Evaluation of Reanalysis Products with in situ GPS Sounding Observations in the Eastern Himalayas

Department of Lower Atmosphere Observation Research (LAOR), Institute of Atmospheric Physics, Chinese Academy of Sciences,Beijing 100029, China

1 Introduction

Located in the southern region of the Tibetan Plateau(TP), the Himalayans are the highest mountain ranges in the world, with an average height of 6 km and a length of 2400 km. Because these mountains border the TP and insulate the warm, moist air in the south from the cold,dry air in the north, they form a boundary of the climate systems of the TP and those of the adjacent Indian subcontinent (Ye and Gao, 1979; Boos and Kuang, 2010).The Yarlung Tsangpo River valley, which is located in the Eastern Himalayas, is a key water vapor channel for the TP (Gao et al., 1985; Yang et al., 1987; Tian et al., 2001).In summer, the southerlies from the Bay of Bengal bring large amounts of water vapor into the hinterland of TP through this channel, which profoundly influences the synoptic and climate systems of the TP as well as the downstream area (Simmonds et al., 1999; Zhou and Yu,2005).

Due to the high altitude, severe weather, and tough environment, few in situ observations have been conducted in the TP, particularly in the high Himalayan region (Zhang et al., 2012). Therefore, reanalysis products, which include full space-time coverage, are widely used in atmospheric research of the TP. However, significant discrepancies exist among reanalysis products due to differences in physical structures, parameterization schemes,resolutions, and assimilation processes in these model systems.

Many studies have evaluated the reliabilities of the reanalysis products in the TP, and most focus on surface variables such as surface temperature, precipitation, and surface heat fluxes (Zhao and Fu, 2006a, b; Ma et al.,2008, 2009; Mao et al., 2010; Wang and Zeng, 2012; Zhu et al., 2012). Such studies have demonstrated that reanalysis products can capture the climatology of most variables in the TP but are unable to reproduce their climatic variations. In addition, the observation data that have been assimilated into reanalysis systems are sparse and unevenly distributed, which contributes to the uncertainties in these products. Moreover, because the performances of reanalysis products in the TP are regionally dependent, no single reanalysis project is appropriate for all of the statistical quantities for surface variables (Wang and Zeng, 2012). Few studies have evaluated reanalysis data at high levels in the TP because of the difficulty in obtaining sounding data in this region. Bao and Zhang(2013) compared the datasets from Climate Forecast System Reanalysis (CFSR), National Center for Environmental Prediction and the Department of Energy Reanalysis 1 (NCEP-R1), Interim European Centre for Medium Range Weather Forecasts Reanalysis (ERAInterim), and ERA 40-year Reanalysis (ERA-40) with sounding observations from 11 stations in the TP from May to August 1998. Their results showed that although the mean structures among the products were consistent with the observation, large diurnal variations were present in their performances in different regions of the TP and at various levels. Thus, reanalysis products should be carefully evaluated for specific application prior to conducting meteorological research in the TP.

In this study, we compared six widely used reanalysis products including NCEP Reanalysis 2 (NCEP-R2), NCEP Final Analysis (FNL), Japanese 25-year Reanalysis (JRA-25), ERA-Interim, Modern Era Retrospective Analysis for Research and Applications (MERRA), and CFSR with the independent sounding observations recorded in the Yarlung Tsangpo River valley in the Eastern Himalayas in June 2010. We evaluated air temperature, specific humidity, u-wind, and v-wind of these reanalysis datasets from 100 hPa to 650 hPa. This research provides a reference for studies in the Eastern Himalayas that consider reanalysis data.

2 Data and method

The Global Positioning System (GPS) sounding data were obtained from an observation campaign conducted in the Yarlung Tsangpo River valley (29.448°N, 94.691°E,2930 m) in the Eastern Himalayas. The Vaisala GPS radiosonde was operated twice daily, at 0000 UTC and 1200 UTC, from 1 June to 30 June 2010. The sounding data were collected with a time interval of 2 s; variables included height, barometric pressure, air temperature, relative humidity, wind speed, and wind direction. The observation data were linearly interpolated between 100 hPa and 700 hPa at intervals of 50 hPa.

Six reanalysis products including CFSR (Saha et al.,2010), ERA-Interim (Simmons et al., 2006), FNL(NCEP/NWS/NOAA, 2000), JRA-25 (Onogi et al., 2007),MERRA (Rienecker et al., 2011), and NCEP-R2 (Kanamitsu et al., 2002) data were evaluated in this study. It should be noted that although FNL is not a reanalysis product, it was regarded as such in this study for convenience. The horizontal resolutions are 0.5° × 0.5° for CFSR, 0.75° × 0.75° for ERA-Interim, 1.0° × 1.0° for FNL,1.25° × 1.25° for JRA-25, 0.67° × 0.50° for MERRA, and 2.5° × 2.5° for NCEP-R2. At each pressure level, the gridded data of 0000 UTC and 1200 UTC were linearly interpolated to the observation site. The surface pressures of these analyses, which were between 652 hPa and 677 hPa, were lower than the 710 hPa observation value. In this study, variables from 100 hPa to 650 hPa were analyzed. All of the products had a vertical resolution of 50 hPa between 100 hPa and 300 hPa; CFSR, ERA-Interim,MERRA, and FNL had a resolution of 50 hPa between 300 hPa and 650 hPa; and JRA-25 and NCEP-R2 had a resolution of 100 hPa between 300 hPa and 600 hPa.

Temperature (T), specific humidity (q), u-wind, and v-wind were evaluated by comparing the statistical quantities of the average, mean bias (MB), root-mean-square difference (RMSD), and correlation coefficient (R) of the reanalysis datasets with those of the observations.

3 Results

3.1 Temperature

Figure 1 a shows that all of the products effectively captured the average profile of the temperature. Large errors were present at low levels below 500 hPa, and all products except for NCEP-R2 underestimated the temperature below 550 hPa. MERRA had the largest negative MB of –3.3°C at 650 hPa (Fig. 1b). NCEP-R2 had the largest positive MB of 0.7°C at 600 hPa and the largest negative MB above 400 hPa. Although JRA-25 had the smallest MB of –0.2°C at 600 hPa, this product had a large MB of 0.8°C at 400 hPa. Throughout the column,CFSR had the smallest MB, followed by ERA-Interim.

Figure 1 Vertical profiles of the temperature statistical quantities between the six reanalysis datasets examined in this study and Global Positioning System (GPS) sounding in the Eastern Himalayas in June 2010 including (a) average, (b) mean bias (MB), (c) root-mean-square difference (RMSD), and (d) correlation coefficient (R). The reanalysis projects include Climate Forecast System Reanalysis (CFSR), Interim European Centre for Medium Range Weather Forecasts Reanalysis(ERA-Interim), Japanese 25-year Reanalysis (JRA-25), Modern Era Retrospective Analysis for Research and Applications (MERRA), National Center for Environmental Prediction and the Department of Energy Reanalysis 2 (NCEP-R2), and NCEP Final Analysis (FNL).

The reanalysis products underestimated the temperature at lower and upper levels, which is consistent with previous studies of the surface temperature (Ma et al.,2008; Wang and Zeng, 2012); however they overestimated the temperature at middle levels, which differs from that reported by Bao and Zhang (2013). In previous studies, NCEP-R2 temperature showed the largest negative MB (Wang et al., 2012; Ma et al., 2008; Zhao and Fu,2006a; Bao and Zhang, 2013). However, our result shows that NCEP-R2 overestimated the temperature between 500 hPa and 600 hPa in the Eastern Himalayas in June 2010.

The RMSD of temperature was small between 150 hPa and 500 hPa and large at low levels (Fig. 1c). MERRA had the largest RMSD at 650 hPa. NCEP-R2 had a larger RMSD at both lower and higher levels. Generally, the RMSDs of JRA-25 and ERA-Interim were the smallest throughout the column. The correlation coefficients for all products except CFSR and MERRA were larger than 0.8 below 250 hPa (Fig. 1d). The correlations increased with height and reached their maxima at 250 hPa and suddenly decreased to their minima at 200 hPa. This correlation coefficient profile differs from that reported by Bao and Zhang (2013), in which the temperature correlation coefficients between observation and CFSR and ERA-Interim increased with height and reached their maxima at 100 hPa.

3.2 Humidity

The mean profile of specific humidity was captured by the reanalysis products (Fig. 2a). The water vapor was overestimated at lower levels below 600 hPa and upper levels above 350 hPa and was underestimated at middle levels (Figs. 2a and 2b). At 650 hPa, CFSR had the largest MB at 0.67 g kg–1, and ERA-Interim had the smallest MB at 0.22 g kg–1. Negative MBs were larger at 500 hPa in NCEP-R2, JRA-25, and FNL; that of NCEP-R2 was the largest at –1.14 g kg–1. The negative MB of CFSR and ERA-Interim were the smallest. Above 300 hPa, the MBs were very small. The MB of reanalysis data for the Eastern Himalayas differs significantly from that reported by Bao and Zhang (2013), in which the RH was overestimated below 250 hPa and underestimated above 250 hPa.Throughout the column, ERA-Interim had the smallest MB for specific humidity, and CFSR has the second-smallest with the exception of a large negative MB at 450 hPa.

The profiles of RMSDs were large and variable with height below 400 hPa and decreased with height above 400 hPa (Fig. 3c). At 650 hPa, CFSR has the largest RMSD, while FNL has the smallest. Throughout the column, CFSR and ERA-Interim had the smallest RMSD whereas NCEP-R2 had the largest; the RMSDs of JRA-25 and FNL were very similar.

<1)，且各件产品是否为不合格品相互独立．

The correlation coefficients of specific humidity decreased with height from 600 hPa. This result differs from the correlation of RH in other regions of the TP, where the correlation increases with height (Bao and Zhang, 2013).The correlation of specific humidity was smaller than that of the temperature, which indicates the reanalysis products were unable to capture the variation of the water vapor as well as the temperature. All analyses had the largest correlation coefficient at 600 hPa. NCEP-R2 had the weakest relationship with the observation at nearly all levels, even showing a negative value of –0.10 at 200 hPa. The correlation coefficients of other datasets were strong below 300 hPa; above this height, the correlation of ERA-Interim, MERRA, and JRA-25 were larger.Throughout the column, ERA-Interim showed the strongest correlation. JRA-25 also had a large correlation coefficient but it had no data at 650 hPa due to its coarse vertical resolution.

Figure 2 Vertical profiles of the specific humidity between the six reanalysis datasets examined in this study (same as Fig. 1) and GPS sounding in the Eastern Himalayas in June 2010 including (a) average,(b) MB, (c) RMSD, and (d) R.

Figure 3 Vertical profiles of the u-wind between the six reanalysis datasets examined in this study (same as Fig. 1) and GPS sounding in the Eastern Himalayas in June 2010 including (a) average, (b) MB, (c)RMSD, and (d) R.

3.3 Wind

The averaged observed u-wind in June 2010 was easterly below 550 hPa and shifted to westerly above (Fig.3a). The west wind increased with height and reached its maximum at 200 hPa, then decreased with height. ERAInterim, CFSR, and FNL reproduced both the low level easterly and the high level westerly; the u-wind of ERAInterim was closest to observation. All reanalysis products captured the variations of the u-wind above 600 hPa but showed large MBs near 550 hPa (Figs. 3a and 3b). All products overestimated the westerly between 600 and 500 hPa. NCEP-R2 had the largest positive MB of 7.6 m s–1at 600 hPa and the largest negative MB of –1.6 m s–1at 200 hPa. Throughout the column, the MB of ERA-Interim was the smallest, within 1.0 m s–1. For the u-wind RMSD,ERA-Interim and NCEP-R2 the largest and smallest, respectively (Fig. 3c). The RMSDs of CFSR and FNL were close and were small below 400 hPa and large above.

The correlation coefficients of u-wind increased with height, showing small values of less than 0.5 below 500 hPa and a maximum of larger than 0.95 at 200 hPa. All reanalysis products except ERA-Interim and CFSR had negative correlation coefficients below 550 hPa; JRA-25 had the largest negative value at –0.27. The weak correlation below 500 hPa indicates that the reanalysis products were not able to capture the daily variation of the u-wind at lower levels in the Eastern Himalayas. Throughout the vertical column, the u-wind of ERA-Interim and CFSR showed the strongest correlation with the observation.

The averaged observed v-wind increased with height from 650 hPa, reached maximum at 550 hPa, then decreased with height and shifted to northerly, showing a negative value at 200 hPa, and reached a maximum at 150 hPa (Fig. 4a). CFSR, ERA-Interim, FNL, and MERRA captured the southerly wind maximum at 550 hPa, because their vertical resolutions are higher than those of JRA-25 and NCEP-R2. All reanalysis products overestimated the v-wind below 550 hPa and underestimate it between 500 hPa and 200 hPa (Fig. 4b). MERRA had the largest positive MB of 5.2 m s–1at 600 hPa, and CFSR had the largest negative MB of –3.2 m s–1at 500 hPa.Throughout the column, ERA-Interim had the smallest bias.

The v-wind RMSDs were between approximately 3.0 m s–1and 4.5 m s–1throughout the column (Fig. 4c). In general, ERA-Interim had the smallest RMSD for v-wind,and CFSR had the largest. The correlation coefficient of v-wind increased with height from 650 hPa, with small values of less than 0.5 appearing below 550 hPa (Fig. 4c).ERA-Interim, CFSR, and FNL showed negative correlation at 650 hPa. Throughout the column, ERA-Interim and JRA-25 had the closest relationship with the observation, whereas CFSR and FNL had the smallest correlation coefficients.

4 Discussion and conclusions

In this study, six reanalysis products were evaluated with independent sounding observations obtained in the Yarlung Tsangpo River valley in the Eastern Himalayas in June 2010. Our results show that the averaged vertical profiles of temperature, specific humidity, u-wind, and v-wind were captured by all reanalysis products; How-ever, large biases appeared at low levels, and each product showed different performances for different variables at different heights.

Figure 4 Vertical profiles of the v-wind between the six reanalysis datasets examined in this study (same as Fig. 1) and GPS sounding in the Eastern Himalayas in June 2010 including (a) average, (b) MB, (c)RMSD, and (d) R.

For the temperature, all products except NCEP-R2 had cold biases at lower and upper levels and warm biases at middle levels. NCEP-R2 showed an overestimation below 400 hPa and an underestimation above, which differs from that in other regions in the TP (Zhao and Fu, 2006a;Ma et al., 2008; Wang and Zeng, 2012; Bao and Zhang,2013). For specific humidity, all products had an overestimation at lower and upper levels and underestimation at middle levels, and their correlations with observations decreased with height from 600 hPa. These result differed from the RH in other region of the TP (Bao and Zhang,2013), in which RH was overestimated at the lower levels and underestimated at higher levels, and the correlations with observations increased with height. For u- and v-wind, all products had an overestimation at lower levels and underestimation at middle levels, and their correlations increased with height. In the study of Bao and Zhang (2013), CFSR and ERA-interim underestimated u-wind at all levels, whereas the MBs of v-wind were small.

Throughout the column, ERA-Interim was closest to the observations for all the four variables in the Eastern Himalayas, whereas NCEP-R2 had the largest uncertainties. In addition, CFSR showed good performance in temperature and specific humidity. Therefore, although horizontal resolution has a relationship with the performances of reanalysis products for the Eastern Himalayas, it is not a decisive factor. Many factors such as model physical parameterizations and assimilation schemes can affect the accuracy of the reanalysis products. We found that in the Eastern Himalayas, although the resolutions of CFSR and MERRA were higher than those of other products, their surface pressures had the largest negative biases. This result may reflect the difficulties in topography processing in the Himalayas and may relate to their performances in this region.

Therefore, no reanalysis product has been determined as superior to others for all variables at all heights.Therefore, users should select the appropriate datasets for their research purposes according to the performances of specific variables and related statistical quantities. Because large biases appeared at lower levels, cautions must be heeded when using these products as the initial and boundary fields for model simulations, particularly simulations of near-surface processes in the Himalayas.

Acknowledgements.This study was supported by the Special Fund for Meteorological Research in the Public Interest(GYHY2012 06041), the National Natural Science Foundation of China (Grant No. 40905067), and the Ministry of Science and Technology of the People’s Republic of China (2009CB421403).The authors would like to acknowledge the data providers including NCEP for NCEP-R2, FNL, and CFSR data; European Centre for Medium-Range Weather Forecasts (ECMWF) for ERA-Interim data; the Japan Meteorological Agency (JMA) and the Central Research Institute of Electric Power Industry (CRIEPI) for JRA-25 data; and the Global Modeling and Assimilation Office (GMAO) for MERRA data.

Bao, X., and F. Zhang, 2013: Evaluation of NCEP-CFSR, NCEPNCAR, ERA-Interim, and ERA-40 reanalysis datasets against independent sounding observations over the Tibetan Plateau, J.Climate, 26, 206–214.

Boos, W. R., and Z. Kuang, 2010: Dominant control of the South Asian monsoon by orographic insulation versus plateau heating,Nature, 463, 218–222.

Gao, D., H. Zou, and W. Wang, 1985: Influence of water vapor pass along the Yarlungzangbo River on precipitation, Mountain Res.(in Chinese), 3(4), 239–249.

Kanamitsu, M., W. Ebisuzaki, J. Woollen, et al., 2002: NCEP-DOE AMIP-II Reanalysis (R-2), Bull. Amer. Meteor. Soc., 83, 1631–1643.

Ma, L., T. Zhang, O. W. Frauenfeld, et al., 2009: Evaluation of precipitation from the ERA-40, NCEP-1, and NCEP-R2 Reanalyses and CMAP-1, CMAP-2, and GPCP-2 with ground-based measurements in China, J. Geophys. Res., 114, doi:10.1029/2008JD 011178.

Ma, L., T. Zhang, Q. Li, et al., 2008: Evaluation of ERA-40,NCEP-1, and NCEP-R2 reanalysis air temperatures with groundbased measurements in China, J. Geophys. Res., 113, doi:10.1029/2007JD009549.

Mao, J., X. Shi, L. Ma, et al., 2010: Assessment of reanalysis daily extreme temperatures with China’s homogenized historical dataset during 1979–2001 using probability density functions, J.Climate, 23, 6605–6623.

NCEP/NWS/NOAA/U.S. Department of Commerce, 2000: NCEP FNL operational model global tropospheric analyses, continuing from July 1999, available at http://rda.ucar.edu/datasets/ds083.2.

Onogi, K., and J. Tsutsui, H. Koide, et al., 2007: The JRA-25 Reanalysis, J. Meteor. Soc. Japan, 85, 369–432.

Rienecker, M. R., M. J. Suarez, R. Gelaro, et al., 2011: MERRANASA’s Modern-ERA retrospective analysis for research and applications, J. Climate, 24, 3624–3648.

Saha, S., S. Moorthi, H.-L. Pan, et al., 2010: The NCEP climate forecast system reanalysis, Bull. Amer. Meteor. Soc., 91, 1015–1057.

Simmonds, I., D. Bi, and P. Hope, 1999: Atmospheric water vapor flux and its association with rainfall over China in summer, J.Climate, 12, 1353–1367.

Simmons, A., S. Uppala, D. Dee, et al., 2006: ERA-Interim: NewECMWF reanalysis products from 1989 onwards, ECMWF News-lett., 110, 25–35.

Tian, L., V. Masson-Delmotte, M. Stievenard, et al., 2001: Tibetan Plateau summer monsoon northward extent revealed by measurements of water stable isotopes, J. Geophys. Res.,106(D22), 28081–28088.

Wang, A., and X. Zeng, 2012: Evaluation of multireanalysis products with in situ observations over the Tibetan Plateau, J.Geophys. Res., 117(D05102), doi:10.1029/2011JD016553

Yang, Y., D. Gao, and B. Li, 1987: Primremary studying water vapor pass along the downstream of Yarlung Zangbo, Sci. China Ser.B-Chem. (in Chinese), 8, 893–902.

Ye, D., and Y. Gao, 1979: Meteorology of the Qinghai-Xizang Plateau (in Chinese), CAS Qinghai-Tibet Plateau Compr. Sci.Invest. Team (Eds.), Science Press, Beijing, 279pp.

Zhang, R., T. Koike, X. Xu, et al., 2012: A China-Japan cooperative JICA atmospheric observing network over the Tibetan Plateau(JICA/Tibet project): An overviews, J. Meteor. Soc. Japan, 90,1–16.

Zhao, T., and C. Fu, 2006a: Preliminary comparison and analysis between ERA-40, NCEP-2 reanalysis and observations over China, Climate Environ. Res. (in Chinese), 11, 14–32.

Zhao, T., and C. Fu, 2006b: Comparison of products from ERA-40,NCEP-2, and CRU station data for summer precipitation over China, Adv. Atmos. Sci., 23(4), 593–604.

Zhou, T.-J., and R.-C. Yu, 2005: Atmospheric water vapor transport associated with typical anomalous summer rainfall patterns in China, J. Geophys. Res., 110, doi:10.1029/2004JD005413.

Zhu, X., Y. Liu, and G. Wu, 2012: An assessment of summer sensible heat flux on the Tibetan Plateau from eight data sets, Sci.China Ser. D-Earth Sci., 55, 779–786.