Capítulo 2. Propuesta metodológica para la dirección del pensamiento táctico en
2.4. Diagnóstico en el pensamiento táctico para determinar las regularidades46
The WRF simulations of 2001 summer climate over the CONUS indicates that the dry bias in SE US summer precipitation is most likely caused by the inaccurately simulated NASH western ridge and associated circulation due to the erroneous distribution of zonal winds in the tropical oceans, i.e., errors in circulation dynamics.
2 The null hypothesis for the Hotelling’s t-‐‑square test is that the WRF simulated NASH western
ridge does not differ significantly from that in reanalysis datasets. According to the test, the null hypothesis can be rejected with a 99.99% confidence level, suggesting that the erroneous northwestward extension of the ridge is significant.
Thus, an improved simulation of large-‐‑scale circulation (especially the NASH western ridge) could potentially reduce the RCM bias in SE US summer precipitation.
To verify the importance of circulation dynamics in generating the SE US summer precipitation bias and to assess the potential improvement in precipitation simulation from an improved large-‐‑scale circulation, two sets of WRF experiments utilizing the FDDA are performed. The FDDA, i.e. interior grid nudging technique, continuously nudges the WRF simulated thermodynamic and dynamic variables towards the driving reanalysis datasets during the simulation (Stauffer and Seaman 1990). The FDDA has been widely applied in regional climate downscaling and has significantly improved climate downscaling skills over the US (Bowden et al. 2013; Lo et al. 2008; Otte et al. 2012). In this analysis, however, the application of the FDDA is not for the purpose of improving precipitation simulation skills but rather for identifying the potential sources of RCM skills in SE US summer precipitation.
The two sets of FDDA experiments are designed as follows: thermodynamic FDDA and dynamic FDDA. In the thermodynamic FDDA experiment, the temperature and specific humidity are nudged towards NCEP-‐‑R2 at each 6-‐‑hr interval during the simulation, while the wind fields are generated by WRF. In the dynamic FDDA experiment, the WRF simulated three-‐‑dimensional wind fields are nudged while the temperature and specific humidity are not. The previous experiment without an FDDA is defined as the control experiment. We run both thermodynamic and dynamic FDDA
with the four different cumulus schemes as in the control experiment. The improvement of the simulated precipitation in thermodynamic (dynamic) FDDA is attributed to the correction of atmospheric thermodynamic (dynamic) structures. Thus, by comparing the simulated precipitation in thermodynamic and dynamic FDDA with that from the control experiment, the relative importance of thermodynamic and dynamic contribution to SE US dry bias can be compared (Li et al. 2013b; Seager et al. 2010).
Figure 4.12 shows the CONUS JJA precipitation in the FDDA experiment. By correcting the WRF simulated circulation fields, the dynamic FDDA experiment substantially reduces the bias in SE US summer precipitation. In the dynamic FDDA experiment, summer precipitation increases to about 5 mm/day over the SE US domain. The domain-‐‑averaged bias is reduced to -‐‑0.3 mm day-‐‑1, indicating that about 80% of the
original dry bias in control experiment has been corrected (Figure 4.12a and c). Furthermore, the spatial distribution of precipitation, especially the southeast-‐‑northwest oriented gradient, is also reasonably simulated in the dynamic FDDA (Figure 4.12a). Thus, the dynamic FDDA experiment further verifies that the errors in wind fields generated during the WRF simulations are responsible for the SE US dry bias in the control experiment.
To confirm that the improved simulation of precipitation in the dynamic FDDA experiment results more from circulation dynamics, the effects of the thermodynamic FDDA are also compared. Generally, the thermodynamic FDDA does not improve the
simulation of SE US summer precipitation as significant as the dynamic FDDA when compared to the control experiment and observations. Specifically, in the thermodynamic FDDA experiment, the SE US dry bias is not meaningfully reduced (Figure 4.12b and d). The areal-‐‑averaged rainfall bias reaches -‐‑2.0 mm day-‐‑1 in the
thermodynamic FDDA experiment, compared to the bias of -‐‑1.3 mm day-‐‑1 in the control
experiment. In addition, the rainfall amount decreases over the coastal regions, and the spatial gradient of rainfall further weakens (Figure 4.12b). Thus, unlike the atmospheric dynamical fields, correcting the atmospheric thermodynamic fields is insufficient to generate a satisfactory skill in SE US summer precipitation simulations. More importantly, the comparison between the thermodynamic and dynamic FDDA experiments indicates that the improvements of the simulations due to the dynamic FDDA are most likely from direct dynamic contributions instead of indirect thermally driven circulation dynamics. In other words, if the thermally driven circulation (i.e. circulation component determined by atmospheric thermal structure) contributes significantly to rainfall simulation, the thermodynamic FDDA should generate similar corrective effects that the dynamic FDDA. However, since the thermodynamic FDDA fails to improve the precipitation simulations, it is the direct dynamic contributions in dynamic FDDA that provides the ultimate sources of simulation skills for SE US summer precipitation.
Figure 4. 12: 2001 JJA summer precipitation (shaded, unit: mm day-‐‑1) as simulated in a)
Thermodynamic FDDA, and b) Dynamic FDDA experiment; and the precipitation bias in c) Thermodynamic and d) dynamic FDDA. The results are shown as the average of the four
cumulus schemes.
The experiments utilizing FDDA collectively suggest that the atmospheric dynamics plays a direct and predominant role in regulating SE US summer precipitation at seasonal scales. The results from FDDA experiment are consistent with the results of SE US summer precipitation based on the regional moisture budget (Chapter 2). Over the SE US, large-‐‑scale circulation contributes to more than 90% of the variance in moisture transport for SE US summer precipitation, whereas thermodynamic (temperature and specific humidity) contribution accounts for the less than 10% (Chapter 2). The observed characteristics of SE US hydrological cycle indicate that errors in large-‐‑scale circulation could easily translate into summer precipitation bias due to its active role in atmospheric
a) Thermodynamic FDDA b) Dynamic FDDA
moisture balance. Thus, the distortion of the large-‐‑scale circulation along the NASH western ridge during the WRF simulation could result in the dry bias over the SE US (Figures. 4.9 and 4.10).
Overall, the FDDA experiments as well as the diagnostic analysis suggest that the WRF simulated dry bias in SE US summer precipitation probably originates from errors in modeled large-‐‑scale circulation. Thus, a better representation of large-‐‑scale dynamics, especially that associated with the NASH western ridge circulation, likely improve the WRF performance in simulation summertime climate over the SE US.