SEARCH

SEARCH BY CITATION

Keywords:

  • spectral nudging;
  • added values;
  • regional climate simulation

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model and Experiments
  5. 3. Sensitivities of Simulated Typhoons to Spectral Nudging
  6. 4. Effect of Intermittent Spectral Nudging on Seasonal Simulation
  7. 5. Summary and Concluding Remarks
  8. Acknowledgments
  9. References
  10. Supporting Information

[1] This study examines simulated typhoon sensitivities to spectral nudging (SN) to investigate the effects on values added by regional climate models, which are not properly resolved by low-resolution global models. SN is suitably modified to mitigate its negative effects while maintaining the positive effects, and the effects of the modified SN are investigated through seasonal simulations. In the sensitivity experiments to nudging intervals of SN, the tracks of simulated typhoons are improved as the SN effect increases; however, the intensities of the simulated typhoons decrease due to the suppression of the typhoon developing process by SN. To avoid such suppression, SN is applied at intermittent intervals only when the deviation between the large-scale driving forcing and the model solution is large. In seasonal simulations, intermittent SN is applied for only 7% of the total time steps; however, this results in not only maintaining the large-scale features of monsoon circulation and precipitation corresponding to observations but also improving the intensification of mesoscale features by reducing the suppression.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model and Experiments
  5. 3. Sensitivities of Simulated Typhoons to Spectral Nudging
  6. 4. Effect of Intermittent Spectral Nudging on Seasonal Simulation
  7. 5. Summary and Concluding Remarks
  8. Acknowledgments
  9. References
  10. Supporting Information

[2] Since the late 1980s, regional climate models (RCMs), which can reproduce regional or local details embedded in low-resolution large-scale driving forcings (e.g., general circulation model and global reanalysis data), have been generally used in a number of climate studies. RCMs can improve simulation at the regional scales, because they can generate added values beyond the highest resolved wavelength of the global model through detailed topography information, higher model resolution, and sophisticated physical processes. Therefore, RCMs have been used not only to reproduce severe weather and extreme climate events but also to project regional climate change by dynamically downscaling low-resolution global model and reanalysis data [e.g., Feser and von Storch, 2008; Giorgi, 1990; Giorgi et al., 1994; Lee et al., 2004; Leung and Ghan, 1999; Wang et al., 2003].

[3] Although considerable efforts have been devoted to their development and improvement, most RCMs have systematic errors that are associated with uncertainties in dynamics, physical parameterization, boundary condition, initialization, domain choice, and model resolution of the numerical models [Giorgi and Mearns, 1999; Wang et al., 2004]. In addition, systematic errors in the RCMs can result from the strong internal forcings generated by the peculiar natural characteristics of specific regions. For example, complex topography and land use, warm local sea surface temperature (SST), and strong monsoon and typhoons can cause significant systematic errors in long-term regional climate simulations over East Asia [Cha and Lee, 2009; Hong and Juang, 1998; Suh and Lee, 2004; Zhong, 2006].

[4] An important issue concerning long-term RCM simulations longer than the seasonal scale is the lateral boundary condition, which controls the consistency between the model solution and the large-scale driving forcing. Since most RCMs have been developed based on limited area models, they generally employ the traditional relaxation method proposed by Davies [1976] as a lateral boundary condition. This consists of applying a Newtonian nudging, which drives the model solution toward the large-scale driving forcing within lateral buffer zones. There have been several studies on the traditional relaxation method that modified the size of the buffer zone and weighting function of Newtonian nudging. For example, Giorgi et al. [1993] modified the relaxation technique by which a wider buffer zone was adopted in the upper troposphere rather than in the middle and lower troposphere.

[5] Spectral nudging (SN) has recently been applied as an addition to the traditional lateral boundary condition (i.e., relaxation method) to ensure coherence of the large scales simulated by the regional climate model with those of the driving data. To provide consistency between large-scale driving fields and nested model solutions, Kida et al. [1991] and Sasaki et al. [1995] introduced an alternative approach, in which the large-scale fields of the model solutions throughout the entire model domain, while the regional model generates the higher frequencies. Von Storch et al. [2000] used different method from these studies in terms of the nudging coefficients, which were applied above 850 hPa with increasing strength for higher model levels. A number of studies have shown that SN can improve the RCM performance in simulating the mean features of the forced large-scale climate. Miguez-Macho et al. [2005] showed that implementing SN can successfully improve the spatial pattern of simulated precipitation. Lee et al. [2004], Kang et al. [2005], and Tang et al. [2010] showed that RCMs based on the MM5 model can reproduce extreme summer precipitation events over East Asia by providing appropriate large-scale circulations resulting from the effect of SN. Feser and von Storch [2008] demonstrated that SN has an positive effect on reducing the track distance error of RCM-simulated typhoons, and Knutson et al. [2007] showed that a RCM with SN improved the interannual variability of hurricane occurrences by decreasing the number of simulated hurricanes in inactive seasons. To reduce systematic large-scale errors in a regional spectral model, Kanamaru and Kanamitsu [2007] recently proposed a similar approach, the scale selective bias correction (SSBC) method, in which the errors in large-scale horizontal wind components are reduced using spectral damping to the tendency, and the area mean biases of mass fields are forced to zero. In addition, Kanamitsu et al. [2010] modified the SSBC method where the nudging is applied only to the rotational wind components itself (not the tendency) and the area average moisture correction is excluded, and they showed new method played a role in reducing the systematic errors in the interannual variabilities of simulated height, temperature, and wind fields.

[6] Up to now, most of previous studies on SN focused on its positive effects in regional climate simulations. In principle, SN should not impede the ability of the RCM to develop regional and small-scale features superimposed on the large-scale driving conditions [Biner et al., 2000; von Storch et al., 2000]. However, SN may also have negative effects by impeding the development of the intrinsic small-scale features reproduced by RCMs, which are not included in the large-scale driving forcing. Alexandru et al. [2009] indicated a noticeable reduction of precipitation extremes as a side effect of SN, and these effects are mostly perceived when SN is the most intense. Therefore, investigating and mitigating the disadvantages of SN is necessary for improving RCM performances.

[7] Since SN is an empirical technique, its effect can be changed by controlling some factors such as nudging interval, weighting, cutoff wavelength, and nudged component. In this study, we use nudging interval to investigate the effects of SN on regional climate simulations, in particular simulated tropical cyclone. SN is applied to the Weather Research and Forecasting (WRF) model [Skamarock et al., 2005] for regional climate simulations over East Asia and the western North Pacific (WNP), and the positive and negative effects of SN are examined through sensitivity experiments on typhoon simulations to the nudging intervals. To reduce the negative effects of SN, nudging is applied intermittently and only when the deviation between the model solution and the large-scale boundary forcing is large, and the effects of intermittent SN on simulating the East Asian summer climate are analyzed. We describe the regional climate model used in this study in section 2. In section 3, the effects of SN on typhoon simulations are analyzed through sensitivity tests to the nudging intervals. In section 4, intermittent SN to reduce the negative effects of SN is introduced, and its effects on seasonal simulation are investigated. Finally, the summary and conclusions are given in section 5.

2. Model and Experiments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model and Experiments
  5. 3. Sensitivities of Simulated Typhoons to Spectral Nudging
  6. 4. Effect of Intermittent Spectral Nudging on Seasonal Simulation
  7. 5. Summary and Concluding Remarks
  8. Acknowledgments
  9. References
  10. Supporting Information

[8] The RCM used in this study is based on the WRF model version 2.2, which is developed by the National Center for Atmospheric Research (NCAR). To improve the performance of the model, the SN method proposed by von Storch et al. [2000] is used along with the relaxation method for an alternative boundary condition. SN is applied to the long-wave spectral regimes (wavelength > 1000 km) of the horizontal wind components itself (not the tendency) at every integral time step over the entire model domain, which is expressed as

  • equation image

where LG and LR are the long-wave spectral regimes in the global and regional models, respectively. αG, αR, and α*R are variables corresponding to large-scale driving fields, nudged fields, and simulated fields from the regional models, respectively. The nudging coefficient η is a function of height and is given by η(σ) = 0.05(1 − σ)2, where σ is the vertical coordinate. Therefore, the nudging weighting is smaller at the lower troposphere than the upper troposphere, indicating the relatively weak impact of SN at low levels. For SN implementation, a module is added to the WRF model; the model solution and large-scale driving forcing are spectrally decomposed by the Fourier transform and only their long-wave regimes (wavelength > 1000 km) are recomposed on the basis of equation (1) at every integral time step.

[9] The model domain has 240 × 200 grid points in each of the zonal and meridional directions with a horizontal grid spacing of 30 km, and covers East Asia and the WNP (Figure 1). Ten grid points are used for lateral boundary forcing in the buffer zone. There are 27 vertical layers between the model top at 50 hPa and the surface, and the time step of the model integration is 90 s. The physical parameterization schemes used in this study are the Kain-Fritsch convective parameterization scheme [Kain and Fritsch, 1990], WSM3 cloud microphysics scheme [Hong et al., 2004], CAM3 radiation scheme [Collins et al., 2004], YSU planetary boundary layer scheme [Hong et al., 2006], and NOAH land surface model [Chen and Dudhia, 2001]. The National Centers for Environmental Prediction/Department of Energy (NCEP/DOE) R-2 reanalysis data [Kanamitsu et al., 2002] are used to obtain the initial and boundary data to drive the RCM, and the Optimal Interpolation SST data [Reynolds et al., 2002] are prescribed every 24 h.

image

Figure 1. Model domain and topography with contour intervals of 200 m.

Download figure to PowerPoint

[10] To investigate the effects of SN, sensitivity experiments to the nudging interval are conducted for the simulations of a strong typhoon that occurred over the WNP in 1994 (1–13 August). The sensitivity experiments are performed for four nudging intervals: 90 s (EXP_1DT run), 450 s (EXP_5DT run), 15 min (EXP_10DT run), and 30 min (EXP_20DT run). To reduce the negative effects of SN, we develop an intermittent SN, where SN is applied intermittently when the deviation between the model solution and the large-scale driving forcing is large. To examine the effects of intermittent SN, we further conduct three seasonal simulations over East Asia for the summer (MJJA) of 1994, when numerous typhoons and extreme drought occurred over East Asia; the NOSP run without SN, the SP run using continuous SN similar to the EXP_1DT run, and the ISP run using intermittent SN.

[11] To evaluate the performance of the model, simulated upper air variables are compared with the reanalysis (R-2) driving fields. In addition, the Global Precipitation Climatology Project (GPCP) monthly data with 2.5° × 2.5° horizontal resolution [Adler et al., 2003] are used for precipitation evaluation. Furthermore, the best track data obtained from the Regional Specialized Meteorological Center (RSMC) Tokyo–Typhoon Center are used for evaluating the simulated typhoons. In this study, simulated typhoons are identified by the following searching method from Camargo et al. [2007]: (1) The potential storm is a local minimum of sea level pressure. (2) Surface wind exceeds 14.1 m s−1, which indicates the grade of tropical storm. (3) A local relative vorticity maximum at 850 hPa exceeds 4.9 × 10−5 s−1. (4) The temperature structure aloft has a marked warm core such that the sum of the temperature deviations at 300, 500 and 700 hPa exceeds 1.2 K. (5) The maximum wind speed at 850 hPa is larger than that at 300 hPa. (6) The duration is not shorter than 2 d.

3. Sensitivities of Simulated Typhoons to Spectral Nudging

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model and Experiments
  5. 3. Sensitivities of Simulated Typhoons to Spectral Nudging
  6. 4. Effect of Intermittent Spectral Nudging on Seasonal Simulation
  7. 5. Summary and Concluding Remarks
  8. Acknowledgments
  9. References
  10. Supporting Information

[12] A typhoon is one of typical mesoscale features that is not well resolved by low-resolution global climate models. Therefore, in this section, we analyze the sensitivities of simulated typhoons to the nudging intervals in order to investigate the effect of SN on values added by a RCM. In 1994, the 13th typhoon, Doug, which was generated over the WNP, moved to the East China Sea via Taiwan and caused serious damages in the East Asian countries. Figure 2 shows the track and life cycle of typhoon Doug simulated using various nudging intervals. The mean track distance errors between 3 and 11 August are 157.4 (EXP_1DT run), 168.6 (EXP_5DT run), 178.1 (EXP_10DT run), and 209.3 km (EXP_20DT run) (Figure 2a). That is, the track distance errors of the simulated typhoons tend to decrease with an increase in the nudging effect (i.e., decrease in the nudging interval). Feser and von Storch [2008] showed that the track error in the reanalysis data was relatively small as compared to that in a RCM forced by the reanalysis data. Therefore, the track error can be reduced by increasing the nudging effect, since the realistic track information embedded in the reanalysis data is strongly nudged to model solution. In contrast, the life cycles of simulated typhoons including the developing processes tend to be improved with a decrease in the nudging effect (i.e., increase in the nudging interval) (Figures 2b and 2c). The EXP_1DT run, in which the nudging effect is relatively large due to a short nudging interval corresponding to the time step of integration cannot reasonably capture the developing processes. Therefore it simulates a higher minimum sea level pressure and weaker maximum surface wind as compared to other experiments. In terms of both the track distance error and intensity of the simulated typhoons, the EXP_10DT run has the best results as compared to other experiments. In addition, the e-folding time corresponding to the EXP_10DT run is approximately 6 h, thus the maximum SN strength in the EXP_10DT run is most reasonable considering the e-folding times of von Storch et al. [2000] and Biner et al. [2000] are approximately 2 h and 9 h, respectively. The EXP_10DT run simulates a maximum surface wind at the peak time that is approximately 20% stronger than the EXP_1DT run, although this is still underestimated as compared with observation. This indicates that the larger SN effects result in weaker intensities of the simulated typhoons.

image

Figure 2. (a) Tracks, (b) time series of minimum sea level pressure (hPa), and (c) time series of maximum surface wind (m s−1) for typhoon Doug as simulated from sensitivity experiments to nudging interval. The best track data set from the RSMC is used as the observed typhoon.

Download figure to PowerPoint

[13] In the EXP_1DT run with a relatively strong SN effect, the track distance error is small, but the intensity of the simulated typhoon is relatively weak. On the contrary, in the EXP_10DT run with a relatively weak SN effect, the life cycle of the simulated typhoon is prominently improved by enhancing typhoon intensity despite the increased track distance error. This indicates that SN can have a negative effect of decreasing the typhoon intensity, while also having the positive effect of improving the typhoon track. To investigate the negative effect of SN decreasing the intensities of simulated typhoons, the vertical structures of simulated typhoons at the peak time of the life cycle are analyzed in Figure 3. In the EXP_1DT run with a relatively strong SN effect, the maximum tangential velocity at about 850 hPa is weakened to below 50 m s−1 and the vertical structure of the eyewall is not well developed (Figure 3a). This is because the vertical development of the typhoon intrinsically generated by the RCM is suppressed by the strong SN effect at the upper levels with a large weighting factor. In the EXP_10DT run with a relatively weak SN effect, on the contrary, the maximum tangential velocity is enhanced to more than 50 m s−1, and the typhoon is vertically well developed (Figure 3b). In the EXP_1DT run, strong winds faster than 30 m s−1 exists only below 500 hPa, while it extends to above 300 hPa in the EXP_10DT run. In addition, the EXP_10DT run simulates not only a stronger upper level divergence but also an enhanced low level convergence as compared with the EXP_1DT run (Figures 3c and 3d); this is because SN decreases upper level divergence by inducing it to large-scale driving forcing due to the stronger effect of SN at the upper levels. Therefore, the intensity of the simulated typhoon at the peak time tends to be enhanced and the vertical structure tends to be better developed as the effect of SN decreases.

image

Figure 3. Vertical cross section of azimuthally averaged (a and b) tangential velocity (m s−1) and (c and d) radial velocity (m s−1) at each peak time for the maximum surface wind speed in the EXP_1DT (Figures 3a and 3c) and EXP_10DT (Figures 3b and 3d) runs.

Download figure to PowerPoint

[14] The intensity of the simulated typhoon is weakened in the EXP_1DT run because SN inhibits the intrinsic intensification, i.e., the self-intensifying feedback that provides a mature typhoon with energy. In the developing stages of the typical typhoon, condensation heating by convection enhances not only upper level divergence but also low level wind and convergence; this in turn results in increased surface evaporation to provide energy from the ocean to the atmosphere. Much of the released energy causes further enhanced convection and consequently more condensation heating. This positive self-intensifying feedback continues as long as the conditions are favorable for typhoon development. Figure 4 shows the temporal variations of surface variables and atmospheric variables associated with the self-intensifying feedback simulated in the EXP_1DT and EXP_10DT runs. In the mature stage of typhoon Doug during 5–7 August, the midlevel convection, low level wind speed and convergence, upper level divergence, boundary layer mixing, and latent heat flux from the sea surface in the EXP_1DT run are underestimated as compared with those in the EXP_10DT run. This implies that the EXP_1DT run simulates a weaker intensification of the simulated typhoon than the EXP_10DT run, since the strong effects of SN in the EXP_1DT run suppresses the positive self-intensifying feedback. As shown in the vertical structure of the simulated typhoon at the peak life time, SN can decrease upper level divergence due to the strong effect of SN at the upper levels, and therefore, the intrinsic self-intensifying feedback generated by RCM can be decreased by the effect of SN.

image

Figure 4. Time series of (a) minimum sea level pressure (hPa), (b) area mean 850 hPa wind speed (m s−1), (c) area mean 925 hPa convergence (10−5 s−1), (d) area mean 500 hPa vertical velocity (cm s−1), (e) area mean surface latent heat flux (W m−2), (f) area mean frictional velocity (cm s−1), and (g) area mean 200 hPa divergence (10−5 s−1) around the typhoon center (within a radius of 150 km from the typhoon center).

Download figure to PowerPoint

4. Effect of Intermittent Spectral Nudging on Seasonal Simulation

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model and Experiments
  5. 3. Sensitivities of Simulated Typhoons to Spectral Nudging
  6. 4. Effect of Intermittent Spectral Nudging on Seasonal Simulation
  7. 5. Summary and Concluding Remarks
  8. Acknowledgments
  9. References
  10. Supporting Information

[15] In section 3, we showed that SN can have a negative effect by suppressing intrinsic mesoscale features, which are values added by RCMs. To overcome the disadvantages of SN while maintaining its advantages, we apply SN only when the deviation between the large-scale driving forcing and the model solution exceeds a certain level. That is, SN is applied intermittently when the pattern correlation of the horizontal wind components (U or V) from field itself at 500 hPa level between the model solution and the large-scale driving forcing (i.e., R-2 reanalysis) over the entire domain becomes less than 0.90. We select the spatial patterns of 500 hPa wind fields as the parameters for intermittent SN, because midlevel wind is closely associated with subtropical high, which is an important synoptic condition over East Asia and the WNP. Also, we investigated sensitivities to pattern correlations between 0.80 and 0.95 and concluded that 0.90 is the best pattern correlation for reducing the negative effects of SN while maintaining the positive effects (not shown).

[16] To examine the effect of intermittent SN, we conduct 4 month (MJJA) simulations over East Asia and the WNP; the NOSP run without SN, the SP run using continuous SN, and the ISP run using intermittent SN. Figure 5 shows the seasonal mean (JJA) precipitation and low level winds. During the summer of 1994, extreme drought occurred over East Asia, and the frequency of typhoon occurrences over the WNP was anomalously larger than climatology. In observation, the precipitation amounts over the midlatitude regions (e.g., central China, Korea, and Japan) are relatively small due to the weak East Asian summer monsoon, whereas those over the tropical regions (e.g., southern China, the Philippines, and tropical WNP) are relatively large due to the strong western North Pacific summer monsoon (Figure 5a). These anomalous precipitation characteristics over East Asia and the WNP during the summer of 1994 can be associated with anomalous synoptic features such as the intensified low level westerly from the Bay of Bengal to the tropical WNP and the weak expansion of the western North Pacific subtropical high. In the NOSP run without SN application, the simulated summer monsoon rainband over the WNP is not properly reproduced as compared with the observed rainband, which well extends from southern China to the Philippines and tropical WNP (Figure 5b). In the NOSP run, the monsoon rainband is unreasonably separated into two weak monsoon rainbands; one extending from the East China Sea to the tropical WNP, and the other extending from the South China Sea to the tropical WNP. Also, the NOSP run cannot capture heavy precipitation over southern China. This precipitation error is because the NOSP run underestimates anticyclonic circulation over the WNP related to the subtropical high and cyclonic circulation from the Bay of Bengal to China associated with severe precipitation over southern China as compared with observation. In contrast, the SP run where continuous SN is applied captures severe floods over southern China and a well-developed summer monsoon rainband over the WNP; this is because SN decreases systematic errors in synoptic features (e.g., subtropical high and low level wind) by prohibiting the distortion of large-scale circulations, as shown in previous studies [Cha and Lee, 2009; Kang et al., 2005; Miguez-Macho et al., 2005; Tang et al., 2010]. In the ISP run with a weaker SN effect than the SP run, the mean summer features over East Asia are properly reproduced, as in the case of the SP run. The ISP run captures heavy precipitation over southern China and a strong monsoon rainband over the WNP reasonably. The statistics for the seasonal mean precipitation in the ISP run are similar to those in the SP run and are prominently better than those in the NOSP run (Table 1). This implies that intermittent SN can maintain the positive effect of reducing the systematic error in large-scale regimes despite the weak nudging effect.

image

Figure 5. Seasonal mean (JJA) precipitation (mm d−1, shading) and 850 hPa wind vector in (a) observation, (b) NOSP run, (c) SP run, and (d) ISP run. Observed precipitation and wind are from the GPCP data and R-2 reanalysis, respectively.

Download figure to PowerPoint

Table 1. Statistics of Seasonal Mean (JJA) Precipitation (mm d−1) Averaged Over the Entire Domain Except for the Buffer Zonea
 NOSPSPISP
  • a

    BIAS, RMSE, and SCC indicate seasonal mean bias, root mean square error, and spatial correlation coefficient between the simulation and the observation, respectively.

BIAS0.140.15−0.03
RMSE3.723.323.37
SCC0.570.710.68

[17] To investigate the temporal variation of systematic errors in simulated synoptic feature related to the subtropical high, we analyze the time series of the pattern correlations of 500 hPa geopotential height between simulated results and the reanalysis data during the entire simulation period (Figure 6). The NOSP run without SN has much lower pattern correlations than the other experiments, indicating larger systematic errors. In particular, the pattern correlations in the NOSP run prominently decrease during July and August, when many typhoons occurred. This result is consistent with that reported by Zhong [2006], according to which the systematic error in regional climate simulations of the 1994 East Asian summer monsoon is associated with the failure of typhoon simulations. In contrast, the SP run with pattern correlations greater than 0.95 during the entire season has smaller systematic errors due to the application of SN. The pattern correlations in the ISP run are slightly lower than that in the SP run, but the temporal variations for the two runs are almost consistent. It should be noted that SN in the ISP run is applied more frequently during July and August when the NOSP run has significantly large systematic errors. Despite the efficient effect of intermittent SN, SN is applied in only 7% (8199 steps) of the total time steps (118,080 steps) in the ISP run.

image

Figure 6. Time series (lines) of pattern correlations of 500 hPa geopotential height between reanalysis and simulations and the nudging frequency (%) per 6 h (solid bars) in the ISP run. The solid, dotted, and dot-dashed lines indicate the SP, the ISP, and the NOSP runs, respectively.

Download figure to PowerPoint

[18] To clarify the effect of intermittent SN in regional climate simulation, we analyze the similarities for large-scale regimes (longer than 1000 km wavelength) of zonal winds between the simulated results and the large-scale driving forcing (Figure 7). Similarity [von Storch et al., 2000] is defined as

  • equation image

where, Ψ(t) is the simulated RCM output field at time t, Ψ*(t) is the input forcing field, and 〈 〉 is a spatial average. The NOSP run has considerably lower similarity than the other runs; in particular, there are some periods when the upper level similarity falls below 0.5, indicating that the fidelity of the RCM with large-scale behavior to the forcing data cannot be maintained in the NOSP run. It is noteworthy that low similarity at the lower troposphere tends to vertically extend to the upper troposphere. In contrast, the SP run has relatively higher similarity at all levels due to the effect of SN as compared with the NOSP run. The upper level and low level similarities are higher than 0.9 and 0.7, respectively, during most simulation periods. The high similarity at the upper level indicates the maintenance of consistency for large-scale regimes between the model solution and the large-scale driving forcing. However, the exaggerated high similarity at the low level means that added values such as enhanced mesoscale features could be suppressed due to the strong effect of continuous SN. The ISP run has lower similarities at all levels as compared with the SP run, but the upper level similarity is still higher than 0.8. This implies that intermittent SN can increase the values added by the RCM at the low level, such as the development and intensification of mesoscale features, while retaining upper level consistency between the model solution and the large-scale driving forcing. Therefore, intermittent SN is a possible solution for maintaining the positive effects while alleviating the negative effects of SN.

image

Figure 7. Similarities (unitless) for large-scale regimes (longer than 1000 km wavelength) of the zonal wind between simulations and R-2 reanalysis averaged over the entire domain except for the buffer zone.

Download figure to PowerPoint

[19] To elucidate the efficiency of intermittent SN in improving the simulations of mesoscale atmospheric features, the total occurrences and mean intensity of simulated typhoons averaged for 4 months are calculated (Table 2). With the identifying method from Camargo et al. [2007], the NOSP run simulates 21 typhoons, which are approximately 30% more than that observed, whereas the SP run reproduces a reasonable 15 typhoons. This is because the occurrences of spurious typhoons increase in the NOSP run due to distorted synoptic features such as a weakened subtropical high and decreased anticyclonic circulation over the WNP (see Figure 5). On the contrary, generation of spurious typhoons is suppressed in the SP run because SN prevents the distortion of synoptic fields. However, the SP run reproduces markedly weaker intensities for the typhoons as compared with that observed and the NOSP run, indicating that SN has the negative effect of decreasing the realistic intensification of the simulated typhoons. The ISP run reasonably reproduces 17 typhoon occurrences, indicating that intermittent SN suppresses the generation of spurious typhoons. In addition, the intensities of the simulated typhoons in the ISP run are significantly enhanced as compared with those in the SP run. In particular, in the ISP run, 8 strong typhoons with a maximum surface wind speed of above 33 m s−1 are reasonably captured, while only 1 strong typhoon is reproduced in the SP run. The mean maximum surface wind speed in the ISP run is about 20% stronger than that in the SP run; this result is more consistent with observation. The enhanced intensities of the simulated typhoons in the ISP run are attributed to the decrease in the inhibition of the self-intensifying feedback caused by continuous SN, as shown in the sensitivity experiments to nudging interval. Therefore, intermittent SN can play a positive role in improving typhoon simulations in terms of both occurrences and intensities.

Table 2. Total Occurrences, Mean Minimum Sea Level Pressure, and Mean Maximum Surface Wind Speeds of Observed and Simulated Typhoons During the Summer (MJJA) of 1994a
 ObservationNOSPSPISP
  • a

    Values in parentheses are occurrences of strong typhoons with maximum surface wind speed above 33 m s−1. Observation data are based on the best track data set from RSMC.

Total occurrences16 (7)21 (7)15 (1)17 (8)
Mean minimum sea level pressure (hPa)967.7977.9986.8978.8
Mean maximum surface wind speed (m s−1)31.927.924.329.1

[20] We apply the spectral analysis [Castro et al., 2005; Errico, 1985] of the vertically integrated kinetic energy at the low and upper levels to clarify the improvement in added values by intermittent SN (Figure 8). The fractional change in the spectral power per wave number is calculated as

  • equation image

where S1(k) and S2(k) are the spectral powers per wave number in two runs. In this study, we compare the fractional change in the ISP (S2(k)) and SP (S1(k)) runs. The ISP run generates prominently more additional information at the low level than the upper level; it simulates more low level kinetic energy than the SP run at scales greater than 120 km (i.e., 4Δx). In particular, the kinetic energy at mesoscales shorter than 1000 km increases drastically, indicating an association with the enhanced typhoon intensity in the ISP run. In the ISP run, the additional information at the upper level is only between the 500 and 1200 km wavelengths. This is probably due to the propagation of additional low level kinetic energy to the upper level by the enhanced typhoons.

image

Figure 8. Fractional changes in the spectral power of the ISP run versus log10(k) and wavelength for the vertically integrated kinetic energy at the low level (1000–500 hPa) and upper level (500–200 hPa) with respect to the spectral power of the SP run. The vertical line indicates the wavelength of 4Δx (120 km).

Download figure to PowerPoint

5. Summary and Concluding Remarks

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model and Experiments
  5. 3. Sensitivities of Simulated Typhoons to Spectral Nudging
  6. 4. Effect of Intermittent Spectral Nudging on Seasonal Simulation
  7. 5. Summary and Concluding Remarks
  8. Acknowledgments
  9. References
  10. Supporting Information

[21] In this study, the impact of SN on values added by RCM was investigated through sensitivity experiments to the nudging interval, and intermittent SN was introduced to alleviate the negative effect of SN. In the sensitivity experiment to the nudging interval, the track distance error of the simulated typhoon tended to be reduced with shorter nudging intervals, indicating the stronger effect of SN. However, the intensities of simulated typhoons also tended to be weaker with a stronger SN effect, since SN suppressed the self-intensifying feedback of developing typhoons. That is, it was shown that SN has not only positive effect of improving regional climate simulation by reducing the systematic error resulting from model uncertainties but also the negative effect of decreasing values added by RCM, such as the intensification of typhoons and convections. To reduce the aforementioned negative effect, SN was applied intermittently only when the deviation between the large-scale driving forcing and the model solution was large. The ISP run using intermittent SN improved the simulations of strong typhoons, which are simulated as too weak in the SP run with continuous SN. In particular, intermittent SN played a role in increasing the kinetic energy at the mesoscale regimes. Therefore, intermittent SN may improve seasonal simulations over East Asia by reducing the negative effect of SN (i.e., suppressing the intensification of mesoscale features) while maintaining the positive effects of SN (i.e., decreasing the systematic errors in large-scale regimes). Intermittent SN also led to an approximately 20% improvement in computational efficiency as compared with continuous SN.

[22] SN is expected to be an effective method for the realistic reproduction of extreme climate and the prevention of climate drift in long-term simulations. However, SN may also have limitations in decreasing values added by RCMs and increasing computational time. Intermittent SN used in this study presents the possibility for the general application of SN in regional climate modeling by reducing its negative effects. Intermittent SN can be one of useful methods for alternative boundary condition in terms of dynamical downscaling, even in studies using global model data as large-scale driving forcing. However, intermittent SN still has a constraint because it is an empirical method associated with only nudging interval. There may be a number of ways to reduce the negative nudging effect. As mentioned in section 1, the effect of SN can be controlled by cutoff wavelength, nudging weighting, nudged components (i.e., rotational or nonrotational fields) and mass field nudging. In other words, the intermittent SN may be a way to improve the performance of model by reducing negative effects of SN, but there may be many others, such as using weaker nudging weighting, using different cutoff wavelength, applying nudging to rotational part of the wind. Also, the intermittent SN may not work in other cases, or can be improved further, because the effect of SN depends on simulated event, target region, and model characteristics. For example, in the low latitudes where the motion is regarded mainly as the small scale, the velocity field is more important than the mass field. Thus the cutoff wavelength of nudging should be considered carefully. Therefore, additional studies on the impact of the factors controlling SN and on the dependency of SN to simulation domain and model characteristics would be required for further improving the SN efficiency. Also, the intermittent SN should be generally improved based on statistics from much more cases. In addition to SN, the impact of a higher model resolution on the developments of mesoscale features should be examined because the 30 km grid spacing used in this study may not be sufficiently high to reproduce realistic self-intensifying feedback of mesoscale features.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model and Experiments
  5. 3. Sensitivities of Simulated Typhoons to Spectral Nudging
  6. 4. Effect of Intermittent Spectral Nudging on Seasonal Simulation
  7. 5. Summary and Concluding Remarks
  8. Acknowledgments
  9. References
  10. Supporting Information

[23] This study was supported by the Global Partnership Program (GPP) through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology. This work was funded by the Korea Meteorological Administration Research and Development Program under grant RACS_2010-2016. This study was also partially supported by the BK21 program of the Korean Government Ministry of Education.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model and Experiments
  5. 3. Sensitivities of Simulated Typhoons to Spectral Nudging
  6. 4. Effect of Intermittent Spectral Nudging on Seasonal Simulation
  7. 5. Summary and Concluding Remarks
  8. Acknowledgments
  9. References
  10. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model and Experiments
  5. 3. Sensitivities of Simulated Typhoons to Spectral Nudging
  6. 4. Effect of Intermittent Spectral Nudging on Seasonal Simulation
  7. 5. Summary and Concluding Remarks
  8. Acknowledgments
  9. References
  10. Supporting Information
FilenameFormatSizeDescription
jgrd16870-sup-0001-t01.txtplain text document0KTab-delimited Table 1.
jgrd16870-sup-0002-t02.txtplain text document1KTab-delimited Table 2.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.