The correlation values between these data are listed in Table 1. This is not a particularly surprising result as agency and automated Dvorak intensity estimates are the primary factor in BT intensity analysis, especially when storms exceed the acceptable scatterometry limit of 50 kt (Knaff et al. 2021). Note that the relatively low precision of Dvorak fixes introduces an additional element of uncertainty to this correlation analysis. For example, T4.0 = 65 kt, while T4.5 = 77 kt. This allows for a range of possible corresponding intensities between the T numbers and likely decreases the overall correlation (Olander and Velden 2007). The correlation overall suggests that these data are useful for making small refinements to the BT intensity, especially in a broad, data-sparse AOR such as the one JTWC operates in.
R2 values regressing intensity estimates from SMAP/SMOS, SAR, and AMSR2 onto the temporally closest JTWC initial (unadjusted) best track intensity, JTWC Dvorak CI, CIMSS ADT CI, and CIMSS SATCON.
These microwave data were used to adjust final BT intensities in the 2019 storm dataset. It is important to note that these intensity adjustments were generally small and fell within the measurement precision limits of the Dvorak technique over 90% of the time. In the few cases where these data were used to adjust the final intensity beyond these limits the adjustments were corroborated by multiple data points. These microwave intensity data aided forecaster confidence in BT intensity by providing additional intensity estimates independent of other remote sensing techniques and at times provided an additional, earlier source of intensity data, prior to Dvorak-based estimates being available.
For operational BT intensities of 80 kt or above, inspection of the data led to increases in intensity more frequently than decreases. When used, analysis of SAR and AMSR2 data resulted in intensity increases 50% of the time (see Table 2) for storms of this intensity or greater. Similarly, for systems 75 kt or below, SAR, SMAP, and SMOS data prompted more increases than decreases in intensity, while AMSR2 data influenced a nearly equal number of increases and decreases. These differences are unlikely to be due to the aforementioned inherent sensor biases. For example, AMSR2 precipitation-based biases, as discussed earlier, would not result in consistent increases in intensity and as stated previously, this is a primary driver behind the low use rate of AMSR2 data overall to make intensity adjustments, especially when other time-coincident microwave data are unavailable. SMAP and SMOS, on the other hand, are known to be less impacted by heavy precipitation and do not have a significant bias (Reul et al. 2017; Meissner et al. 2016).
Percentage of cases in which a reviewer adjusted the JTWC initial intensity upward or downward based on subjective evaluation of SAR, SMAP/SMOS, or AMSR2 imagery during poststorm analysis. Upward adjustments to the final intensity are highlighted in italics, while downward adjustments are highlighted in boldface.
By early April 2020, efforts by NRL-Monterey and CIRA to decrease SAR, SMAP, SMOS, and AMSR2 data latency enabled the JTWC forecaster to use these data operationally and influence the intensity estimation process at key junctures during real-time storm analysis. For example, for Typhoon 07W (Mekkhala), the JTWC forecaster set the initial real-time BT intensity at 55 kt at 1800 UTC 10 Aug 2020. The JTWC Dvorak fix based on IR imagery (not shown) was T3.0 (45 kt), ADT T3.2 (49 kt), and SATCON 47 kt. Shortly after this initial analysis was completed, a 1743 UTC AMSR2 SSW estimate (Fig. 2b) became available through operational data outlets. After assessing the AMSR2 37-GHz microwave image (Fig. 2a) image for artifacts due to precipitation, the forecaster noted the higher intensity which prompted further analysis of additional data sources including buoy data and a final adjustment of the intensity to 65 kt just prior to landfall, 27 nm from Xiamen, China. This intensity was consistent with CMA’s official landfall intensity of 69 kt (10-min average) (which may have incorporated data that JTWC does not have access to) and a buoy observation (number 59334) of 64 kt in proximity to the storm. The forecast intensity would have been 10–20 kt lower if the forecaster only considered the infrared-based Dvorak estimates.
(a) AMSR2 37-GHz (vertical polarization) microwave image for TY 07W (Mekkhala) from 1743 UTC 10 Aug 2020 showing brightness temperature, which is correlated with precipitation and (b) SSW estimate for the same from ATCF at 1743 UTC 10 Aug 2020.
Citation: Bulletin of the American Meteorological Society 103, 10; 10.1175/BAMS-D-21-0180.1
Tropical Cyclone 25P (Harold) provides an example of an intensity increase adjustment based on SAR data. On 0600 UTC 6 April 2020, the initial BT intensity was analyzed at 135 kt based on a PGTW Dvorak estimate of T7.0 (140-kt equivalent) and the near-concurrent ADT estimate of T6.5 (127 kt). Shortly thereafter, by 0940 UTC, ADT estimates increased to T7.4 (152 kt). Furthermore, data from a 0714 UTC SAR pass (Fig. 3a) depicted SSWs near 160 kt in the northeast quadrant of the system, with some higher values approaching 193 kt along the immediate coastline of Pentacost Island. The higher values approaching 195 kt were considered to be spurious and not representative of the TC intensity. The BT intensity at 0600 UTC 6 April was adjusted upward to 145 kt, partially to account for higher winds evident in the SAR data, which provided additional data to clarify that the BT intensity was on the higher end of the available (Dvorak and ADT) intensity estimates. This example demonstrates the utility of SAR imagery as an additional data point for very intense systems.
(a) Graphic of NESDIS STAR TC winds generated SAR radial winds averaged over all quadrants for 0714 UTC 6 Apr 2020. (b) SAR overview imagery from 0714 UTC 6 Apr 2020 as viewed in ATCF with the numerical SSW values for each pixel superimposed, with areas of 160 kt or higher winds highlighted.
Citation: Bulletin of the American Meteorological Society 103, 10; 10.1175/BAMS-D-21-0180.1
Summary
JTWC is now utilizing SAR, SMAP/SMOS, and AMSR2 derived SSW as additional data sources for real-time and poststorm analysis. Previously, only sparse observations from ships, island weather stations, scatterometry sensors (50 kt and below), and application of the Dvorak technique were available for estimating TC intensity. The aforementioned SBEM SSW derived intensity estimation methods now benefit the JTWC poststorm best track review and as of 2020, the real-time storm analysis. These data are useful as they were shown (this study) to correlate with 2019 JTWC best track intensities and Dvorak CI estimates. Importantly, over half the time in the 2019 BT review, these data resulted in small increases of the final intensities (<10 kt), though in a few cases these SBEM estimates provided support to make larger changes (>15 kt), particularly for intense (> 80 kt) TCs which lacked data other than traditional Dvorak-based fixes. The value of these new satellite-based intensity estimates to an operational TC forecasting center cannot be overstated, as they provide an outstanding source of measured SSW data in otherwise data-sparse regions. While these sensors contributed greatly to JTWC’s operational and poststorm assessment of TC intensity, and changes were made to the poststorm TC intensity records, the overall impact to the historical record is small, with 411 of the 425 postanalysis adjustments being 10 kt or smaller, well within accepted uncertainty ranges for best track intensity estimates (e.g., Landsea and Franklin 2012). Overall, these methods lend valuable SSW and thus storm intensity estimations that are now trusted by JTWC to enhance storm intensity analysis, both in real time and during the poststorm review process. Concerns about heterogeneity in the TC record are certainly valid from a research point of view, since the introduction of new datasets can alter the statistical characteristics of that record. However, it is worth pointing out that such heterogeneity is not new, and is present throughout the historical record due to the appearance (and disappearance) of new satellite sensors, aircraft reconnaissance, etc., and therefore, researchers will always have to be cautious when incorporating data based on new sensors.
Acknowledgments.
This study was made possible by input from Buck Sampson and Mindy Surratt at NRL, John Knaff, Chris Jackson, and Tyler Ruff at NOAA Satellite Applications and Research (STAR), and the teams who support the JTWC mission at NRL-Monterey, CIRA, and the NOAA STAR. Publication funding was provided by the U.S. Air Force. We thank Dr. Levi Cowan (JTWC), Brian Strahl (JTWC), Matthew Kucas (JTWC), and Owen Shieh (JTWC) for their helpful comments and suggestions that improved this manuscript.
References
Combot, C. , A. Mouche , J. A. Knaff , Y. Zhao , Y. Zhao , L. Vinour , Y. Quilfen , and B. Chapron , 2020: Extensive high-resolution synthetic aperture radar (SAR) data analysis of tropical cyclones: Comparisons with SFMR flights and best track. Mon. Wea. Rev., 148, 4545–4563, https://doi.org/10.1175/MWR-D-20-0005.1.
Dvorak, V. F. , 1975: Tropical cyclone intensity analysis and forecasting from satellite imagery. Mon. Wea. Rev., 103, 420–430, https://doi.org/10.1175/1520-0493(1975)103<0420:TCIAAF>2.0.CO;2.
Hihara, T. , M. Kubota , and A. Okuro , 2015: Evaluation of sea surface temperature and SSW observed by GCOM-W1/AMSR2 using in situ data and global products. Remote Sens. Environ., 164, 170–178, https://doi.org/10.1016/j.rse.2015.04.005.
Katsaros, K. , P. W. Vachon , P. G. Black , P. P. Dodge , and E. W. Uhlhorn , 2000: Wind fields from SAR: Could they improve our understanding of storm dynamics? Johns Hopkins APL Tech. Dig., 21, 86–93.
Knaff, J. A. , D. P. Brown , J. Courtney , G. M. Gallina , and J. L. Beven , 2010: An evaluation of Dvorak technique–based tropical cyclone intensity estimates. Wea. Forecasting, 25, 1362–1379, https://doi.org/10.1175/2010WAF2222375.1.
Knaff, J. A. , and Coauthors, 2021: Estimating tropical cyclone surface winds: Current status, emerging technologies, historical evolution, and a look to the future. Trop. Cyclone Res. Rev., 10, 125–150, https://doi.org/10.1016/j.tcrr.2021.09.002.
Kossin, J. P. , and C. S. Velden , 2004: A pronounced bias in tropical cyclone minimum sea level pressure estimation based on the Dvorak technique. Mon. Wea. Rev., 132, 165–173, https://doi.org/10.1175/1520-0493(2004)132<0165:APBITC>2.0.CO;2.
Landsea, C. W. , and J. L. Franklin , 2012: Atlantic Hurricane Database uncertainty and presentation of a new database format. Mon. Wea. Rev., 141, 3576–3592, https://doi.org/10.1175/MWR-D-12-00254.1.
Li, X. , J. A. Zhang , X. Yang , W. G. Pichel , M. Demaria , D. Long , and Z. Li , 2013: Tropical cyclone morphology from spaceborne synthetic aperture radar. Bull. Amer. Meteor. Soc., 94, 215–230, https://doi.org/10.1175/BAMS-D-11-00211.1.
Meissner, T. , F. J. Wentz , J. Scott , and J. Vazquez-Cuervo , 2016: Sensitivity of ocean surface salinity measurements from spaceborne L-band radiometers to ancillary sea surface temperature. IEEE Trans. Geosci. Remote Sens., 54, 7105–7111, https://doi.org/10.1109/TGRS.2016.2596100.
Meissner, T. , L. Ricciardulli , and A. Manaster , 2021: Tropical cyclone wind speeds from WindSat, AMSR and SMAP: Algorithm development and testing. IEEE Trans. Geosci. Remote Sens., 54, 7105–7111, https://doi.org/10.1109/TGRS.2016.2596100.
Mouche, A. , B. Chapron , J. Knaff , Y. Zhao , B. Zhang , and C. Combot , 2019: Copolarized and cross-polarized SAR measurements for high-resolution description of major hurricane wind structures: Application to Irma category 5 hurricane. J. Geophys. Res. Oceans, 124, 3905–3922, https://doi.org/10.1029/2019JC015056.
Olander, T. L. , and C. S. Velden , 2007: The advanced Dvorak technique: Continued development of an objective scheme to estimate tropical cyclone intensity using geostationary infrared satellite imagery. Wea. Forecasting, 22, 287–298, https://doi.org/10.1175/WAF975.1.
Olander, T. L. , and C. S. Velden , 2019: The advanced Dvorak technique (ADT) for estimating tropical cyclone intensity: Update and new capabilities. Wea. Forecasting, 34, 905–922, https://doi.org/10.1175/WAF-D-19-0007.1.
Portabella, M. , A. Stoffelen , A. Verhoef , and J. Verspeek , 2012: A new method for improving scatterometer wind quality control. IEEE Geosci. Remote Sens. Lett., 9, 579–583, https://doi.org/10.1109/LGRS.2011.2175435.
Reul, N. , and Coauthors, 2017: A new generation of tropical cyclone size measurements from space. Bull. Amer. Meteor. Soc., 98, 2367–2385, https://doi.org/10.1175/BAMS-D-15-00291.1.
Velden, C. S. , and D. Herndon , 2020: A consensus approach for estimating tropical cyclone intensity from meteorological satellites: SATCON. Wea. Forecasting, 35, 1645–1662, https://doi.org/10.1175/WAF-D-20-0015.1.
