1. Introduction
The diurnal cycle is one of the most regular and basic modes in Earth’s climate system (Dai and Deser 1999; Kikuchi and Wang 2008). Over East Asia (Fig. 1), the pronounced diurnal variations of winds, convective activities, and precipitation are recognized as a key aspect of the warm-season climate. In particular, the diurnal cycle of rainfall is found to vary considerably with regions and seasons (Yu et al. 2007; Chen et al. 2009a). Nocturnal rain episodes prefer to occur at the lees of the Tibetan Plateau and propagate eastward, producing a coherent shift of diurnal phase and implying an intrinsic predictability of warm-season rainfall (Wang et al. 2004; Bao et al. 2011). Monsoon flow also displays an evident diurnal mode and becomes efficient at regulating moisture transport and rainfall systems in the night hours (Chen et al. 2009b, 2013). Morning-peak rainfall comes to dominate in the active monsoon period and undergoes distinct variations associated with the long-term monsoon activities (Yuan et al. 2010, 2013a; Yin et al. 2011). It is thus essential to improve our understanding of the diurnal cycle and related physics for the better forecast of regional climate variability. One of the major issues is using diurnal cycles as a benchmark to evaluate reanalysis data (Bao and Zhang 2013; Chung et al. 2013) and climate models (Dai 2006; Sato et al. 2009; Dirmeyer et al. 2012; Satoh and Kitao 2013; Yuan et al. 2013b).
East Asian topography and SCSMEX sounding sites over southern China.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
East Asian topography and SCSMEX sounding sites over southern China.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
East Asian topography and SCSMEX sounding sites over southern China.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Global reanalyses, a kind of long-term observationally constrained data with homogenous nature, are useful for studying the atmospheric systems. Because reanalyses have uncertainties resulting from the forecast model, data assimilation, and data source(s) used, it is very important to assess their quality in representing the weather and climate variations (e.g., Trenberth and Guillemot 1998; Hodges et al. 2011; Lin et al. 2014). A need to evaluate reanalyses in capturing diurnal cycle is highlighted, as the routine soundings used in data assimilation are basically collected twice daily (too coarse to resolve subdaily variations). For the diurnal cycle over East Asia, Chen et al. (2010) evaluated the reanalyses by using 12-hourly soundings and 3-hourly wind records at mountain sites over southern China. Chen et al. (2013) compared reanalysis data with 6-hourly soundings for the diurnal cycle of low-level winds during active monsoon days. Dai et al. (2011) verified the diurnal cycle of summer rainfall in reanalyses with rain gauge observation over eastern China. Although these evaluation efforts shed light on the quality of reanalyses at a diurnal time scale, most of them have been made either for the old-generation datasets or for limited periods/variables.
In recent years, several new reanalysis datasets have become available: the 55-yr Japanese Reanalysis Project (JRA-55), the Interim European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-Interim, herein ERA-I), the National Centers for Environmental Prediction (NCEP) Climate Forecast System Reanalysis (CFSR), and the National Aeronautics and Space Administration (NASA) Modern-Era Retrospective Analysis for Research and Applications (MERRA) (Ebita et al. 2011; Dee et al. 2011; Saha et al. 2010; Rienecker et al. 2011). These reanalyses are derived with the state-of-the-art atmospheric models and advanced data assimilation schemes, which aim to achieve an improved depiction of the weather/climate variations. More recently, Bao and Zhang (2013) assessed ERA-I, CFSR, and their predecessors against 6-hourly sounding data over the Tibetan Plateau. They found that two new reanalyses generally have a small root-mean-square (RMS) error. Both mean bias and RMS error exhibit strong diurnal variations that differ considerably among reanalyses, indicating a large uncertainty in representing diurnal variability. Over East Asia where the diurnal cycle is most evident, however, the systematic comparison of several new reanalyses is still sparse. It is unclear what differences and similarities these datasets capture the major features of the diurnal cycle over this key climate regime.
Given the importance of diurnal cycle in regional climate and the increasing use of new reanalyses, it is essential to continually assess the reanalyses to understand their uncertainties and reliability. In particular, JRA-55 is very new, so that analysis of its quality and aspects of its differences from other datasets will be informative to the community. In this study, we extensively intercompare four recent reanalyses (JRA-55, ERA-I, CFSR, and MERRA) and evaluate them by using intense soundings and satellite-derived rainfalls. The goal is to clarify their quality and efficiency in the representation of the warm-season diurnal variability over East Asia. Section 2 introduces the datasets and methods. In section 3, an intercomparison of the four reanalyses is made to identify their uncertainties as to the climate mean of the wind diurnal cycle. Section 4 evaluates the diurnal variability in reanalyses using intense rawinsonde observations. Section 5 compares the reanalysis rainfall forecasts with the satellite rainfall estimates to reveal their differences in describing the diurnal cycle of rainfall. The summary remarks are finally given in section 6.
2. Data and methods
In this study, four recent reanalyses (JRA-55, ERA-I, CFSR, and MERRA) are intercompared and evaluated. One important improvement of these new reanalyses is their high spatial resolution (Ebita et al. 2011; Dee et al. 2011; Saha et al. 2010; Rienecker et al. 2011). The horizontal resolution of the forecast models is ~60 km for JRA-55, ~79 km for ERA-I, ~38 km for CFSR, and ~50 km for MERRA. The number of the model levels ranges from 60 to 72. To produce analysis, recent reanalyses also apply modern techniques to assimilate a wide range of data sources including conventional and satellite measurements. JRA-55 and ERA-I employ four-dimensional variational data assimilation (4DVAR) with variational bias correction for satellite radiances (Ebita et al. 2011; Dee et al. 2011). CFSR and MERRA apply three-dimensional variational data assimilation (3DVAR) based on gridpoint statistical interpolation (GSI), with flow dependence for background error variances (Wu et al. 2002; Purser et al. 2003; Saha et al. 2010; Rienecker et al. 2011). The advanced data assimilation, model physics, and increase of resolution help to reduce the bias/error in recent reanalyses and improve the forecast skills. For more details on the configuration and general performance of these new reanalyses, see the references listed above.
In this study, we use the 6-hourly pressure-level analysis products of ERA-I, CFSR, and MERRA at a full resolution, whereas the early release version of JRA-55 at a reduced resolution (1.25° longitude/latitude) is used. All these products have about 37 pressure levels, with an interval of 25 hPa below 750 hPa. They thus offer a good resolution of the variables in the lower troposphere where a strong diurnal cycle is usually observed. All four reanalyses cover the satellite era from 1979 to present, and JRA-55 even provides an outstanding long archive back to 1958. On the other hand, the hourly (3 hourly) rainfall forecasts in CFSR and MERRA (ERA-I and JRA-55) are used and their quality in representing the diurnal cycle of warm-season rainfall is assessed.
To evaluate reanalyses, we use the 6-hourly rawinsonde observation data from 22 sites over southern China (Fig. 1), which were collected during the intensive observing period (IOP) of the South China Sea Monsoon Experiment (SCSMEX) from 1 May to 30 June 1998 (Lau et al. 2000). These quality-controlled data include the wind speed, wind direction, temperature, and dewpoint temperature. Wind speed and direction are applied to derive the zonal and meridional wind components. Relative humidity is derived using the formula as in the observational preprocess of ERA-I. As these sounding sites are part of the routine observation network, their 12-hourly Global Telecommunication System (GTS) records are used in all four reanalyses. Meanwhile, the 6-hourly offline SCSMEX dataset is assimilated into JRA-55 but not into other three reanalyses. Therefore, the intense SCSMEX soundings offer us a rare opportunity to estimate what degree the reanalyses, with different assimilation schemes and data source input, capture the diurnal cycle. To make comparisons, reanalyses are interpolated to each sounding site with a distance-reversed weightiness of the nearest four grids. The match-up of reanalyses and soundings is made for any pressure level and any synoptic hour when both data sources are available.
To illustrate the ability of reanalyses to reproduce rainfall diurnal cycle, we compare the forecast rainfall of reanalyses with the observed rainfall. We use the satellite-derived rainfall from the 3B42 product (version 7) of the Tropical Rainfall Measurement Mission (TRMM). The 3B42 offers the 3-hourly rainfall estimate with a resolution of 0.25° longitude/latitude (Huffman et al. 2007). It provides important data to detect the diurnal cycle of rainfall globally (Kikuchi and Wang 2008). It also performs well over East Asia, except for an underestimate of some fraction of morning rainfall in the low-lying areas (Chen et al. 2012; Yuan et al. 2012). It offers full coverage of East Asia including remote lands and oceans where rain gauges are sparse or unavailable. For consistency, all rainfall data of 3B42 and reanalyses are matched to the same grid points (1.25° longitude/latitude) as for JRA-55.
Over East Asia, the warm-season climate exhibits an evident subseasonal change associated with the progress of the summer monsoon (Ding 1992). The rainband appears over southern/central China and southern Japan in presummer (May–June); it proceeds to northern China, the Korean Peninsula, and the main islands of Japan in midsummer (July–August). Diurnal variations of wind and precipitation undergo a corresponding change along with the summer progress (Yuan et al. 2010; Chen et al. 2012, 2013). To depict the summer progress, we composite the monthly mean of diurnal cycle, and then group the data into those of presummer and midsummer. Such a division of the warm season allows us to clarify whether reanalyses capture the summer progress of the diurnal variations of winds and precipitation. Note that, over East Asia (local time LT = UTC + 7–9 h), the synoptic hours of 0000, 0600, 1200, and 1800 UTC denote the morning, afternoon, evening, and late night, respectively.
3. Intercomparison of reanalyses for mean diurnal cycle
The intercomparison of multiple reanalyses is one important way to identify the uncertainties (e.g., Hodges et al. 2011). Over East Asia, we examine the difference and similarity of new reanalyses in representing the climatology of diurnal cycle, with emphasis on the horizontal winds because of their key role in regional climate (e.g., Dai and Deser 1999; Chen et al. 2013). First, we estimate the mean diurnal cycle of horizontal winds for each reanalysis. We then compare the diurnal phase/amplitude, spatial pattern, vertical structures, and summer progress among four reanalyses.
Figure 2 show the mean diurnal variation of horizontal winds at 850 hPa in four reanalyses. The daily mean, mostly southerly or southwesterly (Ding 1992; Chen et al. 2013), has been removed at each grid. The diurnally deviated wind vectors generally undergo a clockwise rotation in both presummer and midsummer. At the low latitudes south of 30°N, the wind vectors display as northerly at 0600 UTC, easterly at 1200 UTC, southerly at 1800 UTC, and westerly at 0000 UTC. The wind vectors rotate relatively fast at the middle latitudes, being southwesterly at 1800 UTC and northerly at 0000 UTC. Such diurnal veering and latitudinal difference are consistently seen in Fig. 2, suggesting that the four reanalyses present similar diurnal phases of wind oscillations.
Climate mean of the diurnal variations of horizontal winds at 850 hPa in presummer (May–June) and midsummer (July–August). The diurnal amplitude (shaded) is defined by a speed difference between 1800–0000 UTC and 0600–1200 UTC. The climate mean is estimated by the average of 1979–2009 from CFSR and of 1979–2012 from other three reanalyses.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Climate mean of the diurnal variations of horizontal winds at 850 hPa in presummer (May–June) and midsummer (July–August). The diurnal amplitude (shaded) is defined by a speed difference between 1800–0000 UTC and 0600–1200 UTC. The climate mean is estimated by the average of 1979–2009 from CFSR and of 1979–2012 from other three reanalyses.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Climate mean of the diurnal variations of horizontal winds at 850 hPa in presummer (May–June) and midsummer (July–August). The diurnal amplitude (shaded) is defined by a speed difference between 1800–0000 UTC and 0600–1200 UTC. The climate mean is estimated by the average of 1979–2009 from CFSR and of 1979–2012 from other three reanalyses.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
The diurnal amplitude can be expressed as a speed difference of wind vectors between 1800–0000 UTC and 0600–1200 UTC. In presummer (Figs. 2a–d), diurnal amplitude is evident over the East Asian continent, especially southern China, the eastern regions of the Indochinese Peninsula, and the periphery of the Tibetan Plateau. These correspond to the preferred locations of the diurnally varying low-level jets (e.g., Monaghan et al. 2010). The diurnal amplitude is relatively small over adjacent seas and less visible over open oceans. Although the four reanalyses present similar spatial patterns of diurnal amplitude, they exhibit a large difference in magnitude. MERRA has the largest amplitude over southern China, with a maximum of 2–2.5 m s−1 centered at 26°N, 110°E. CFSR and JRA-55 seem to give the next-largest amplitude (~2 m s−1) over southern China, while ERA-I presents the smallest amplitude (1.5–2 m s−1).
In midsummer (Figs. 2e–h), all four analyses express an increase of diurnal amplitude from presummer. The most evident is an extension of large amplitude over central eastern China, with a maximum at 30°N, 118°E. This is related to a northward march of the monsoon flow from presummer to midsummer (Chen et al. 2013). The reanalyses are thus consistent in depicting a summer progress of wind diurnal cycle, associated with the summer monsoon. However, we still see that the magnitude of diurnal amplitude differs greatly among the reanalyses, as in presummer.
To illustrate vertical structures, we examine the longitude–pressure sections of diurnal wind components in the reanalyses. One section is made for southern China (Fig. 3), where the largest diurnal amplitude occurs in presummer. Another is made for central China (Fig. 4), where the diurnal amplitude strengthens significantly in midsummer. As indicated by the diurnal veering in Fig. 2, we pay emphasis on the major diurnal components (i.e., zonal wind at 1200 UTC and meridional wind at 1800 UTC). At 1200 UTC in presummer (Figs. 3a–d), the four reanalyses show a dominant easterly wind at lower levels over southern China. The largest amplitude of 1.5–2 m s−1 is seen at the eastern slope (~105°E) of the Yun-Gui Plateau, while the secondary center of 0.75–1 m s−1 is established at the coast (~122°E), suggesting a strong topographic dependence. In general, the four reanalyses present a low-level easterly of comparable magnitude (about 1.3 m s−1 at 850 hPa) and vertical extent over southern China. At upper levels, they all report a westerly component. The low-level easterly and upper westerly (return flow) together are recognized as a continental-scale “sea breeze” circulation (Dai and Deser 1999; Huang et al. 2010; Chen et al. 2013). The high similarity in Figs. 3a–d thus suggests that four recent reanalyses are very consistent in representing the large-scale solenoidal circulation over East Asia.
Diurnal components of horizontal winds averaged at a zone of 25°–27.5°N in presummer: (left) diurnal component of zonal wind at 1200 UTC and (right) that of meridional wind at 1800 UTC.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Diurnal components of horizontal winds averaged at a zone of 25°–27.5°N in presummer: (left) diurnal component of zonal wind at 1200 UTC and (right) that of meridional wind at 1800 UTC.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Diurnal components of horizontal winds averaged at a zone of 25°–27.5°N in presummer: (left) diurnal component of zonal wind at 1200 UTC and (right) that of meridional wind at 1800 UTC.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
As in Fig. 3, but for the average at a zone of 30°–31.5°N in midsummer.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
As in Fig. 3, but for the average at a zone of 30°–31.5°N in midsummer.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
As in Fig. 3, but for the average at a zone of 30°–31.5°N in midsummer.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Figures 3e–h show the diurnally deviated meridional wind at 1800 UTC. The four reanalyses represent a low-level southerly and an upper northerly at 105°–118°E. This low-level southerly is collocated with an easterly at 1200 UTC, while the upper northerly is preceded by a westerly (cf. Figs. 3e–h and 3a,b). MERRA has the largest amplitude of low-level southerly wind and ERA-I reports the smallest. The averaged amplitude at 850 hPa over southern China is 0.7, 0.9, 1.1, and 1.5 m s−1 in ERA-I, JRA-55, CFSR, and MERRA, respectively. At the coast near 122°E, CFSR presents the strongest low-level southerly, and ERA-I reports the weakest one. JRA-55 has a moderate amplitude to east of 105°E, but gives the largest amplitude of midlevel southerly over the Yun-Gui Plateau near 100°E. It seems clear that a considerable difference among the reanalyses occurs in the diurnal amplitude of nocturnal southerly. Because the southerly wind is in phase with summer monsoon, different reanalyses appear to report different strengths of the diurnal cycle of monsoon flow at different regions.
Figure 4 shows the vertical structures of wind diurnal cycle over central China in midsummer. For the zonal wind at 1200 UTC, four reanalyses present an extensive easterly below 500 hPa (Figs. 4a–d). There are three regional maxima of easterlies at 100°–102°E, 107°–110°E, and 112°–121°E. The western maximum is seen at 600–500 hPa, and the middle (eastern) one is located at ~800 hPa (~850 hPa), as seen consistently in four reanalyses. These regional maxima also have a comparable magnitude of 1–1.5 m s−1 in four reanalyses.
For the meridional wind at 1800 UTC (Figs. 4e–h), the four reanalyses present a similar pattern of southerly winds in the lower troposphere, except for slight discrepancies in the locations of regional maxima. Clearly, the amplitude of the low-level southerly differs greatly among reanalyses. The mean amplitude at 850 hPa at 106°–120°E is 0.7, 1.0, 1.2, and 1.7 m s−1 in ERA-I, JRA-55, CFSR, and MERRA, respectively. The southerly wind in ERA-I also has the lowest vertical extent. At upper levels, a northerly wind is seen in four reanalyses, with a relatively large diurnal amplitude in CFSR and MERRA. Similar to the situation in presummer, the reanalyses seem to have a large uncertainty in representing the mean diurnal amplitude of meridional wind in midsummer.
4. Evaluation of reanalyses using intense soundings
a. Overall bias and RMS error
To further clarify the quality, reanalyses are evaluated with intense soundings during the IOP of SCSMEX. We examine their overall performance before assessing the quality at diurnal time scale. Figure 5 shows the mean value, bias, and RMS error averaged over 22 sites of southern China. During the two-month period, mean zonal wind (U) increases with height, from ~−1 m s−1 at 1000 hPa to ~15 m s−1 at 200 hPa (Fig. 5a). The meridional wind (V) exhibits as a southerly at mid to lower levels with a peak of ~4 m s−1 at 850 hPa, and a northerly at upper levels (Fig. 5d). A low-level southerly/southwesterly wind and an upper northwesterly wind are seen consistently in reanalyses and soundings. It suggests that reanalyses are reliable for describing the regional mean climate at seasonal scale, as also noted in Bao and Zhang (2013).
Vertical profiles of U, V, T, and RH averaged over 22 sounding sites over southern China during May–June 1998: (top) the mean values for the SCSMEX soundings and matched reanalyses, and (middle),(bottom) the mean bias and RMS error, respectively, of the matched reanalyses against soundings.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Vertical profiles of U, V, T, and RH averaged over 22 sounding sites over southern China during May–June 1998: (top) the mean values for the SCSMEX soundings and matched reanalyses, and (middle),(bottom) the mean bias and RMS error, respectively, of the matched reanalyses against soundings.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Vertical profiles of U, V, T, and RH averaged over 22 sounding sites over southern China during May–June 1998: (top) the mean values for the SCSMEX soundings and matched reanalyses, and (middle),(bottom) the mean bias and RMS error, respectively, of the matched reanalyses against soundings.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
The bias of horizontal winds of reanalyses against verifying soundings is plotted in Figs. 5b and 5e. The U wind generally has a small negative bias within −0.5 m s−1, except at lower levels in ERA-I and MERRA (Fig. 5b). The bias of V wind is also rather small and mostly within ±0.5 m s−1 in the troposphere above surface (Fig. 5e). At lower levels, the bias profiles diverge largely between reanalyses, with a positive (negative) bias in ERA-I and MERRA (JRA-55 and CFSR). All four reanalyses are nearly free from bias for the V wind at the middle levels. They generally present a positive bias at upper levels, with a very small bias in JRA-55 and ERA-I.
Figure 5c shows that the RMS error of U wind in four reanalyses is about 2–2.5 m s−1 at mid to lower levels. The RMS error increases with height, especially at upper levels. The differences among the reanalyses become evident at 200 hPa, with the smallest error (2.5 m s−1) in JRA-55 and the largest error (3.5 m s−1) in MERRA. Among four reanalyses, JRA-55 has the smallest RMS error throughout the vertical column. The CFSR has the second smallest error at lower levels, while ERA-I does so at mid to upper levels. Figure 5f shows that the RMS error of V is as small as 2–2.5 m s−1 below 700 hPa in four reanalyses, except near the surface in MERRA. The RMS error increases at mid to upper levels and reaches 2.8–3.7 m s−1 at 200 hPa. Broadly speaking, JRA-55 has the smallest RMS error, while ERA-I follows with the second smallest. CFSR is comparable to ERA-I at lower levels, but it gives a relatively large RMS error at mid to upper levels as in MERRA. We also see that the RMS error of U and V winds is smaller than that over the Tibetan Plateau, where the error is about 3–4.5 m s−1 in ERA-I and CFSR (Bao and Zhang 2013). It is expected that the relatively dense routine soundings over southern China have helped in reducing the error in reanalyses.
Besides horizontal winds, other available variables (temperature and humidity) are also evaluated for a comprehensive image of reanalysis quality. Figure 5g shows that mean temperature (T) decreases with height, with a relatively large lapse at upper levels. Figure 5h shows that the four reanalyses have a cold bias through the troposphere, except for MERRA near the surface. The bias is mostly within −0.5°C at lower levels and becomes about −1.5°C at upper levels. The cold bias is most evident in CFSR among the reanalyses. Figure 5i shows that the RMS error of T is ~1.2°C near the surface and decreases to ~0.9°C at 850 hPa, with less difference among reanalyses. The RMS error remains stable or increases slowly at middle levels. It increases relatively fast at upper levels and reaches 1.6°–2.2°C at 200 hPa. The largest error is seen in CFSR, while the smallest one is seen in JRA-55. A comparison of Figs. 5h and 5i suggests that the systematic cold bias is the major cause of the RMS error of temperature.
Figure 5j shows that relative humidity (RH) decreases with height, from ~85% near the surface to ~45% at 200 hPa. For the RH bias (Fig. 5k), JRA-55 has a dry bias except near the surface, most evident at 700 and 300 hPa. It corresponds to an underestimate of specific humidity, with the largest value −0.6 g kg−1 at 700 hPa (not shown). The three other reanalyses generally have a small wet bias, reflecting the combined result of a cold bias and a small positive bias of specific humidity. Figure 5l shows that the RMS error in reanalyses usually increases with height, from ~7% at 1000 hPa to ~16% at 200 hPa. The RMS error is smaller in JRA-55 and ERA-I than that in CFSR and MERRA. Again, we see that the RMS error of RH over southern China is smaller than that over the Tibetan Plateau (20%–32%) reported by Bao and Zhang (2013).
Figure 6 shows the correlation between soundings and reanalysis matched data. The correlation coefficient for U and V winds is ~0.70 near surface and increases to ~0.90 at 850 hPa. It decreases slightly from 850 to 500 hPa, but returns above 0.90 at upper levels. The correlation coefficient for T remains high above 0.85 in the troposphere, with the maxima of ~0.93 at 850 and 300 hPa. The correlation for RH is 0.60–0.80 at lower levels, increasing to 0.80–0.85 at 400–500 hPa, and decreasing to 0.40–0.60 at 200 hPa. JRA-55 has the highest correlation with soundings for four variables. The other three reanalyses display similar profiles of the correlation with soundings for U, V, and T. Clearly, ERA-I gives the second highest correlation with soundings for RH. In general, the rank of the reanalyses for correlation with soundings is consistent with that of the RMS error; that is, the smaller the RMS error the reanalysis has, the higher the correlation with soundings it exhibits. We also note that the reanalysis with smaller RMS error usually tends to have a smaller relative error (not shown).
As in Fig. 5, but for the correlation coefficient between soundings and matched reanalyses.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
As in Fig. 5, but for the correlation coefficient between soundings and matched reanalyses.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
As in Fig. 5, but for the correlation coefficient between soundings and matched reanalyses.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
b. Diurnal variation of bias and its impact on mean diurnal cycle
Figure 7 shows the mean diurnal cycles of U, V, T, and RH, with the daily mean removed at each level. Figure 7a shows that, in soundings, the zonal wind exhibits as an easterly at 1200 UTC and a westerly at 0000–0600 UTC at lower levels. Figures 7b–e show that reanalyses generally present a mean diurnal cycle similar to soundings. The 6-hourly averaged bias of reanalyses against soundings is overlapped as contours. The four reanalyses tend to give a positive bias at 1800 UTC at lower levels. In JRA-55 and ERA-I, the negative bias is clearly seen at 1200 UTC and in phase with mean diurnal cycle, which results in a diurnal amplitude slightly larger than sounding observations. In CFSR and MERRA, the negative bias is most visible at 0600 UTC, which reduces the diurnal component at 0600 UTC but has less effect on those at 0000 and 1200 UTC.
The 6-hourly variations of U, V, T, and RH in the SCSMEX soundings and matched reanalyses, averaged over 22 sites over southern China (marked in Fig. 1) during May–June 1998. The shaded denotes the mean diurnal cycle with daily mean removed. The contour stands for the bias of reanalyses against soundings, with an interval of 0.25 m s−1 for U and V, 0.25 K for T, and 2% for RH.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
The 6-hourly variations of U, V, T, and RH in the SCSMEX soundings and matched reanalyses, averaged over 22 sites over southern China (marked in Fig. 1) during May–June 1998. The shaded denotes the mean diurnal cycle with daily mean removed. The contour stands for the bias of reanalyses against soundings, with an interval of 0.25 m s−1 for U and V, 0.25 K for T, and 2% for RH.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
The 6-hourly variations of U, V, T, and RH in the SCSMEX soundings and matched reanalyses, averaged over 22 sites over southern China (marked in Fig. 1) during May–June 1998. The shaded denotes the mean diurnal cycle with daily mean removed. The contour stands for the bias of reanalyses against soundings, with an interval of 0.25 m s−1 for U and V, 0.25 K for T, and 2% for RH.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Figure 7f portrays that the V wind in soundings reaches its maximum at 1800–0000 UTC and minimum at 0600–1200 UTC at lower levels. Figures 7g–j show that the low-level southerly at 1800 UTC in JRA-55 and ERA-I (CFSR and MERRA) is some weaker (stronger) than that in soundings. The amplitude difference is very similar to that of climate mean in presummer (Fig. 3). Figure 7g shows that the underestimate of mean diurnal cycle in JRA-55 is due to a negative bias at 1800–0000 UTC. The small diurnal cycle in ERA-I instead results from a positive bias at 1200 UTC (Fig. 7h). In contrast, MERRA tends to overestimate the diurnal amplitude at lower levels, due to a positive bias at 1800 UTC (Fig. 7j). The CFSR also gives a diurnal amplitude larger than the sounding observation, but because of a negative bias at 0600 UTC (Fig. 7i). In summary, we see that the diurnal variation of V wind bias differs greatly among the four reanalyses, which explains the difference in mean diurnal amplitude.
Figure 7k shows that the mean T exhibits as a warm phase at 0600–1200 UTC and a cold phase at 1800–0000 UTC. The diurnal cycle is evident through the troposphere, with the largest amplitude at lower levels and near the surface. Figures 7l–o show that all four reanalyses present a mean diurnal cycle very similar to the sounding observation. In JRA-55, the cold bias at lower levels is most evident at 1800 UTC and in phase with the mean diurnal cycle, enhancing the mean diurnal amplitude (Fig. 7l). In ERA-I, near the surface there is a cold bias at 0600–1200 UTC and a warm bias at 1800 UTC, reducing the mean diurnal amplitude (Fig. 7m). In CFSR and MERRA, the diurnal variation of bias at lower levels is quite small (Figs. 7n,o) and thus it has less impact on the mean diurnal amplitude.
Figure 7p shows that the RH reaches a diurnal peak at 1800–0000 UTC at mid to lower levels with T diurnal minimum, whereas its peak appears at 1200 UTC at upper levels, likely due to a moistening of daytime convection. In JRA-55, the dry bias at 700 and 300 hPa mainly occurs at 1200 UTC (Fig. 7q), giving a delayed mean diurnal cycle. The wet bias near the surface has a weak diurnal variation and less impact on mean diurnal cycle. In ERA-I, the wet bias is relatively large at 0600 UTC in the troposphere (Fig. 7r), which offsets the mean diurnal cycle. In CFSR, the wet bias is smallest at 1800 UTC and reduces the mean diurnal cycle at lower levels (Fig. 7s). In MERRA, an evident oscillation of bias is seen at about 400 hPa and strongly disturbs the mean diurnal cycle (Fig. 7t).
c. Diurnal variation of RMS error and its impact on the varying diurnal amplitude
Figure 8 shows the 6-hourly variations of RMS error for reanalyses. In JRA-55, the diurnal variations of RMS error for U, V, and T are very weak (Figs. 8a,e,i). The RMS error profiles of ERA-I are very similar to those of JRA-55, with a small diurnal variation (Figs. 8b,f, j). The error profiles of CFSR and MERRA, as shown in the third and fourth rows of Fig. 8, however, differ significantly from those in JRA-55 and ERA-I. They exhibit an evident diurnal variation, with a small error at 1200 and 0000 UTC and a relatively large error at 0600 and 1800 UTC. The RMS error for RH is shown in Figs. 8m–p. JRA-55 has a relatively large error at mid to upper levels at 1200 UTC (Fig. 8m). In contrast, the other three reanalyses (ERA-I, CFSR, and MERRA) have a larger error at 0600 and 1800 UTC than that at 0000 and 1200 UTC (Figs. 8n–p).
The 6-hourly variations of the RMS error of U, V, T, and RH in the matched reanalyses against the SCSMEX soundings, averaged over sounding 22 sites over southern China during May–June 1998.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
The 6-hourly variations of the RMS error of U, V, T, and RH in the matched reanalyses against the SCSMEX soundings, averaged over sounding 22 sites over southern China during May–June 1998.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
The 6-hourly variations of the RMS error of U, V, T, and RH in the matched reanalyses against the SCSMEX soundings, averaged over sounding 22 sites over southern China during May–June 1998.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
The diurnal variations in data quality that differ among reanalyses are also seen over the Tibetan Plateau (Bao and Zhang 2013). It is still unclear to what extent such phenomena affect the representation of diurnal cycle. As the RMS error relates to a capture of variation, its diurnal cycle indicates a possible difference of reanalyses in capturing the diurnal amplitude at different hours. Here, we compare reanalyses to the verifying soundings, with respect to the diurnal ranges of four variables that vary in 22 sites for 61 days of the SCSMEX IOP. As shown by the mean diurnal cycle in Fig. 7, the diurnal range of U wind can be defined as the difference between 0000 and 1200 UTC, while the diurnal range of V, T, and RH is estimated by the difference between 1800 and 0600 UTC.
Figure 9 shows all available records of the diurnal range at 850 hPa derived from reanalyses against soundings. The diurnal range of U wind in the four reanalyses exhibits a good association with sounding observations, with a high correlation coefficient of 0.73–0.8 (Figs. 9a–d). The correlation coefficient is slightly higher in CFSR and MERRA than that in JRA-55 and ERA-I, which corresponds to a small RMS error at 850 hPa at 0000 and 1200 UTC. Nevertheless, the reanalyses are shown to represent well the diurnal amplitude of the solenoidal circulation that varies in sites/days.
Diurnal difference of U, V, T, and RH at 850 hPa in the matched reanalyses against the SCSMEX soundings, for 22 sites and all days during May–June 1998. The correlation coefficient is marked at the left upper corner of each figure. The diurnal difference of U is defined as the difference between 1200 and 0000 UTC next day, while that of V, T, and RH is defined as the difference between 0600 and 1800 UTC.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Diurnal difference of U, V, T, and RH at 850 hPa in the matched reanalyses against the SCSMEX soundings, for 22 sites and all days during May–June 1998. The correlation coefficient is marked at the left upper corner of each figure. The diurnal difference of U is defined as the difference between 1200 and 0000 UTC next day, while that of V, T, and RH is defined as the difference between 0600 and 1800 UTC.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Diurnal difference of U, V, T, and RH at 850 hPa in the matched reanalyses against the SCSMEX soundings, for 22 sites and all days during May–June 1998. The correlation coefficient is marked at the left upper corner of each figure. The diurnal difference of U is defined as the difference between 1200 and 0000 UTC next day, while that of V, T, and RH is defined as the difference between 0600 and 1800 UTC.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
For the diurnal range of V wind between 0600 and 1800 UTC, the good relation with soundings is kept in JRA-55 and ERA-I (Figs. 9e,f). The correlation coefficient of CFSR/MERRA with soundings, however, decreases to ~0.6 (Figs. 9g,h). The correlation with soundings for the diurnal range of T and RH is also somewhat lower in CFSR and MERRA than in JRA-55 and ERA-I (cf. Figs. 9k,l and 9i,j). These differences correspond to the strong diurnal variations of the RMS error in CFSR and MERRA (Figs. 8g,h, 8k,l, and 8o,p). The quality decay at 0600 and 1800 UTC in CFSR and MERRA (as indicated by the relatively large RMS error) is thus responsible for the decline in the capture of the diurnal range of V, T, and RH.
Figure 10 shows the correlation between reanalyses and soundings for the diurnal range at different levels. The correlation coefficient for U wind diurnal range is higher than 0.7 above 900 hPa (Fig. 10a), and that for V wind is also higher than 0.5 above the surface (Fig. 10b); both are above the 99.9% confidence level. All four reanalyses seem somewhat reliable for representing the temporal/spatial variations of wind diurnal amplitude through the troposphere. In particular, they may help depict the diurnally varying low-level jet that has a strong link to the high-impact extreme heavy rainfalls over East Asia (Monaghan et al. 2010; Chen et al. 2014). It is also shown in Fig. 10b that the correlation with soundings differs among reanalyses for V. Similar features also appear in the correlation profiles for T and RH (Figs. 10c,d). In general, JRA-55 exhibits the highest correlation with soundings, and ERA-I follows as the second highest except near the surface. They thus offer an excellent capture of the varying diurnal amplitude over southern China in a two-month period. As a comparison, the correlation of CSFR and MERRA with soundings is some lower for V, T, and RH in the troposphere above the surface (Figs. 10b–d), which corresponds to the relatively large RMS error at 0600 and 1800 UTC in these two reanalyses (Figs. 8g,h, 8k,l, and 8o,p). The results suggest that the diurnal cycle of RMS error greatly affects a representation of the temporal and spatial variations of diurnal amplitude.
Correlation coefficient between reanalyses and soundings for the diurnal amplitude of U, V, T, and RH at different vertical levels.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Correlation coefficient between reanalyses and soundings for the diurnal amplitude of U, V, T, and RH at different vertical levels.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Correlation coefficient between reanalyses and soundings for the diurnal amplitude of U, V, T, and RH at different vertical levels.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
As horizontal winds are the assimilated variables, their analysis error may arise from the uncertainties of assimilation scheme and data source. As shown in Fig. 8, the RMS error is small at 0000 and 1200 UTC when routine soundings are available, indicating that data assimilation schemes perform reasonably well in four reanalyses. In contrast, a large difference of RMS error is seen at 0600 and 1800 UTC. It is clear that this difference largely occurs between JRA-55/ERA-I and CFSR/MERRA. A possible cause is these two reanalysis groups that implement different concepts of data assimilation (4DVAR versus 3DVAR), which may involve the different handlings of the variable consistency for the hours (0600 and 1800 UTC) with less routine sounding input. Another cause is that JRA-55 further assimilates the offline SCSMEX soundings at 0600 and 1800 UTC. It explains that a combination of advanced assimilation scheme 4DVAR and intense soundings usage in JRA-55 contributes to the smallest RMS error in these two hours and the best capture of diurnal cycle variations during SCSMEX IOP. More studies are need for clarification of the detailed impacts of data assimilation and observation network on wind diurnal cycle. Such efforts, while beyond the scope of this study, would help improve the representation of diurnal cycle in the future generation of reanalyses.
5. Evaluation of rainfall diurnal cycle in reanalyses using satellite observation
Precipitation is one of the key variables, which is evaluated to depict the ability of reanalyses and climate models in representing the hydrological cycle (e.g., Randall et al. 1991; Trenberth and Guillemot 1998; Dai 2006). An early validation effort of Dai et al. (2011) shows that several old-generation reanalyses have a deficiency in reproducing the diurnal cycle of summer rainfall over East Asia. Here, we address whether four recent reanalyses capture the spatial pattern, regional contrast, summer progress, and interannual/interdecadal variability of the rainfall diurnal cycle.
Over East Asia, rainfall often reaches a peak from afternoon to evening over the elevated lands and from late night to morning over the valleys and oceans (Yu et al. 2007; Chen et al. 2009a; Yuan et al. 2010; Mao and Wu 2012). To facilitate validation, rainfall diurnal difference (Rd) is estimated by the rainfall of afternoon to evening (0400–1500 UTC) minus that of late night to morning (1600–0300 UTC), expressed as a percentage of daily rainfall amount [i.e., Rd = 100(ΣR04–15 − ΣR16–03)/ΣR00–23]. Similar to Yuan et al. (2013b), the diurnal component at any hour (Rc) can be defined as the rain rate (R) normalized by daily mean rain rate (Rm) [i.e., Rc = (R − Rm)/Rm].
a. Spatial patterns of rainfall diurnal difference
Figure 11a shows that, in presummer, the afternoon-hour rainfall (Rd > 0, shaded in red) is generally observed over the Tibetan Plateau, the highlands of northern China, and the tropical coastal lands. The morning-hour rainfall (Rd < 0, shaded in blue) is seen over the oceans, the valleys at the plateau peripheries, and the low-lying plains. Figures 11b–e show that similar patterns of rainfall diurnal difference appear in four reanalyses. The large-scale modes relating to land–sea contrast and major terrains can be clearly identified in the reanalyses. Their spatial correlation coefficient with 3B42 is as high as 0.65–0.80, above a significance level at 99.9% confidence. The correlation coefficient is also higher than that between old-generation reanalyses and rain gauge observation reported by Dai et al. (2011). It suggests that the new reanalyses capture well the broad pattern of rainfall diurnal difference over East Asia. This may benefit from a good representation of the large-scale dynamics (such as the continental sea-breeze circulation in Figs. 2 and 3) and terrain thermal contrast that regulate the spatial pattern of rainfall diurnal cycle over East Asia (Huang et al. 2010; Yuan et al. 2012).
Diurnal difference of rainfall in 3B42 and reanalyses during (left) presummer and (right) midsummer of 1998–2012. Dashes outline the elevations of 750, 1500, and 3000 m. See the definition of diurnal difference in the text.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Diurnal difference of rainfall in 3B42 and reanalyses during (left) presummer and (right) midsummer of 1998–2012. Dashes outline the elevations of 750, 1500, and 3000 m. See the definition of diurnal difference in the text.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Diurnal difference of rainfall in 3B42 and reanalyses during (left) presummer and (right) midsummer of 1998–2012. Dashes outline the elevations of 750, 1500, and 3000 m. See the definition of diurnal difference in the text.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
In Figs. 11a–e, we also see that reanalyses exhibit discrepancies in the diurnal amplitude at regional scale. For most lands with dominant afternoon rainfall, JRA-55 has a diurnal difference of ~50% (Fig. 11b), which is comparable to that observed in 3B42. The locations of regional maxima also match fairly with those in 3B42. ERA-I and CFSR reproduce well the afternoon-hour rainfall on the tropical lands, but they give the small diurnal difference (~20%) over the Tibetan Plateau and the highlands of northern China (Figs. 11c,d). MERRA presents an extensive afternoon-hour rainfall on land (Fig. 11e). The regional maxima in MERRA, compared to those in 3B42, are some smooth in distribution and have a relatively small amplitude (~30%) over the Indochinese Peninsula and eastern China.
At the eastern lee of the Tibetan Plateau, a local maximum of morning rainfall, with a diurnal difference of −50%, is observed in 3B42 (Fig. 11a). Correspondingly, the four reanalyses, especially ERA-I and MERRA, tend to report a weak one of about −25% (Figs. 11b–e). Over the east China plain (27°–36.5°N, 113.5°–120°E), where afternoon and morning rainfalls are comparable, CFSR tends to present more morning rainfall (Fig. 11d); JRA-55 and MERRA reproduce more afternoon rainfall (Figs. 11b,e). Over oceans, CFSR and MERRA present an extensive morning rainfall with a diurnal difference of about −10% as in 3B42 (Figs. 11d,e), whereas JRA-55 and ERA-I seem to give a weaker morning rainfall over western Pacific (Figs. 11b,c).
Figure 11f shows that, in midsummer, 3B42 reports an evident increase of the afternoon-hour rainfall over southeast contiguous China, except some morning-hour rainfall remaining over the Sichuan Basin (~30°N, ~105°E). In contrast, morning-hour rainfall strengthens over the adjacent seas of northern China, the Korean Peninsula, and Japan. In Figs. 11g–j, it seems that reanalyses present well both the increasing afternoon rainfall over southeastern China and the increasing morning rainfall over midlatitude oceans. This northward shift of morning rainfall has been connected with a progress of summer monsoon over East Asia (Chen et al. 2009a, 2013). It thus suggests that new reanalyses generally capture well the summer progress of rainfall diurnal cycle.
In midsummer, we still see a clear difference between reanalyses at regional scale (Figs. 11g–j). For instance, the afternoon-hour rainfall over the Tibetan Plateau is more obvious in JRA-55 and MERRA than that in ERA-I and CFSR. The morning rainfall over the Sichuan Basin is apparent in JRA-55, quite weak in CFSR, and almost absent in ERA-I and MERRA. These discrepancies indicate that some cautions should be taken for the data application at regional scale.
It is recognized that the rainfall diurnal cycle may vary from year to year, due to the climate variations such as monsoon activities (Yuan et al. 2010, 2013a; Chen et al. 2012, 2013). The quality of reanalyses in individual years can be illustrated by the spatial correlation between reanalyses and 3B42 for the rainfall diurnal difference over East Asia. Figure 12 shows that correlation coefficient ranges from 0.35 to 0.75 during 1998–2012, above the 99.9% confidence level. For most years, JRA-55 has the highest correlation with 3B42, giving the best capture of the pattern of rainfall diurnal cycle. The other three reanalyses have a comparable value in presummer, while MERRA exhibits the second highest correlation with 3B42 in midsummer. Interestingly, the correlations between 3B42 and reanalyses generally increase as the years progress, indicating an improved capture of the pattern of rainfall diurnal cycle in recent years.
Spatial correlation of the rainfall diurnal difference over East Asia between reanalyses and 3B42 for the individual years in 1998–2012.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Spatial correlation of the rainfall diurnal difference over East Asia between reanalyses and 3B42 for the individual years in 1998–2012.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Spatial correlation of the rainfall diurnal difference over East Asia between reanalyses and 3B42 for the individual years in 1998–2012.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
b. Longitudinal difference of rainfall diurnal cycle
Over East Asia, rainfall is characterized by an eastward delayed diurnal phase (Chen et al. 2010, 2012; Bao et al. 2011; Yuan et al. 2012). To show a capture of this key feature, we examine the longitudinal variation of rainfall diurnal component (Rc). Figure 13a shows that, in 3B42, rainfall with a peak at 1200 UTC dominates the Tibetan Plateau (~100°E). JRA-55 reproduces this early-evening rainfall in terms of both diurnal peak and amplitude (Fig. 13b). ERA-I and CFSR instead present a weak afternoon rainfall over the plateau (Figs. 13c,d). The MERRA gives an afternoon rainfall with a peak at 0700 UTC (Fig. 13e), somewhat earlier than in 3B42. At the plateau lee, 3B42 reports a coherent shifting of diurnal peak from ~100°E at 1200 UTC to ~110°E at 0000 UTC the next day (Fig. 13a). It is impressive to see that such an eastward shift of diurnal phase is well captured by JRA-55 (Fig. 13b). The other three reanalyses also present a propagation feature (Figs. 13c–e), but with an amplitude smaller than that in 3B42 and JRA-55.
Normalized diurnal cycle of rainfall derived from 3B42 and reanalyses, averaged in a zone of 27.5–32.5°N. See the definition of normalized diurnal cycle in the text.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Normalized diurnal cycle of rainfall derived from 3B42 and reanalyses, averaged in a zone of 27.5–32.5°N. See the definition of normalized diurnal cycle in the text.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Normalized diurnal cycle of rainfall derived from 3B42 and reanalyses, averaged in a zone of 27.5–32.5°N. See the definition of normalized diurnal cycle in the text.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Over eastern China (110°–120°E), 3B42 rainfall has double peaks, one at morning and another at afternoon (Fig. 13a). ERA-I and CFSR roughly reproduce the signature of double peaks (Figs. 13c,d), whereas JRA-55 and MERRA present the afternoon peak and miss the morning peak (Figs. 13b,e). In particular, MERRA gives a very strong early-afternoon peak (Fig. 13e), which is distinct from 3B42 and other three reanalyses. Over ocean to the east of 122°E where morning rainfall is dominant in 3B42 (Fig. 13a), the four reanalyses generally capture the diurnal peak time and amplitude (Figs. 13b–e). ERA-I tends to report an additional peak at early afternoon (Fig. 13c).
In midsummer, 3B42 reports a weak propagation feature at the plateau lee and a strong afternoon rainfall over eastern China (Fig. 13f). JRA-55 reproduces well the early-evening rainfall over the Tibetan Plateau and the weak propagation at the plateau lee (Fig. 13g). The other three reanalyses, however, fail to capture the propagation feature (Figs. 13h–j). As in presummer, ERA-I and CFSR give a weak afternoon rainfall over the plateau (Figs. 13h,i), while MERRA presents an early-afternoon rainfall (Fig. 13j). Clearly, four reanalyses capture the increasing afternoon rainfall over eastern China and the dominant morning rainfall over oceans (Figs. 13g–j). Consequently, they represent well the strong contrast of rainfall diurnal cycle at the coastline at ~122°E.
Figures 11 and 13 show that considerable differences occur among the reanalyses for the diurnal variation at regional scale. It is supportive to a common understanding that global climate models have deficiencies in the simulation of rainfall diurnal cycle (e.g., Dai 2006; Lee et al. 2007). The deficiency generally involves the various model physics including resolution, parameterization, and depiction of mesoscale circulation related to diurnal cycle propagation (Sato et al. 2009; Dirmeyer et al. 2012; Satoh and Kitao 2013; Yuan et al. 2013b). The forecast models of four reanalyses have a spatial resolution ranging from 38 to 79 km, and three of them (JRA-55, CFSR, and MERRA) implement the Arakawa–Schubert convection parameterization but diverge in detailed settings. It is speculated that model resolution and tuning of convection scheme may affect the diurnal cycle of rainfall in these reanalyses. On the other hand, the eastward shift of diurnal cycle over East Asia is often linked to the propagating rainfall episodes that are supported by mountain–valley circulation (Bao et al. 2011), nocturnal low-level jets, and related moist processes (Chen et al. 2014). An improved representation of these regional forcings in JRA-55 (as indicated in Figs. 9 and 10) thus may contribute, in part, to the good capture of diurnal cycle propagation. As rainfall episodes occur in limited days, case evaluations at regional scale may provide further insights. Nevertheless, hope remains for simulating the diurnal cycle of rainfall with a high-resolution model and improved representations of convection and mesoscale processes (Sato et al. 2009; Dirmeyer et al. 2012; Satoh and Kitao 2013; Yuan et al. 2013b; Chen et al. 2014).
c. Long-term variations of rainfall diurnal cycle over the east China plain
It is well known that the summer rainfall over the east China plain has undergone an increasing trend over the past four decades. Yuan et al. (2013a) found that this long-term trend is clearly accompanied by an increasing morning rainfall percentage, indicating an important role of the diurnal cycle in climate change. Here, we examine whether new reanalyses capture the long-term variations of the diurnal cycle over this key region. Figure 14 shows the average morning percentage of summer rainfall over the east China plain. As shown in Fig. 14a, although the reanalyses have different mean values, they present the variations of morning rainfall as highly similar to 3B42. In particular, JRA-55 almost follows 3B42 in recent decades. The correlation coefficient with 3B42 is as high as 0.73 in JRA-55, 0.82 in ERA-I, 0.59 in CFSR, and 0.84 in MERRA. Therefore, four reanalyses represent well the interannual variability of the rainfall diurnal cycle over the east China plain.
Morning percentage of summer rainfall averaged over east China plain during (a) 1998–2012 and (b) 1958–2012. The dashed lines in (b) denote the linear trend during 1966–2005 for JRA-55 and during 1979–2005 for other three reanalyses.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Morning percentage of summer rainfall averaged over east China plain during (a) 1998–2012 and (b) 1958–2012. The dashed lines in (b) denote the linear trend during 1966–2005 for JRA-55 and during 1979–2005 for other three reanalyses.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Morning percentage of summer rainfall averaged over east China plain during (a) 1998–2012 and (b) 1958–2012. The dashed lines in (b) denote the linear trend during 1966–2005 for JRA-55 and during 1979–2005 for other three reanalyses.
Citation: Journal of Climate 27, 14; 10.1175/JCLI-D-14-00005.1
Figure 14b shows an extended view of the morning rainfall percentage for all available years, back to 1958 for JRA-55 and to 1979 for other three reanalyses. The series exhibit similar variance and highlight a good consistency among the reanalyses in representing interannual variability. During 1979–2005, morning rainfall undergoes an evident interdecadal increase in CFSR (0.92% decade−1) and MERRA (1.24% decade−1), while the increase is less visible in ERA-I. The long-archive JRA-55 further identifies that the interdecadal increase originates from 1966, with a ratio of 1.14% decade−1. It is striking that this long-term trend is in good agreement with rain gauge analysis, which reports an increase of ~4% in the past 40 years (Yuan et al. 2013a).
In Fig. 14, it is encouraging to see that most reanalyses reproduce well both the year-to-year variation and the long-term trend of the morning rainfall percentage, despite some discrepancies in mean value. As we know, the diurnal cycle of rainfall over China is greatly modulated by the climate state as shown by variables such as summer monsoon activity (Chen et al. 2009b, 2013; Yuan et al. 2010). The long-term increase of the morning rainfall over the east China plain, along with a decrease in northern China, is also linked to a weakening summer monsoon in the past decades that has a southward location (Yin et al. 2011; Yuan et al. 2013a). As the reanalyses are some reliable in representing the monsoon variability (Lin et al. 2014), they may serve as a reasonable depiction of the climate forcings that regulate the diurnal variation of rainfall systems. Taken together, an implication of Fig. 14 is the possibility that new reanalyses, especially the long-archive JRA-55, provide a good opportunity for studying the climate variations and associated diurnal cycle over East Asia.
6. Conclusions
In this study, four recent reanalyses (JRA-55, ERA-I, CFSR, and MERRA) have been compared to identify their uncertainties in the representation of the warm-season diurnal cycle over East Asia. It is found that all four reanalyses present similar spatial patterns and summer progress of the mean diurnal cycle in winds. They are very consistent in describing the diurnal establishment/reversal and structures of large-scale breeze circulation. However, the differences among the reanalyses are clearly seen in the diurnal amplitude of the low-level meridional wind, indicating a large uncertainty in representing the nocturnal speeding up (and/or daytime slowing down) of monsoon flow.
The reanalyses are evaluated with the intense soundings data during May–June 1998. All four reanalyses are shown to have a small mean bias and RMS error over southern China, compared to the other regions with sparse routine soundings. The differences between reanalyses in this period are similar to those in the climate mean. The reanalyses exhibit strong diurnal variations in the mean bias that differ among them. JRA-55 has a negative bias of meridional wind at 1800 UTC, and ERA-I displays a positive bias at 0600 UTC. Both are out of phase with the mean diurnal cycle, resulting in a small mean diurnal amplitude. In contrast, MERRA gives a positive bias at 1800 UTC and CFSR has a negative at 0600 UTC, overestimating the mean diurnal amplitude. On the other hand, the RMS error exhibits a strong diurnal variation in CFSR and MERRA, with a maximum (minimum) at 0600 and 1800 UTC (0000 and 1200 UTC). This seems to decline a capture of the spatiotemporal variations of diurnal difference at 0600 and 1800 UTC. As a comparison, the diurnal variation of the RMS error is less evident in JRA-55 and ERA-I. Correspondingly, JRA-55 and ERA-I are good at capturing the individual records of diurnal amplitude that vary in sites/days.
In this study, we reveal that the differences among reanalyses clearly indicate the uncertainties in the representation of diurnal cycle, consistent with recent studies (Chung et al. 2013; Bao and Zhang 2013). We further clarify that the diurnal variation of mean bias mainly explains the difference of mean diurnal amplitude among the four reanalyses, while the diurnal variation of RMS error greatly affects the capture of the temporal/spatial variations of diurnal amplitude. These findings strongly suggest that the different preferences of reanalyses at diurnal time scale and related impacts should be a concern when using reanalyses to describe the wind diurnal cycle over East Asia.
Compared with satellite rainfall estimate, the four reanalyses are found to reproduce well both the broad pattern (related to large-scale terrains and land–sea contrast) and summer progress of rainfall diurnal cycle over East Asia. In particular, JRA-55 gives a good capture of the early-evening rainfall over the Tibetan Plateau and the coherent eastward shift of diurnal phase over the eastern lowlands. The four reanalyses are also consistent in reproducing the interannual variability of rainfall diurnal cycle over East Asia. Three reanalyses (JRA-55, CFSR, and MERRA) even present an interdecadal increase of morning rainfall over the east China plain, agreeing well with rain gauge observation. Our assessments suggest that new reanalyses are potentially useful for studying the response of diurnal cycle of hydrological processes to the climate change over East Asia. As indicated here and in previous studies, further investigations of the role of diurnal cycle in climate system are warranted to provide more insights into both the initiation/evolution mechanism of precipitation systems and the dynamics of climate variability. Finally, as new reanalyses still have obvious uncertainty at regional scale, caution should be taken when applying them for specific regions.
Acknowledgments
The authors wish to acknowledge the help of three anonymous reviewers and the technical support of JRA-55 project staffs. They thank JMA, ECMWF, NCEP, and NASA for providing reanalysis datasets JRA-55, ERA-Interim, CFSR, and MERRA. Thanks also go to NASA GSFC for providing satellite rainfall and Department of Atmospheric Science at Colorado State University for providing the SCSMEX dataset. This study was partly supported by the Strategic Programs for Innovative Research (field 3, proposal number hp120282) funded by the Ministry of Education, Culture, Sports, Science and Technology (MEXT) in Japan.
REFERENCES
Bao, X., and F. Zhang, 2013: Evaluation of NCEP–CFSR, NCEP–NCAR, ERA-Interim, and ERA-40 reanalysis datasets against independent sounding observations over the Tibetan Plateau. J. Climate, 26, 206–214, doi:10.1175/JCLI-D-12-00056.1.
Bao, X., F. Zhang, and J. Sun, 2011: Diurnal variations of warm-season precipitation east of the Tibetan Plateau over China. Mon. Wea. Rev., 139, 2790–2810, doi:10.1175/MWR-D-11-00006.1.
Chen, G., W. Sha, and T. Iwasaki, 2009a: Diurnal variation of precipitation over southeastern China: Spatial distribution and its seasonality. J. Geophys. Res., 114, D13103, doi:10.1029/2008JD011103.
Chen, G., W. Sha, and T. Iwasaki, 2009b: Diurnal variation of precipitation over southeastern China: 2. Impact of the diurnal monsoon variability. J. Geophys. Res., 114, D21105, doi:10.1029/2009JD012181.
Chen, G., W. Sha, T. Iwasaki, and K. Ueno, 2012: Diurnal variation of rainfall in the Yangtze River Valley during the spring–summer transition from TRMM measurements. J. Geophys. Res., 117, D06106, doi:10.1029/2011JD017056.
Chen, G., W. Sha, M. Sawada, and T. Iwasaki, 2013: Influence of summer monsoon diurnal cycle on moisture transport and precipitation over eastern China. J. Geophys. Res., 118, 3163–3177, doi:10.1002/jgrd.50337.
Chen, G., R. Yoshida, W. Sha, T. Iwasaki, and H. Qin, 2014: Convective instability associated with the eastward-propagating rainfall episodes over eastern China during the warm season. J. Climate, 27, 2331–2339, doi:10.1175/JCLI-D-13-00443.1.
Chen, H., R. Yu, J. Li, W. Yuan, and T. Zhou, 2010: Why nocturnal long-duration rainfall presents an eastward-delayed diurnal phase of rainfall down the Yangtze River valley. J. Climate, 23, 905–917, doi:10.1175/2009JCLI3187.1.
Chung, E.-S., B. J. Soden, B. J. Sohn, and J. Schmetz, 2013: An assessment of the diurnal variation of upper tropospheric humidity in reanalysis data sets. J. Geophys. Res., 118, 3425–3430, doi:10.1002/jgrd.50345.
Dai, A., 2006: Precipitation characteristics in eighteen coupled climate models. J. Climate, 19, 4605–4630, doi:10.1175/JCLI3884.1.
Dai, A., and C. Deser, 1999: Diurnal and semidiurnal variations in global surface wind and divergence fields. J. Geophys. Res., 104, 31 109–31 125, doi:10.1029/1999JD900927.
Dai, Z., R. Yu, J. Li, and H. Chen, 2011: The characteristics of summer precipitation diurnal variations in three reanalysis datasets over China (in Chinese). Meteor. Monthly, 37, 21–30, doi:10.7519/j.issn.1000-0526.2011.1.003.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, doi:10.1002/qj.828.
Ding, Y., 1992: Summer monsoon rainfalls in China. J. Meteor. Soc. Japan, 70, 373–396.
Dirmeyer, P. A., and Coauthors, 2012: Simulating the diurnal cycle of rainfall in global climate models: Resolution versus parameterization. Climate Dyn., 39, 399–418, doi:10.1007/s00382-011-1127-9.
Ebita, A., and Coauthors, 2011: The Japanese 55-Year Reanalysis “JRA-55”: An interim report. SOLA, 7, 149–152, doi:10.2151/sola.2011-038.
Hodges, K. I., R. W. Lee, and L. Bengtsson, 2011: A comparison of extratropical cyclones in recent reanalyses ERA-Interim, NASA MERRA, NCEP CFSR, and JRA-25. J. Climate, 24, 4888–4906, doi:10.1175/2011JCLI4097.1.
Huang, W.-R., J. C. L. Chan, and S.-Y. Wang, 2010: A planetary-scale land–sea breeze circulation in East Asia and the western North Pacific. Quart. J. Roy. Meteor. Soc., 136, 1543–1553, doi:10.1002/qj.663.
Huffman, G. J., and Coauthors, 2007: The TRMM Multisatellite Precipitation Analysis: Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales. J. Hydrometeor., 8, 38–55, doi:10.1175/JHM560.1.
Kikuchi, K., and B. Wang, 2008: Diurnal precipitation regimes in the global tropics. J. Climate, 21, 2680–2696, doi:10.1175/2007JCLI2051.1.
Lau, K. M., and Coauthors, 2000: A report of the field operations and early results of the South China Sea Monsoon Experiment (SCSMEX). Bull. Amer. Meteor. Soc., 81, 1261–1270, doi:10.1175/1520-0477(2000)081<1261:AROTFO>2.3.CO;2.
Lee, M.-I., and Coauthors, 2007: An analysis of the warm-season diurnal cycle over the continental United States and northern Mexico in general circulation models. J. Hydrometeor., 8, 344–366, doi:10.1175/JHM581.1.
Lin, R., T. Zhou, and Y. Qian, 2014: Evaluation of global monsoon precipitation changes based on five reanalysis datasets. J. Climate, 27, 1271–1289, doi:10.1175/JCLI-D-13-00215.1.
Mao, J., and G. Wu, 2012: Diurnal variations of summer precipitation over the Asian monsoon region as revealed by TRMM satellite data. Sci. China Earth Sci., 55, 554–566, doi:10.1007/s11430-011-4315-x.
Monaghan, A. J., D. L. Rife, J. O. Pinto, C. A. Davis, and J. R. Hannan, 2010: Global precipitation extremes associated with diurnally varying low-level jets. J. Climate, 23, 5065–5084, doi:10.1175/2010JCLI3515.1.
Purser, R. J., W.-S. Wu, D. F. Parrish, and N. M. Roberts, 2003: Numerical aspects of the application of recursive filters to variational statistical analysis. Part I: Spatially homogeneous and isotropic Gaussian covariances. Mon. Wea. Rev., 131, 1524–1535, doi:10.1175//1520-0493(2003)131<1524:NAOTAO>2.0.CO;2.
Randall, D. A., Harshvardhan, and D. A. Dazlich, 1991: Diurnal variability of the hydrologic cycle in a general circulation model. J. Atmos. Sci., 48, 40–62, doi:10.1175/1520-0469(1991)048<0040:DVOTHC>2.0.CO;2.
Rienecker, M. M., and Coauthors, 2011: MERRA: NASA’s Modern-Era Retrospective Analysis for Research and Applications. J. Climate, 24, 3624–3648, doi:10.1175/JCLI-D-11-00015.1.
Saha, S., and Coauthors, 2010: The NCEP Climate Forecast System Reanalysis. Bull. Amer. Meteor. Soc., 91, 1015–1057, doi:10.1175/2010BAMS3001.1.
Sato, T., H. Miura, M. Satoh, Y. N. Takayabu, and Y. Wang, 2009: Diurnal cycle of precipitation in the tropics simulated in a global cloud-resolving model. J. Climate, 22, 4809–4826, doi:10.1175/2009JCLI2890.1.
Satoh, M., and Y. Kitao, 2013: Numerical examination of the diurnal variation of summer precipitation over southern China. SOLA, 9, 129–133, doi:10.2151/sola.2013-029.
Trenberth, K. E., and C. J. Guillemot, 1998: Evaluation of the atmospheric moisture and hydrological cycle in the NCEP/NCAR reanalyses. Climate Dyn., 14, 213–231, doi:10.1007/s003820050219.
Wang, C.-C., G. T.-J. Chen, and R. E. Carbone, 2004: A climatology of warm-season cloud patterns over East Asia based on GMS infrared brightness temperature observations. Mon. Wea. Rev., 132, 1606–1629, doi:10.1175/1520-0493(2004)132<1606:ACOWCP>2.0.CO;2.
Wu, W.-S., R. J. Purser, and D. F. Parrish, 2002: Three-dimensional variational analysis with spatially inhomogeneous covariances. Mon. Wea. Rev., 130, 2905–2916, doi:10.1175/1520-0493(2002)130<2905:TDVAWS>2.0.CO;2.
Yin, S., G. Gao, W. Li, D. Chen, and L. Hao, 2011: Long-term precipitation change by hourly data in Haihe River Basin during 1961–2004. Sci. China Earth Sci., 54, 1576–1585, doi:10.1007/s11430-011-4232-z.
Yu, R., T. Zhou, A. Xiong, Y. Zhu, and J. Li, 2007: Diurnal variations of summer precipitation over contiguous China. Geophys. Res. Lett., 34, L01704, doi:10.1029/2006GL028129.
Yuan, W., R. Yu, H. Chen, J. Li, and M. Zhang, 2010: Subseasonal characteristics of diurnal variation in summer monsoon rainfall over central eastern China. J. Climate, 23, 6684–6695, doi:10.1175/2010JCLI3805.1.
Yuan, W., R. Yu, M. Zhang, W. Lin, H. Chen, and J. Li, 2012: Regimes of diurnal variation of summer rainfall over subtropical East Asia. J. Climate, 25, 3307–3320, doi:10.1175/JCLI-D-11-00288.1.
Yuan, W., R. Yu, and J. Li, 2013a: Changes in the diurnal cycles of precipitation over eastern China in the past 40 years. Adv. Atmos. Sci., 30, 461–467, doi:10.1007/s00376-012-2092-x.
Yuan, W., R. Yu, M. Zhang, W. Lin, J. Li, and Y. Fu, 2013b: Diurnal cycle of summer precipitation over subtropical East Asia in CAM5. J. Climate, 26, 3159–3172, doi:10.1175/JCLI-D-12-00119.1.
