1. Introduction
There is a long history of initializing tropical cyclones (TCs) in the Met Office Unified Model (MetUM) in order to improve the model’s representation of TCs in both the analysis and the forecast. In the late 1980s and early 1990s forecasters had a tool available that inserted so-called bogus observations of central pressure, surrounded by four values of wind speed and direction at the surface and three lower-tropospheric levels. In 1994, this was superseded by a new initialization scheme that involved the insertion of bogus observations of wind speed and direction at the surface and three lower-tropospheric levels. This technique proved extremely successful and reduced TC track forecast errors by 34% on average in trials (Heming et al. 1995). The following year the MetUM produced better guidance for TC track prediction to the National Hurricane Center than any other numerical weather prediction model for the extremely active Atlantic hurricane season of that year (Gross 1996; Heming and Radford 1998).
In 2007, a complete reevaluation of the initialization scheme was undertaken to assess whether it was still proving beneficial to forecasts of TCs from the MetUM. In the years since the scheme was first introduced there had been many improvements in model formulation, increases in model resolution, and the introduction of new observational data, particularly from satellites. These changes were likely to have diminished the need for artificial initialization of TCs. In the event, this evaluation first found that the initialization scheme was still reducing TC track forecast errors by an average of 12.2%. Furthermore, a modification to the scheme that reduced the horizontal spread of “bogus” observations generated for small TCs resulted in a further reduction in TC track forecast errors of 4.7% (Heming 2009). This configuration of the initialization scheme will be known as the 2007 scheme hereafter. A diagrammatic representation of the scheme is shown in Fig. 1. After a few years in operation a further evaluation of the 2007 scheme was undertaken to ensure it was still providing benefit to the model forecasts. Section 2 of this paper presents the results of this evaluation.
The configuration of the 2007 scheme.
Citation: Weather and Forecasting 31, 5; 10.1175/WAF-D-16-0040.1
A major change to the MetUM was implemented in 2014, which was the culmination of many years’ work (Met Office 2014; Walters et al. 2016, manuscript submitted to Geosci. Model Dev.). This included changes to the model dynamics, physics, horizontal resolution, and satellite data usage. Early trials indicated that this change would have a significant impact on TC intensity predictions. Section 3 of this paper evaluates the impact of the model change on TC predictions and explains how this resulted in the development of a completely new form of TC initialization in the MetUM using estimates of central pressure from TC warning centers (discussed in section 4).
2. Evaluation of the 2007 scheme
a. Trial configuration and results
In 2012 experiments were undertaken to assess the impact on TC forecasts of transplanting analysis fields from another model into the MetUM. The European Centre for Medium-Range Forecasts (ECMWF) model was chosen since it had performed better than the MetUM for TC track prediction in the previous few years. The results indicated that the performance of the forecast was very sensitive to the lower-tropospheric winds in the analysis (Heming 2012). Given this result, it was decided to undertake another evaluation of the 2007 scheme to ensure it was still providing benefit to MetUM forecasts of TCs since the scheme primarily adjusted the model’s lower-tropospheric wind fields. Standard control runs of approximately 4–8 weeks are available to model developers at the Met Office. The 2007 scheme was evaluated by using three control runs available at the time for periods during 2010, 2011, and 2012 and running trial forecasts without the 2007 scheme. These periods covered a significant amount of TC activity in both the Northern and Southern Hemispheres. The first period was from 23 August to 19 September 2010, the second from 20 August to 15 October 2011, and the third from 19 January to 18 March 2012. In total there were 57 TCs during these three periods comprising 17 in the western North Pacific, 10 in the eastern North Pacific, 19 in the Atlantic, 4 in the western south Indian Ocean (west of 90°E), and 7 in the eastern south Indian Ocean and South Pacific. A sizeable number of forecasts were verified at each forecast lead time: for example, 460 at 24 h, 243 at 96 h, and 87 at 168 h. The control and trial used the configuration of the MetUM operational from July 2011 to January 2012 (known as OS27), which had a horizontal grid spacing of approximately 25 km at midlatitudes and 70 vertical levels. The control included use of the 2007 scheme to initialize TCs, as was done in the operational model at the time, whereas in the trial this scheme was switched off. Table 1 shows the various verification scores for TC track prediction for the control and trial. Scores were calculated at 12-hourly forecast intervals, but only the 24-hourly values are shown in Table 1. Details of the TC tracking method and verification scores can be found in Heming (2016).
2007 scheme control versus trial results. Mean track forecast statistics, where boldface indicates the better score. The 12-hourly statistics are calculated, but only 24-hourly results are shown. The t values and significance levels (%) of track error differences are shown. The trial had the 2007 scheme switched off.
For TC track the results show that despite much larger analysis errors in the trial, which might be expected because of the removal of the initialization, track forecast errors were lower at all lead times. When averaged over all forecasts from 12 to 168 h at 12-hourly intervals, the trial track forecast errors were 8.4% lower. The reduction in track errors up to 96 h was significant beyond the 1% level. The reduction in errors at longer lead times was significant beyond the 7% level. Model skill in predicting the track of TCs against climatology and persistence (CLIPER; Neumann 1972) was calculated for the first 72 h of the forecast. The trial skill scores were on average 6.0% higher than the control results. The frequency of superior performance for TC track forecasts shows that the trial was superior in 57% of all forecasts compared to 42% for the control. This latter measure indicates the percentage of TC track forecasts that were improved or worsened as a result of switching off the 2007 scheme. Results show that switching off the scheme was beneficial in 15% more forecasts than it was detrimental.
For intensity forecasts (not shown in Table 1) the differences between the control and trial were fairly modest. Trial forecasts on average had a lower central pressure by 0.8 hPa and a higher maximum 10-m wind speed by 0.7 kt (1 kt = 0.51 m s−1). The trial reduced the mean absolute error in maximum 10-m wind speed by just 0.6 kt. Intensity tendency skill (the skill of the model to predict strengthening or weakening) was 1.8% higher in the control than the trial.
b. Individual cases
These results indicate overall that switching off the 2007 scheme reduces TC track forecast errors and has little impact on TC intensity forecasts. However, it is worth looking at a few individual cases to see how switching off the scheme affects some characteristics of the MetUM’s predictions.
Cross-track errors (not shown in Table 1) indicate that control forecasts had an equatorward bias relative to the observed track, particularly in the North Atlantic. Several cases bear this out and show that without the initialization scheme the equatorward bias in forecasts is eradicated. For example, Fig. 2 shows forecast tracks for individual forecasts for Hurricanes Igor, Katia, and Maria. For Hurricane Igor the control persisted with a westward track for too long and was also a little too fast. The trial was slower and turned Igor northwestward sooner, resulting in a much better forecast. For Hurricane Katia the control showed an erroneous turn toward the west, whereas the trial correctly persisted with a northwestward track. Finally, for Hurricane Maria the control again did not turn the system northwest and then north soon enough, resulting in an equatorward bias. The trial was much closer to the observed track.
Control (red circles) and trial (green squares) forecast tracks (24-h steps) for (a) Hurricane Igor from 0000 UTC 11 Sep 2010, (b) Hurricane Katia from 0000 UTC 4 Sep 2011, and (c) Hurricane Maria from 1200 UTC 8 Sep 2011. Corresponding analysis positions are shown as triangles and best-track observed positions as TC symbols (24-h steps). The trial had the 2007 scheme switched off.
Citation: Weather and Forecasting 31, 5; 10.1175/WAF-D-16-0040.1
In another forecast for Hurricane Igor, the trial showed a much better prediction of recurvature and acceleration into the midlatitudes at longer lead times. Figure 3 shows the 168-h forecast from data at 1200 UTC 14 September 2010 for the control and trial alongside the verifying analysis. The control did not engage the hurricane with the midlatitude jet, resulting in it remaining in the subtropics close to Bermuda. Although the trial forecast was a little to the southeast of the verifying position for Hurricane Igor, it was a much better forecast.
(a) Control (solid red lines) and trial (dashed green lines) 168-h forecast of mean sea level pressure from data time 1200 UTC 14 Sep 2010 for Hurricane Igor. (b) Verifying analysis. Trial had the 2007 scheme switched off.
Citation: Weather and Forecasting 31, 5; 10.1175/WAF-D-16-0040.1
In some cases the lack of initialization of the TC in the trial resulted in a poorer analysis position, which in turn resulted in a poorer forecast. For example, Fig. 4a shows forecast tracks for Hurricane Danielle. The trial had a larger initial error than the control, which resulted in a persistent westward bias in the forecast compared to the control, which had a better forecast track. The problem of poorer analyses without initialization was particularly an issue during the formative stages of TCs in the western North Pacific. Figure 4b shows forecast tracks for Tropical Storm Nanmadol at the time when it was a relatively weak tropical storm. There was a large analysis error in the position of the storm in the trial, and the forecast took Nanmadol toward the north and then northeast, whereas the storm actually moved west and then northwest. The control forecast was much better in this case. However, it should be noted that by 2 days later the trial forecast track was slightly better than the control (not shown).
Control (red circles) and trial (green squares) forecast tracks (24-h steps) for (a) Hurricane Danielle from 0000 UTC 26 Aug 2010 and (b) Tropical Storm Nanmadol from 1200 UTC 23 Aug 2011. Corresponding analysis positions are shown as triangles and best-track observed positions as TC symbols (24-h steps). The trial had the 2007 scheme switched off.
Citation: Weather and Forecasting 31, 5; 10.1175/WAF-D-16-0040.1
Cyclone Funso developed in the Mozambique Channel and was initially slow moving but later accelerated southward into the open Indian Ocean. Control forecasts persistently tracked Funso westward toward the coast of Mozambique at a fairly slow pace. The trial forecasts were much better both in terms of direction and speed of movement. This is illustrated in Fig. 5, which shows all forecast tracks from 0000 UTC runs of the control and trial configurations of the model for Cyclone Funso.
(a) Control and (b) trial forecast tracks (colored symbols 24 h apart all valid at 0000 UTC) for Cyclone Funso in January 2012. Best-track observed positions are shown as TC symbols (24-h steps). The trial had the 2007 scheme switched off.
Citation: Weather and Forecasting 31, 5; 10.1175/WAF-D-16-0040.1
c. Conclusion from evaluation of the 2007 scheme
The evidence from the trial presented above was carefully considered in mid-2012 in order to make a decision about the future of the 2007 scheme for initializing TCs.
The mean error statistics for TC tracks give a clear indication that switching off the scheme was beneficial, with an overall reduction of error of 8.4%. This reduction in track error was achieved without any significant detriment to TC intensity forecasts. Individual cases show that with the scheme switched off the model performed much better for the recurvature of strong TCs—particularly in the Atlantic. There was benefit seen in both Northern and Southern Hemisphere TC predictions. On the negative side, there was evidence that switching off the scheme sometimes produced worse forecasts during the formative stages of TCs, most notably in the western North Pacific.
As discussed in the introduction, the 2007 scheme was found to be beneficial to MetUM forecasts of TCs, but to a lesser degree than when the first configuration of the initialization scheme was introduced in 1994. The reduced benefit was likely due to the many improvements in model formulation, increases in model resolution, and the introduction of new observational data, particularly from satellites, that occurred in the intervening period. In the few years immediately following the implementation of the 2007 scheme there were further changes to the MetUM, including increases in model horizontal and vertical resolution, increases in data assimilation resolution, the introduction of hybrid four-dimensional variational data assimilation (4DVAR; Clayton et al. 2013), and further improvements in observational coverage and usage, particularly satellite data. Some of these changes would have acted to constrain the model’s analysis of TCs and reduce the requirement for artificial initialization of the low-level wind field to the point that this form of initialization actually became detrimental to the model’s analysis and forecast capabilities.
Alongside evidence from the trials of the 2007 scheme, consideration was also given to the fact that a major model change, which had been in development for several years and was planned for implementation within the following two years, was showing signs of performing much better for TC prediction in initial tests (Heming 2014). In conjunction with this change there were also plans to test a completely new technique for TC initialization that would be targeted at improving the analysis of TC intensity (discussed later in this paper). Having considered all of these factors, a decision was made to switch off the 2007 scheme for initialization of TCs beginning 17 July 2012.
3. Major upgrade to the MetUM
a. Background to the model upgrade
Upgrades to the dynamical core of the MetUM typically take many years to develop. In 2002 the “New Dynamics” upgrade was implemented (Davies et al. 2005). The successor to New Dynamics was named ENDGame and was implemented in the MetUM operationally in July 2014 (Wood et al. 2014; Met Office 2014). A major benefit from ENDGame was an increase in atmospheric variability. This manifested itself in improved detail and intensity of large-scale storms, which arose from the use of less artificial damping in the ENDGame formulation. A simple measure of the atmospheric variability is provided by the globally integrated eddy kinetic energy. Figure 6, taken from Met Office (2014) (P. Earnshaw 2015, personal communication) illustrates the improved results for ENDGame against its predecessor New Dynamics when compared to independent analyses from ECMWF for a set of twelve 3-day forecasts.
Globally integrated eddy kinetic energy (EKE) for various resolutions from a set of twelve 3-day ENDGame and New Dynamics forecasts initialized from ECMWF analyses. Comparable EKE results from ECMWF analyses are also shown.
Citation: Weather and Forecasting 31, 5; 10.1175/WAF-D-16-0040.1
In addition to the new dynamical core, a number of other changes were also implemented. There was an increase in the model’s horizontal resolution from a grid spacing at midlatitudes of approximately 25 km (N512) to 17 km (N768). The resolution of the data assimilation component was also changed from approximately 60 km (N216) to 40 km (N320). A package of changes to the satellite data assimilation included reduced thinning of some types of data and improved usage of some other types. Changes to model physics included an increase in entrainment rate for deep convection and improvements to several other physical parameterization schemes. The complete package was known as Global Atmosphere 6 (GA6) and is described in detail in work by Walters et al. 2016, manuscript submitted to Geosci. Model Dev.).
b. Trial of the model upgrade
Prior to implementation of this package of changes, the model was trialed using periods from June to September and November to December 2012. There were 36 TCs during these trial periods. The package of changes was tested with and without the increase in horizontal resolution. The two trials were known as Trial25 (25-km grid spacing) and Trial17 (17-km grid spacing). Details of the impact of the trial configurations on TC forecasts from the MetUM can be found in Heming (2014) and are summarized in Table 2.
Summary of differences between the control and trials during the assessment of GA6. Positive indicates trial values higher than control. Negative indicates trial values lower than control.
For TC track prediction the key result was that Trial25 reduced track errors by 7.3% and Trial17 reduced track errors by 8.6% relative to the control. Thus, there was a clear benefit to TC track forecasting from GA6, with most of the benefit coming from the dynamics, physics and satellite data usage changes and with a small additional benefit from the resolution increase.
Results for TC intensity prediction showed that in both trials TCs were much stronger and absolute errors and biases in forecast intensities were reduced compared to the control. In Trial25 the mean forecast central pressure was 7.1 hPa lower and 10-m winds 8.9 kt higher than for the control. The mean absolute error in central pressure forecasts was reduced by 3.0 hPa, and the mean absolute error in 10-m wind forecasts was reduced by 6.7 kt. In Trial17 the mean forecast central pressure was 11.1 hPa lower and 10-m winds 13.4 kt higher than in the control. The mean absolute error in central pressure forecasts was reduced by 3.6 hPa and the mean absolute error in 10-m wind forecasts reduced by 9.0 kt.
For TC intensity prediction it is also worth examining the variation in bias with forecast lead time. Figure 7 is taken from Heming (2014) and shows the mean bias in central pressure for the trials. Figure 7 first shows that the control (New Dynamics with 25-km grid spacing) had a weak bias in the analysis of approximately 15 hPa. This bias initially increased with lead time, as a result of a tendency to spin up TCs at a slower rate than reality, before reducing again at longer lead times to a value close to 15 hPa. Trial25 had a similar weak bias in the analysis, but the bias reduced with increasing forecast lead time to a value of just below 5 hPa by 144 h. For Trial17 there was again a large weak bias in the analysis, but the bias reduced at a faster rate with increasing lead time than for Trial25. In fact, the bias actually became negative at lead times of 120 h and above (i.e., the forecast TCs were too intense on average).
Mean TC central pressure bias during the trial of GA6: Trial25, model upgrade without horizontal resolution increase; Trial17, model upgrade with horizontal resolution increase.
Citation: Weather and Forecasting 31, 5; 10.1175/WAF-D-16-0040.1
GA6 was operationally implemented in the MetUM on 15 July 2014. The impact of GA6 on operational TC forecasts is discussed in section 5.
4. TC initialization using central pressure estimates
a. Background to the new initialization technique
In section 2, we saw that switching off the 2007 scheme in the MetUM was beneficial to TC track forecasting overall with no detriment to TC intensity. Two years after switching off the initialization scheme, a major model upgrade (GA6) resulted in a further reduction in TC track forecast errors as well as a significant reduction in TC intensity forecast bias—particularly at longer lead times, as explained in section 3 above. However, it was notable that GA6 had virtually no impact on the model’s weak bias in the analysis for TCs. While the intensity bias at short lead times was reduced by GA6, TC predictions were handicapped by having to spin up from a relatively weak analysis. Thus, attention was turned to how this weak bias could be reduced, possibly through the development of a new form of TC initialization.
Given that there was clear evidence that the 2007 scheme, which involved the generation of lower-tropospheric wind observations, was now detrimental to TC track forecasts in the MetUM, it was decided not to return to a form of initialization that involved direct adjustments to the wind structure of the model. However, usage of central pressure estimates appeared to have the potential to reduce the model’s weak bias while allowing the model to make its own balanced adjustments to the wind structure. Thus, a new form of TC initialization was developed and tested, which is hereafter referred to as the 2015 scheme.
b. Formulation of the 2015 scheme
TC warning centers around the globe (e.g., Japan Meteorological Agency, National Hurricane Center) produce estimates of the position and structure of all active TCs every 3 or 6 h. These include estimates of maximum sustained wind and central pressure. While in most cases these are estimates (i.e., not directly measured) based on a combination of techniques such as Dvorak (1975, 1984) and Knaff and Zehr (2007), they provide information that is potentially of value to numerical models. While the 2007 scheme made use of the estimates of maximum sustained wind and radii of 34-, 50-, and 64-kt winds provided by TC warning centers, the estimates of central pressure had remained unused by the MetUM.
The 2015 scheme was designed to ingest estimates of the central pressures of all active TCs from a variety of TC warning centers around the globe. These are available at 6-hourly intervals and sometimes at 3-hourly intervals. It was considered that assimilating a single central pressure observation every 6 h may have a limited impact on the MetUM analysis or forecast. Hence, the scheme was designed to produce hourly values of TC central pressure. These are based on a combination of interpolation and extrapolation of the estimates from TC warning centers. For example, the 1200 UTC run of the MetUM, which uses a hybrid 4DVAR data assimilation system (Clayton et al. 2013), has an observational time window from 0900 to 1459 UTC. If estimates of position and central pressure for an active TC are available at 0600 and 1200 UTC, these are used to derive estimates of central pressure at 0900, 1000, and 1100 UTC by linear interpolation. The 1200 UTC value from the warning center is used directly and values for 1300 and 1400 UTC are derived by linear extrapolation. Hence, six values of TC central pressure and position are presented for data assimilation during the 6-h time window and are used in the production of a model analysis, being treated in a similar way to other surface observations, including passing through quality control procedures.
c. Trial of the 2015 scheme
The trial period chosen was 26 September–12 November 2013. In total there were 23 TCs during the trial period, comprising 13 in the western North Pacific, 5 in the eastern North Pacific, 3 in the Atlantic, and 2 in the north Indian Ocean. Forecasts were run to 144 h at 0000 and 1200 UTC and in addition forecasts were run to 48 h at 0600 and 1800 UTC. The numbers of forecasts verified at each forecast lead time were 324 at 24 h, 83 at 72 h, and 24 at 144 h. Although the trial period was from 2013, the control and trial simulations used the configuration of the MetUM operational from July 2014, which included the major model upgrade described in section 3 (GA6). The GA6 configuration of the model was used since it was vital to assessing the 2015 scheme against this baseline, which was already in operation at the time the trials were conducted. Table 3 shows the various verification scores for TC track and intensity prediction for the control and trial runs. Scores were calculated at 6-hourly forecast intervals, but only the 24-hourly values are shown in Table 3.
Results of the 2015 scheme control versus trial. Mean track and intensity forecast statistics, where boldface indicates the better score. The 6-hourly statistics are calculated, but only the 24-hourly results are shown. The t values and significance levels (%) of track error differences are shown. The trial included use of the 2015 scheme.
The results for TC track show that forecast errors were lower in the trial at all lead times. When averaged over all forecasts from 6 to 144 h at 6-hourly intervals, the trial track forecast errors were 6.2% lower. However, statistical significance was mostly not high. At 48 h the reduction in track errors was significant beyond the 2% level, but at other lead times the reduction was significant only between the 10% and 30% levels. The trial track forecast skill scores were on average 2.7% higher than the control. The frequency of superior performance for TC track forecasts shows that the trial was superior in over 50% of all forecasts compared to less than 43% for the control. Track forecast errors for the control and trial are shown in Fig. 8a and also include the values for the configuration of the model that was operational during the trial period (i.e., the version before the major model upgrade described in section 3). This illustrates the combined impact of GA6 and the 2015 scheme.
(a) TC track forecast errors and (b) TC central pressure forecast bias during the trial of the 2015 scheme. The control and trial used the GA6 configuration of the MetUM. The trial included the 2015 scheme. The operational results were for an earlier model configuration.
Citation: Weather and Forecasting 31, 5; 10.1175/WAF-D-16-0040.1
For TC intensity, the statistics indicate that the weak bias in the analysis was markedly reduced in the trial with the central pressure bias of 11.5 hPa cut to 6.3 hPa. With increasing lead time the difference between the control and trial narrowed and beyond 108 h both had a bias very close to zero. Thus, the introduction of assimilation of central pressure estimates has reduced the weak bias in the analysis and short-lead-time forecasts without resulting in an overdeepening at longer lead times when the control already had a very small bias. Figure 8b shows these results together with the central pressure bias for the configuration of the model, which was operational during the trial period. This shows that the combination of GA6 and the 2015 scheme slashes the bias in forecast central pressure. Biases that ranged between approximately 17 and 28 hPa before these two changes now ranged between approximately 0 and 10 hPa.
d. Case studies
Examination of some individual cases illustrates some of the characteristics of the 2015 scheme.
1) Hurricane Raymond
The reduction in track forecast errors is exemplified in a forecast for Hurricane Raymond in the eastern North Pacific from data at 1200 UTC 23 October 2013, shown in Fig. 9. The control forecast had a fast westward movement for the hurricane whereas the trial had a slower movement with a gradual curve toward the north. The latter matches the observed track far better and thus produced much lower track forecast errors. At the analysis time in this case the observed central pressure was 995 hPa. The control analysis had a central pressure of 1003 hPa whereas the trial had a central pressure of 996 hPa. The lower central pressure in the trial was accompanied by a stronger and vertically deeper vortex, as seen in the cross sections of the u component of the wind shown in Fig. 10. The difference in vortex depth between the control and trial resulted in differences in the steering level. Thus, the control and trial forecast tracks started to diverge. By 96 h into the forecast the control was erroneously tracking Raymond westward whereas the trial was correctly turning it to the northwest and then north. The cross sections of the u component of the wind at this time (Fig. 11) show that the control had a weak, shallow, and sheared vortex whereas the trial had a strong and vertically deep vortex. In reality, Hurricane Raymond was going through a period of intensification at this time and had developed a strong and deep vortex, as in the trial forecast. Clearly, in this case the assimilation of central pressure estimates helped develop a stronger vortex and a more accurate vertical structure and steering level, which in turn resulted in a better forecast track.
Control (red circles) and trial (green squares) forecast tracks (24-h steps) from data time 1200 UTC 23 Oct 2013 plotted against best-track observed positions (24-h steps) for Hurricane Raymond. Corresponding analysis positions are shown as triangles. The trial included the 2015 scheme.
Citation: Weather and Forecasting 31, 5; 10.1175/WAF-D-16-0040.1
(a) Control and (b) trial vertical cross sections of the analyzed u component of the wind for Hurricane Raymond valid at 1200 UTC 23 Oct 2013. The trial included the 2015 scheme.
Citation: Weather and Forecasting 31, 5; 10.1175/WAF-D-16-0040.1
(a) Control and (b) trial vertical cross sections of the 96-h forecast u component of the wind for Hurricane Raymond valid at 1200 UTC 27 Oct 2013. The trial included the 2015 scheme.
Citation: Weather and Forecasting 31, 5; 10.1175/WAF-D-16-0040.1
2) Typhoons Wutip and Wipha
The impact of the 2015 scheme on TC intensity can be seen in the central pressure predictions for Typhoons Wutip and Wipha. During the period from 27 to 29 September 2013 Typhoon Wutip deepened from 1000 to 965 hPa. The control analysis did not keep pace with the rate of deepening and the subsequent forecasts were consequently unable to predict the intensity of the typhoon. However, the assimilation of central pressure estimates resulted in the analysis being much closer to the observed intensity and even being too deep in a couple of runs. The resulting forecasts had central pressures much closer to the observed values in most cases, as seen in Fig. 12.
Control (solid red lines) and trial (dashed green lines) central pressure forecasts plotted against the best-track observed data (solid blue line) for Typhoon Wutip (September 2013). The trial included the 2015 scheme.
Citation: Weather and Forecasting 31, 5; 10.1175/WAF-D-16-0040.1
For Typhoon Wipha (Fig. 13) the control analysis was unable to represent the intensity of the TC and by 1200 UTC 13 October 2013 had a central pressure of 969 hPa compared to an observed value of 930 hPa. In contrast the trial analysis central pressure was 935 hPa. Consequently, trial forecasts of central pressure were much better in most cases. However, this case illustrates a known characteristic of the MetUM in that it tends to continue deepening the TCs beyond the point at which they reach their peak intensity in reality as they move into the subtropics. This resulted in an overdeepening in some forecasts, particularly for the trial. There were also a couple of trial analyses that had large negative central pressure biases (i.e., low centers too deep). This can happen as a result of assimilating central pressure observations that have observation minus background values larger (in absolute terms) than the difference between the observed and background central pressure because of a positional error in the location of the TC in the background field.
Control (solid red lines) and trial (dashed green lines) central pressure forecasts plotted against the best-track observed data (solid blue line) for Typhoon Wipha (October 2013). The trial included the 2015 scheme.
Citation: Weather and Forecasting 31, 5; 10.1175/WAF-D-16-0040.1
3) Typhoon Haiyan
Typhoon Haiyan devastated parts of the central Philippines and was likely the most intense TC recorded at landfall (Lander 2014). It occurred in November 2013, which fell during the latter part of the trial period for the 2015 scheme. The control and trial forecasts of the track of Typhoon Haiyan were both very good with 72–96-h forecasts of landfall having an error around 100 km. Typhoon Haiyan went through rapid intensification with the central pressure dropping 60 hPa in 24 h and 87 hPa in 48 h. The MetUM was unable to simulate these extreme rates of deepening, even over a period of 6 h—the length of forecast used as the “background” for the next model cycle. Thus, the observation minus background values for the central pressure observations created by the 2015 scheme were very large. These observations are subject to the same quality control procedures as other conventional observations. On this basis, once rapid intensification of Typhoon Haiyan was under way (around 0600 UTC 5 December 2013) all central pressure observations were flagged because of the large observation minus background values and were not assimilated into the model. The consequence of this is that the MetUM predictions of the central pressure from about 12–24 h after this point were no better in the trial than the control, as seen in Fig. 14.
Control (solid red lines) and trial (dashed green lines) central pressure forecasts plotted against the best-track observed data (solid blue line) for Typhoon Haiyan (November 2013). The trial included the 2015 scheme.
Citation: Weather and Forecasting 31, 5; 10.1175/WAF-D-16-0040.1
e. Conclusion from trial of the 2015 scheme
The 2015 scheme was developed primarily with the aim of reducing the weak bias in model forecast intensities at short lead times. Evidence from the trial indicates that this was achieved without causing a significant overdeepening at longer lead times. The 2015 scheme was not developed with the primary aim of reducing TC track forecast errors. However, the trial track forecast errors were reduced by 6.2%, although the statistical significance of this result was mostly not high.
The issue of central pressure observations being flagged by quality control when TCs rapidly intensify was seen in the case of Typhoon Haiyan and several other cases in the trial. While this is not ideal, it is considered that in a case such as Typhoon Haiyan when the central pressure value was close to 900 hPa and the pressure gradient near the center of the TC was near 42 hPa over one model grid length (17 km) (Morgerman 2014), it may not be appropriate to assimilate central pressure values in the current configuration of the MetUM. This is a matter that will be the subject of further investigation and experimentation.
The evidence from the trial presented above was considered in late 2014. Given the positive results overall the decision was made to implement the 2015 scheme in the MetUM on 3 February 2015.
5. Operational impact of GA6 and the 2015 scheme
Figure 15 shows the sequence of changes to TC initialization and the MetUM that have been discussed thus far. Examination is now made of the impact of the last two of these changes on operational forecasts of TCs in the MetUM.
Timeline of model changes and TC initialization in the MetUM.
Citation: Weather and Forecasting 31, 5; 10.1175/WAF-D-16-0040.1
GA6 was implemented in the MetUM on 15 July 2014. Thus, most of the 2014 Northern Hemisphere TC season occurred after the implementation date. The 2015 scheme was implemented on 3 February 2015 meaning the whole of the 2015 Northern Hemisphere TC season occurred after the implementation date. Hence, a time series of Northern Hemisphere TC forecast errors from the MetUM gives a good perspective on the impact of these two changes after they became operational.
The mean TC track forecast error for Northern Hemisphere TCs in 2014 was almost 25% lower than the mean for the previous five seasons (2009–13) and in 2015 was a further 3.2% below the 2014 figure. Even when examining the 5-yr running mean of TC track forecast errors, which normally smooths out large interannual variability, there was still a sharp drop in the errors in 2014–15, as seen in Fig. 16.
The 5-yr running mean of MetUM Northern Hemisphere TC track forecast errors.
Citation: Weather and Forecasting 31, 5; 10.1175/WAF-D-16-0040.1
Figure 17 shows the Northern Hemisphere TC central pressure bias for the years 2011–15. In 2014, although analyses were still too weak, the central pressure bias dropped steadily with forecast lead time to a value close to zero by 168 h. In 2015 the bias in the analysis was more than halved compared to previous years, and the bias in the forecast was also markedly reduced. At longer lead times (beyond 96 h), the bias was very close to zero.
Mean TC central pressure bias in the Northern Hemisphere for the MetUM.
Citation: Weather and Forecasting 31, 5; 10.1175/WAF-D-16-0040.1
On an international level the Met Office participates in the World Meteorological Organization’s Working Group on Numerical Experimentation (WGNE), which has undertaken an annual comparison of numerical model TC forecasts since the early 1990s. Results are available to the end of 2014, which includes the period immediately after implementation of GA6 in the MetUM. The MetUM performed well in 2014 against other global models, particularly in the western North Pacific region. Figure 18 (taken from WGNE’s “Intercomparison of Tropical Cyclone Track Forecasts Using Operational Global Models” website) is a time series of 72-h track forecast errors for this region. It shows that the MetUM (labeled UKMO in Fig. 18) outperformed all other models in 2014 including that of the ECMWF, which has been the best-performing model in recent years.
The 72-h TC track forecast errors in the western North Pacific for global models (taken from WGNE’s “Intercomparison of Tropical Cyclone Track Forecasts Using Operational Global Models” website). The MetUM is labeled as UKMO.
Citation: Weather and Forecasting 31, 5; 10.1175/WAF-D-16-0040.1
These operational results support the results seen in the separate trials of both GA6 and the 2015 scheme.
6. Conclusions
Between 2012 and 2015 three changes were made to the MetUM that had large impacts on TC prediction. First, the old TC initialization scheme (2007 scheme) was switched off. There was then a major model upgrade to the model dynamics, physics, resolution, and satellite data usage (GA6). Finally, a new form of TC initialization was introduced involving the assimilation of central pressure estimates (2015 scheme). Results presented in this paper show that each of these changes resulted in a significant reduction in TC forecast errors (for track, intensity, or both) in forecasts from the MetUM.
Issues that require further research include how the 2015 scheme handles rapidly intensifying TCs and the MetUM’s propensity on occasions to continue deepening TCs beyond their actual points of peak intensity as they move into the subtropics. This will be undertaken against a backdrop of wider model development that in the coming years will include the development of the convective parameterization, further increases in horizontal and vertical resolution, and coupling to the ocean, all of which are likely to have an impact upon the TC forecast performance of the MetUM.
REFERENCES
Clayton, A. M., Lorenc A. C. , and Barker D. M. , 2013: Operational implementation of a hybrid ensemble/4D-Var global data assimilation system at the Met Office. Quart. J. Roy. Meteor. Soc., 139, 1445–1461, doi:10.1002/qj.2054.
Davies, T., Cullen M. J. P. , Malcolm A. J. , Mawson M. H. , Staniforth A. , White A. A. , and Wo N. , 2005: A new dynamical core for the Met Office’s global and regional modelling of the atmosphere. Quart. J. Roy. Meteor. Soc., 131, 1759–1782, doi:10.1256/qj.04.101.
Dvorak, V., 1975: Tropical cyclone intensity analysis and forecasting from satellite imagery. Mon. Wea. Rev., 103, 420–430, doi:10.1175/1520-0493(1975)103<0420:TCIAAF>2.0.CO;2.
Dvorak, V., 1984: Tropical cyclone intensity analysis using satellite data. NOAA Tech. Rep. NESDIS 11, 47 pp. [Available from NOAA/NESDIS, 5200 Auth Rd., Washington, DC 20233.]
Gross, J., 1996: 1995 National Hurricane Center forecast verification. National Hurricane Center, Miami, FL, 14 pp. [Available online at http://www.nhc.noaa.gov/verification/pdfs/Verification_1995.pdf.]
Heming, J. T., 2009: Evaluation of and improvements to the Met Office tropical cyclone initialisation scheme. Meteor. Appl., 16, 339–351, doi:10.1002/met.129.
Heming, J. T., 2012: Tropical cyclone predictions from the Met Office and ECMWF global models - Relative performance and error diagnosis. Proc. 30th Conf. on Hurricanes and Tropical Meteorology, Ponte Vedra Beach, FL, Amer. Meteor. Soc., 4A.6. [Available online at https://ams.confex.com/ams/30Hurricane/webprogram/Paper205026.html.]
Heming, J. T., 2014: The impact on tropical cyclone predictions of a major upgrade to the Met Office Global Model. Proc. 31st Conf. on Hurricanes and Tropical Meteorology, San Diego, CA, Amer. Meteor. Soc., 11A.3. [Available online at https://ams.confex.com/ams/31Hurr/webprogram/Paper243428.html.]
Heming, J. T., 2016: The method used to verify the forecast tracks of tropical cyclones in Met Office numerical weather prediction models. Meteor. Appl., in press.
Heming, J. T., and Radford A. M. , 1998: The performance of the United Kingdom Meteorological Office global model in predicting the tracks of Atlantic tropical cyclones in 1995. Mon. Wea. Rev., 126, 1323–1331, doi:10.1175/1520-0493(1998)126<1323:TPOTUK>2.0.CO;2.
Heming, J. T., Chan J. C. L. , and Radford A. M. , 1995: A new scheme for the initialisation of tropical cyclones in the UK Meteorological Office global model. Meteor. Appl., 2, 171–184, doi:10.1002/met.5060020211.
Knaff, J. A., and Zehr R. M. , 2007: Reexamination of tropical cyclone pressure–wind relationships. Wea. Forecasting, 22, 71–88, doi:10.1175/WAF965.1.
Lander, M. A., 2014: Super Typhoon Haiyan’s 170 kt peak intensity: Has Super Typhoon Tip been dethroned? Proc. 31st Conf. on Hurricanes and Tropical Meteorology, San Diego, CA, Amer. Meteor. Soc., 9C.8. [Available online at https://ams.confex.com/ams/31Hurr/webprogram/Paper245225.html.]
Met Office, 2014: ENDGame: A new dynamical core for seamless atmospheric prediction. [Available online at http://www.metoffice.gov.uk/research/news/2014/endgame-a-new-dynamical-core.]
Morgerman, J., 2014: iCyclone chase report: Super Typhoon HAIYAN in Tacloban City & Leyte, Philippines. iCyclone.com Chase Rep., 47 pp. [Available online at http://www.icyclone.com/upload/chases/haiyan/iCyclone_HAIYAN_in_Tacloban_City_040314.pdf.]
Neumann, C. J., 1972: An alternative to the HURRAN tropical cyclone forecast system. Mon. Wea. Rev., 100, 245–255, doi:10.1175/1520-0493(1972)100<0245:PAOTHT>2.3.CO;2.
Wood, N., and Coauthors, 2014: An inherently mass-conserving semi-implicit semi-Lagrangian discretization of the deep-atmosphere global non-hydrostatic equations. Quart. J. Roy. Meteor. Soc., 140, 1505–1520, doi:10.1002/qj.2235.
