• Arakawa, A., 2004: The cumulus parameterization problem: Past, present, and future. J. Climate, 17, 24932525, doi:10.1175/1520-0442(2004)017<2493:RATCPP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Argüeso, D., J. M. Hidalgo-Muñoz, S. R. Gámiz-Fortis, M. J. Esteban-Parra, and Y. Castro-Díez, 2012: Evaluation of WRF mean and extreme precipitation over Spain: Present climate (1970–99). J. Climate, 25, 48834897, doi:10.1175/JCLI-D-11-00276.1.

    • Search Google Scholar
    • Export Citation
  • Barnston, A. G., and M. K. Tippett, 2013: Predictions of Nino3.4 SST in CFSv1 and CFSv2: A diagnostic comparison. Climate Dyn., 41, 16151633, doi:10.1007/s00382-013-1845-2.

    • Search Google Scholar
    • Export Citation
  • Bauer, P., A. Thorpe, and G. Brunet, 2015: The quiet revolution of numerical weather prediction. Nature, 525, 4755, doi:10.1038/nature14956.

    • Search Google Scholar
    • Export Citation
  • Brisson, E., K. Van Weverberg, M. Demuzere, A. Devis, S. Saeed, and M. Stengel, and N. P. M. van Lipzig, 2016: How well can a convection-permitting climate model reproduce decadal statistics of precipitation, temperature and cloud characteristics? Climate Dyn., 47, 30433061, doi:10.1007/s00382-016-3012-z.

    • Search Google Scholar
    • Export Citation
  • Caldwell, P., H.-N. S. Chin, D. C. Bader, and G. Bala, 2009: Evaluation of a WRF dynamical downscaling simulation over California. Climatic Change, 95, 499521, doi:10.1007/s10584-009-9583-5.

    • Search Google Scholar
    • Export Citation
  • Chen, H., S. L. Guo, C. Y. Xu, and V. P. Singh, 2007: Historical temporal trends of hydroclimatic variables and runoff response to climate variability and their relevance in water resource management in the Hanjiang basin. J. Hydrol., 344, 171184, doi:10.1016/j.jhydrol.2007.06.034.

    • Search Google Scholar
    • Export Citation
  • Cretat, J., B. Pohl, Y. Richard, and P. Drobinski, 2012: Uncertainties in simulating regional climate of southern Africa: Sensitivity to physical parameterizations using WRF. Climate Dyn., 38, 613634, doi:10.1007/s00382-011-1055-8.

    • Search Google Scholar
    • Export Citation
  • Cui, J.-X., S.-Z. Xu, R.-Q. Wang, and H.-Y. Zhang, 2007: A study on meteorological and hydrological characteristics of autumnal flood in the Hanjiang River in 2005. Torrential Rain Disasters, 26, 289294.

    • Search Google Scholar
    • Export Citation
  • Di Luca, A., R. D. Elia, and R. Laprise, 2012: Potential for added value in precipitation simulated by high-resolution nested regional climate models and observations. Climate Dyn., 38, 12291247, doi:10.1007/s00382-011-1068-3.

    • Search Google Scholar
    • Export Citation
  • Dirmeyer, P. A., and Coauthors, 2012: Simulating the diurnal cycle of rainfall in global climate models: Resolution versus parameterization. Climate Dyn., 39, 399418, doi:10.1007/s00382-011-1127-9.

    • Search Google Scholar
    • Export Citation
  • Dudhia, J., 1989: Numerical study of convection observed during the Winter Monsoon Experiment using a mesoscale two-dimensional model. J. Atmos. Sci., 46, 30773107, doi:10.1175/1520-0469(1989)046<3077:NSOCOD>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Ek, M. B., K. E. Mitchell, Y. Lin, E. Rogers, P. Grunmann, V. Koren, G. Gayno, and J. D. Tarpley, 2003: Implementation of Noah land surface model advances in the National Centers for Environmental Prediction operational mesoscale Eta model. J. Geophys. Res., 108, 8851, doi:10.1029/2002JD003296.

    • Search Google Scholar
    • Export Citation
  • Fonseca, R. M., T. Zhang, and K. T. Yong, 2015: Improved simulation of precipitation in the tropics using a modified BMJ scheme in the WRF Model. Geosci. Model Dev., 8, 29152928, doi:10.5194/gmd-8-2915-2015.

    • Search Google Scholar
    • Export Citation
  • Grell, G. A., 1993: Prognostic evaluation of assumptions used by cumulus parameterizations. Mon. Wea. Rev., 121, 764787, doi:10.1175/1520-0493(1993)121<0764:PEOAUB>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Hong, S. Y., Y. Noh, and J. Dudhia, 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 23182341, doi:10.1175/MWR3199.1.

    • Search Google Scholar
    • Export Citation
  • Hu, Z.-Z., A. Kumar, B. Huang, W. Wang, J. Zhu, and C. Wen, 2013: Prediction skill of monthly SST in the North Atlantic Ocean in NCEP Climate Forecast System version 2. Climate Dyn., 40, 27452759, doi:10.1007/s00382-012-1431-z.

    • Search Google Scholar
    • Export Citation
  • Janjić, Z. I., 2000: Comments on “Development and evaluation of a convection scheme for use in climate models.” J. Atmos. Sci., 57, 3686, doi:10.1175/1520-0469(2000)057<3686:CODAEO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Jiang, X. W., S. Yang, Y. Q. Li, A. Kumar, X. W. Liu, Z. Y. Zuo, and B. Jha, 2013: Seasonal-to-interannual prediction of the Asian summer monsoon in the NCEP Climate Forecast System version 2. J. Climate, 26, 37083727, doi:10.1175/JCLI-D-12-00437.1.

    • Search Google Scholar
    • Export Citation
  • Kain, J. S., 2004: The Kain–Fritsch convective parameterization: An update. J. Appl. Meteor., 43, 170181, doi:10.1175/1520-0450(2004)043<0170:TKCPAU>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Krishnamurti, T. N., and J. Sanjay, 2003: A new approach to the cumulus parameterization issue. Tellus, 55A, 275300, doi:10.1034/j.1600-0870.2003.00021.x.

    • Search Google Scholar
    • Export Citation
  • Lang, S., W. K. Tao, J. Simpson, and B. Ferrier, 2003: Modeling of convective–stratiform precipitation processes: Sensitivity to partition methods. J. Appl. Meteor., 42, 505527, doi:10.1175/1520-0450(2003)042<0505:MOCSPP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Lang, Y., and Coauthors, 2014: Evaluating skill of seasonal precipitation and temperature predictions of NCEP CFSv2 forecasts over 17 hydroclimatic regions in China. J. Hydrometeor., 15, 15461559, doi:10.1175/JHM-D-13-0208.1.

    • Search Google Scholar
    • Export Citation
  • Liang, X. Z., L. Li, A. Dai, and K. E. Kunkel, 2004: Regional climate model simulation of summer precipitation diurnal cycle over the United States. Geophys. Res. Lett., 31, L24208, doi:10.1029/2004GL021054.

    • Search Google Scholar
    • Export Citation
  • Liang, X. Z., M. Xu, K. E. Kunkel, G. A. Grell, and J. S. Kain, 2007: Regional climate model simulation of U.S.–Mexico summer precipitation using the optimal ensemble of two cumulus parameterizations. J. Climate, 20, 52015207, doi:10.1175/JCLI4306.1.

    • Search Google Scholar
    • Export Citation
  • Liu, H., D.-L. Zhang, and B. Wang, 2008: Daily to submonthly weather and climate characteristics of the summer 1998 extreme rainfall over the Yangtze River basin. J. Geophys. Res., 113, D22101, doi:10.1029/2008JD010072.

    • Search Google Scholar
    • Export Citation
  • Luo, L., W. Tang, Z. Lin, and E. F. Wood, 2013: Evaluation of summer temperature and precipitation predictions from NCEP CFSv2 retrospective forecast over China. Climate Dyn., 41, 22132230, doi:10.1007/s00382-013-1927-1.

    • Search Google Scholar
    • Export Citation
  • Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A. Clough, 1997: Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102, 16 66316 682, doi:10.1029/97JD00237.

    • Search Google Scholar
    • Export Citation
  • Moncrieff, M. W., and M. J. Miller, 1976: The dynamics and simulation of tropical cumulonimbus and squall lines. Quart. J. Roy. Meteor. Soc., 102, 373394, doi:10.1002/qj.49710243208.

    • Search Google Scholar
    • Export Citation
  • Park, D. S. R., C. H. Ho, and J. H. Kim, 2014: Growing threat of intense tropical cyclones to East Asia over the period 1977–2010. Environ. Res. Lett., 9, 014008, doi:10.1088/1748-9326/9/1/014008.

    • Search Google Scholar
    • Export Citation
  • Pei, L. S., N. Moore, S. Y. Zhong, L. F. Luo, D. W. Hyndman, W. E. Heilman, and Z. Q. Gao, 2014: WRF Model sensitivity to land surface model and cumulus parameterization under short-term climate extremes over the southern Great Plains of the United States. J. Climate, 27, 77037724, doi:10.1175/JCLI-D-14-00015.1.

    • Search Google Scholar
    • Export Citation
  • Pérez, J. C., J. P. Díaz, A. González, J. Expósito, F. Rivera-López, and D. Taima, 2014: Evaluation of WRF parameterizations for dynamical downscaling in the Canary Islands. J. Climate, 27, 56115631, doi:10.1175/JCLI-D-13-00458.1.

    • Search Google Scholar
    • Export Citation
  • Pohl, B., and H. Douville, 2011: Diagnosing GCM errors over West Africa using relaxation experiments. Part II: Intraseasonal variability and African easterly waves. Climate Dyn., 37, 13131334, doi:10.1007/s00382-011-1106-1.

    • Search Google Scholar
    • Export Citation
  • Saha, S., and Coauthors, 2010: The NCEP Climate Forecast System Reanalysis. Bull. Amer. Meteor. Soc., 91, 10151057, doi:10.1175/2010BAMS3001.1.

    • Search Google Scholar
    • Export Citation
  • Saha, S., and Coauthors, 2014: The NCEP Climate Forecast System version 2. J. Climate, 27, 21852208, doi:10.1175/JCLI-D-12-00823.1.

  • Stéfanon, M., P. Drobinski, F. D’Andrea, C. Lebeaupin-Brossier, and S. Bastin, 2013: Soil moisture-temperature feedbacks at meso-scale during summer heat waves over western Europe. Climate Dyn., 42, 13091324, doi:10.1007/s00382-013-1794-9.

    • Search Google Scholar
    • Export Citation
  • Wang, Z. W., P. M. Zhai, and H. T. Zhang, 2003: Variation of drought over northern China during 1950–2000. J. Geogr. Sci., 13, 480487, doi:10.1007/BF02837887.

    • Search Google Scholar
    • Export Citation
  • Webster, P. J., and S. Yang, 1992: Monsoon and ENSO: Selectively interactive systems. Quart. J. Roy. Meteor. Soc., 118, 877926, doi:10.1002/qj.49711850705.

    • Search Google Scholar
    • Export Citation
  • Xie, P. P., A. Yatagai, M. Y. Chen, T. Hayasaka, Y. Fukushima, C. M. Liu, and S. Yang, 2007: A gauge-based analysis of daily precipitation over East Asia. J. Hydrometeor., 8, 607626, doi:10.1175/JHM583.1.

    • Search Google Scholar
    • Export Citation
  • Xue, Y. K., Z. Janjic, J. Dudhia, R. Vasic, and F. De Sales, 2014: A review on regional dynamical downscaling in intraseasonal to seasonal simulation/prediction and major factors that affect downscaling ability. Atmos. Res., 147–148, 6885, doi:10.1016/j.atmosres.2014.05.001.

    • Search Google Scholar
    • Export Citation
  • Yuan, C., T. Tozuka, W. A. Landman, and T. Yamagata, 2014: Dynamical seasonal prediction of southern African summer precipitation. Climate Dyn., 42, 33573374, doi:10.1007/s00382-013-1923-5.

    • Search Google Scholar
    • Export Citation
  • Zhong, Z., Y. Hu, J. Min, and H. Xu, 2007: Numerical experiments on the spin-up time for seasonal-scale regional climate modeling. Acta Meteor. Sin., 21, 409419.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Location of Hanjiang basin and the middle route of the SNWDP in China.

  • View in gallery

    The climatology of (a) spring, (b) summer, (c) autumn, and (d) winter precipitation over the Hanjiang basin during the period of 1981–2010.

  • View in gallery

    Configuration of the WRF Model nested domains and the regional topography of the Hanjiang basin. The outer domain is D01 with a 30-km grid spacing, and the inner domain is D02 with a 10-km grid spacing.

  • View in gallery

    The mean biases of CFSv2 forecast 2001–09 (a) spring, (b) summer, (c) autumn, and (d) winter precipitation when compared with the observations.

  • View in gallery

    Mean (a) spring, (b) summer, (c) autumn, and (d) winter precipitation biases of the WRF Model coupled with the KF–Noah, the BMJ–Noah, the GF–Noah, and the NSAS–Noah convective schemes during the period of 2001–09.

  • View in gallery

    Spearman’s correlation coefficients of (a) spring, (b) summer, (c) autumn, and (d) winter precipitation between the CFSv2 and observations during the period of 2001–09.

  • View in gallery

    As in Fig. 6, but between WRF Model coupled with the four convection schemes and the observations.

  • View in gallery

    ACCs between the forecast and observed (top left) spring, (top right) summer, (bottom left) autumn, and (bottom right) winter precipitation over Hanjiang basin for the CFSv2 (label OBS) and the WRF Model coupled with the KF–Noah, the BMJ–Noah, the GF–Noah, and the NSAS convective schemes during the period of 2001–09.

  • View in gallery

    The differences between the forecast and observed upper, middle, and lower subbasin-averaged seasonal precipitation for the CFSv2 and the WRF Model coupled with the KF–Noah, BMJ–Noah, CFS, NSAS–Noah, and GF–Noah convective schemes during the period of 2001–09. The dashed lines denote the climatic normal precipitation during the period of 1981–2010.

  • View in gallery

    Mean monthly precipitation biases of (top) the CFSv2 and (bottom) the WRF Model coupled with the KF–Noah convective schemes during the period of 2001–09 at different lead times (LT).

  • View in gallery

    As in Fig. 10, but for Spearman’s correlation coefficient with the observations.

  • View in gallery

    As in Fig. 8, but for monthly precipitation.

  • View in gallery

    As in Fig. 9, but for the upper basin-averaged monthly precipitation.

  • View in gallery

    Q–Q plots of the forecast and observed daily precipitation in (top left) spring, (top right) summer, (bottom left) autumn, and (bottom right) winter; the forecast daily precipitation is derived from the CFSv2 and the WRF Model coupled with the KF–Noah, BMJ–Noah, GF–Noah, and NSAS–NOAA convective schemes during the period of 2001–09.

  • View in gallery

    A comparison of the forecast 99th percentile of daily precipitation and the observations in summer, where the forecasts are from the CFSv2 and from the WRF Model coupled with the KF–Noah, BMJ–Noah, GF–Noah, and NSAS–Noah convective schemes during the period of 2001–09.

  • View in gallery

    Mean (a) spring, (b) summer, (c) autumn, and (d) winter precipitation biases of the WRF Model coupled with the KF–Noah, the BMJ–Noah, the GF–Noah, and the NSAS–Noah convective schemes relative to the CFSv2 during the period of 2001–09.

  • View in gallery

    Spearman’s correlation coefficients of (a) spring, (b) summer, (c) autumn, and (d) winter precipitation between the WRF Model coupled with the KF–Noah, the BMJ–Noah, the GF–Noah, and the NSAS–Noah convective schemes and the CFSv2 during the period of 2001–09.

  • View in gallery

    A comparison of average summer (a) CAPE and (b) CIN between the KF–Noah, the BMJ–Noah, the GF–Noah, and the NSAS–Noah convective schemes during the period of 2001–09.

  • View in gallery

    As in Fig. 18, but for 2007.

  • View in gallery

    The climatology of (a) spring, (b) summer, (c) autumn, and (d) winter convective precipitation over Hanjiang basin during the period of 1981–2010.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 1 1 1
PDF Downloads 0 0 0

High-Resolution Dynamical Downscaling of Seasonal Precipitation Forecasts for the Hanjiang Basin in China Using the Weather Research and Forecasting Model

View More View Less
  • 1 College of Hydrology and Water Resources, Hohai University, Nanjing, China
  • 2 Jiangsu Water Resources Department, Nanjing, China
© Get Permissions
Full access

Abstract

Management of water resources may benefit from seasonal precipitation forecasts, but for obtaining high enough resolution, dynamical downscaling is necessary. This study investigated the downscaling capability of the Weather Research and Forecasting (WRF) Model ARW, version 3.5, on seasonal precipitation forecasts for the Hanjiang basin in China during 2001–09, which was the water source of the middle route of the South-to-North Water Diversion Project (SNWDP). The WRF Model is forced by the National Centers for Environmental Prediction Operational Climate Forecast System, version 2 (CFSv2), and it performs at a high horizontal resolution of 10 km with four selected convection schemes. The National Oceanic and Atmospheric Administration’s Climate Prediction Center global daily precipitation data were employed to evaluate the WRF Model on multiple scales. On average, when large biases were removed, the WRF Model slightly outperformed the CFSv2 in all seasons, especially summer. In particular, the Kain–Fritsch convective scheme performed best in summer, whereas little difference was found in winter. The WRF Model showed similar results in monthly precipitation, but no time-dependent characteristics were observed for all months. The spatial anomaly correlation coefficient showed greater uncertainty than the bias and the temporal correlation coefficient. In addition, the performance of the WRF Model showed considerable regional variations. The upper basin always showed better agreement with observations than did the middle and lower parts of the basin. A comparison of the forecast and observed daily precipitation revealed that the WRF Model can provide more accurate extreme precipitation information than the CFSv2.

© 2017 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Zhiyong Wu, wzyhhu@gmail.com

Abstract

Management of water resources may benefit from seasonal precipitation forecasts, but for obtaining high enough resolution, dynamical downscaling is necessary. This study investigated the downscaling capability of the Weather Research and Forecasting (WRF) Model ARW, version 3.5, on seasonal precipitation forecasts for the Hanjiang basin in China during 2001–09, which was the water source of the middle route of the South-to-North Water Diversion Project (SNWDP). The WRF Model is forced by the National Centers for Environmental Prediction Operational Climate Forecast System, version 2 (CFSv2), and it performs at a high horizontal resolution of 10 km with four selected convection schemes. The National Oceanic and Atmospheric Administration’s Climate Prediction Center global daily precipitation data were employed to evaluate the WRF Model on multiple scales. On average, when large biases were removed, the WRF Model slightly outperformed the CFSv2 in all seasons, especially summer. In particular, the Kain–Fritsch convective scheme performed best in summer, whereas little difference was found in winter. The WRF Model showed similar results in monthly precipitation, but no time-dependent characteristics were observed for all months. The spatial anomaly correlation coefficient showed greater uncertainty than the bias and the temporal correlation coefficient. In addition, the performance of the WRF Model showed considerable regional variations. The upper basin always showed better agreement with observations than did the middle and lower parts of the basin. A comparison of the forecast and observed daily precipitation revealed that the WRF Model can provide more accurate extreme precipitation information than the CFSv2.

© 2017 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Zhiyong Wu, wzyhhu@gmail.com

1. Introduction

Accurate seasonal precipitation forecasts are of great importance to flood prevention as well as water resource management in China, where most areas are highly affected by East Asian monsoons. Massive extreme rainfall events frequently happen in the southeast part of China. In 1998, two different heavy mei-yu rainy phases occurred over the Yangtze River basin, lasting for about 31 days and leading to the most severe floods over the region since the year of 1954 (Liu et al. 2008). However, northern China usually suffers from severe droughts (Wang et al. 2003). To solve the water crisis in China, the South-to-North Water Diversion Project (SNWDP) has been built to provide water annually for cities in northern China.

However, the management of SNWDP remains a challenge. The Danjiangkou reservoir, which is located in the upper Hanjiang basin, is the water source of the middle route of the SNWDP (Fig. 1). Although the diversion of water resources from the Danjiangkou reservoir may benefit northern cities, it also influences the water usage of cities located in the Hanjiang basin. To balance the water requirement of northern China and guarantee the usage of the water sources areas, water supply plans are needed. The seasonal precipitation forecast, which is used as input into the distributed hydrological models to estimate inflows for the Danjiangkou reservoir, plays a key role in the establishment of water supply plans. However, it remains a challenge to forecast seasonal precipitation because of the complexity of the climate system over the Hanjiang basin. The Hanjiang basin is not only influenced by the East Asian summer monsoon (EASM) but also affected by a specific climate that is known as autumnal flood season, and the precipitation in the upper subbasin is usually heavier because of the stationary front of fall (Cui et al. 2007).

Fig. 1.
Fig. 1.

Location of Hanjiang basin and the middle route of the SNWDP in China.

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

During the last two decades, global climate models (GCMs) have become increasingly important for seasonal predictions and the prime sources of information for seasonal forecasts. Furthermore, they are extremely useful in providing comprehensive knowledge of the large-scale climate (Argüeso et al. 2012). In particular, coupled ocean–atmosphere GCMs (CGCMs) considering interactions between the ocean, land, and atmosphere have achieved significant improvement in recent years (Lang et al. 2014). The Climate Forecast System, version 2 (CFSv2), developed by the National Centers for Environmental Prediction, is a fully coupled atmosphere–ocean–land model, which is used for seasonal predictions (Saha et al. 2014), and its performance has been investigated and confirmed by numerous studies. Hu et al. (2013) evaluated the skill of the CFSv2 in predicting monthly sea surface temperature in the North Atlantic Ocean, and they found that it showed high prediction skills in high-latitude areas and in the tropical North Atlantic. Barnston and Tippett (2013) compared the ability of the Climate Forecast System, version 1 (CFSv1), and CFSv2 to predict monthly anomalies of the ENSO-related Niño-3.4 sea surface temperature index. Their results indicated that the overall skill of the CFSv2 is higher than the CFSv1.

Although CGCMs have been proven to be capable of forecasting large-scale climatic features, local- to regional-scale features that are crucial to local populations and the natural environment remain poorly represented in such models. Luo et al. (2013) suggested that forecasts of summer precipitation at the seasonal temporal scale and catchment spatial scale using dynamical models such as the CFSv2 remain challenging. Improvements of the model physics and parameterization are needed for better prediction of Asian monsoon rainfall. Recently, regional climate models (RCMs), whose lateral boundary conditions are derived from GCMs, have been widely applied for the dynamic downscaling of future climate to provide regional-scale information (Xue et al. 2014). Within the framework of the European PRUDENCE project, RCMs of the European area are run with horizontal resolution of 50 km, while the ENSEMBLES project compares the results of RCMs with 50-km resolution run for both Europe and West Africa (Pohl and Douville 2011). Research has indicated that RCMs are capable of reproducing the formation of mesoscale phenomena and high-resolution climatic features (Di Luca et al. 2012; Stéfanon et al. 2013). However, the large uncertainties that remain in dynamical downscaling prevent RCMs from adding valuable information.

Convective parameterization schemes, which are used to represent small-scale processes by simplification, are crucial factors that significantly affect dynamic downscaling capability. Cretat et al. (2012) compared simulations of summer precipitation in southern Africa using different physical parameterizations. Their results indicated that the intensity and other characteristics of summer precipitation are predominantly sensitive to the convective schemes and affected less by the planetary boundary layer and microphysical schemes. Pei et al. (2014) investigated the effects of the land surface model and convective parameterization on the simulation of short-term climate extremes. The results indicated that land surface processes strongly affect the precipitation amount, while the convective parameterization has considerable impact on the precipitation pattern. It is obvious that no single convective scheme could represent seasonal precipitation well throughout the world, indicating that further experiments are necessary to reveal the sensitivity of parameterization on regional-scale precipitation forecasts.

The aim of this paper was to study how the nested CFSv2 seasonal retrospective forecast with 10-km horizontal resolution performed over Hanjiang basin; this nesting was conducted using the WRF Model ARW, version 3.5. Analysis was focused on precipitation statistics on multiple temporal (seasonal, monthly, and daily) and spatial scales to evaluate the downscaling capability of the WRF Model and the influence of convective schemes on precipitation forecasts. The remainder of this paper is structured as follows: Section 2 describes the study area and observation dataset, and section 3 gives a detailed description of the forecast scheme in this study. Section 4 compares the differences when the WRF Model is coupled with different physical parameterizations. The results are discussed and conclusions are offered in section 5.

2. Study area and data

a. Study area

The Hanjiang basin belongs to the subtropical monsoon climate. The basin-averaged annual precipitation varies from 700 to 1000 mm. For this study, the Hanjiang basin was divided into three regions: the upper basin (Danjiangkou reservoir basin), middle basin, and lower basin, based on the work of Chen et al. (2007). The area of each subbasin has been listed in Table 1.

Table 1.

The area of each subbasin over Hanjiang basin.

Table 1.

The upper basin is within the northern subtropical monsoon climatic zone, and it has remarkable transitional climatic characteristics. The climatology of summer precipitation over this region ranges from 3.6 to 4.8 mm day−1 (Fig. 2). This is especially so in the autum flood season, which is a particular climate regime found in the upper basin. The convection during this period is as strong at that in summer (Cui et al. 2007), and the autumnal precipitation is between 2.4 and ~3.0 mm day−1.

Fig. 2.
Fig. 2.

The climatology of (a) spring, (b) summer, (c) autumn, and (d) winter precipitation over the Hanjiang basin during the period of 1981–2010.

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

The middle subbasin starts from the Danjiangkou reservoir and ends at the city of Zhongxiang. The spring and autumn precipitation varies from 1.8 to 3.5 mm day−1, while the summer precipitation is almost the same as that in the upper basin.

The lower subbasin is mainly composed of recent alluvial plain, and precipitation over this region is heaviest when compared with the two other subbasins. The summer precipitation can reach 5.0 mm day−1. Meanwhile, the spring and autumn precipitation is about 3.0–0.4.8 mm day−1, slightly higher than that in the upper and middle basin.

b. Data

The CFSv2 is a quasi-global, fully coupled atmosphere–ocean–land model, which has been implemented within the operations at the National Centers for Environmental Prediction since August 2011 (Saha et al. 2014). The resolution of the CFSv2 atmospheric model is T126 (~1.0°) with 64 sigma-pressure hybrid layers. Previous studies have shown that the CFSv2 has increased skills in predicting large-scale monsoonal circulations relative to CFSv1 (Jiang et al. 2013). The CFSv2 retrospective forecasts have initial conditions from the 0000 UTC cycle each day over the 12-yr period from January 1999 to December 2010 with a lead time of nearly 4 months.

The observed precipitation data used in this study were extracted from the National Oceanic and Atmospheric Administration’s Climate Prediction Center (CPC) unified gauge-based analysis of the global daily precipitation dataset developed by the U.S. National Center for Atmospheric Research (Xie et al. 2007). This dataset is based on gauge observations from over 30 000 stations around the globe, and it has been used in a variety of previous studies (Park et al. 2014; Saha et al. 2010). This dataset consists of daily precipitation from 1979 to the present with resolution of 0.5°. (We retrieved this dataset from http://climatedataguide.ucar.edu/climate-data/cmap-cpc-merged-analysis-precipitation.) The climatic normal is based on the 1981–2010 period, which is now commonly used.

In addition, National Centers for Environmental Prediction–U.S. Department of Energy (NCEP–DOE) Reanalysis-2 data products provided by the NOAA/OAR/ESRL Physical Sciences Division (from their website at http://www.esrl.noaa.gov/psd/), are used to calculate the climatic normal of the convective precipitation rate over the Hanjiang basin.

3. Methodology

a. General configuration of the WRF Model

The Weather Research and Forecasting (WRF) Model is a next-generation mesoscale numerical weather prediction system designed for both atmospheric research and operational forecasting needs (http://www.wrf-model.org/index.php). In the model, a set of equations that describe the evolution of the atmosphere is solved by numerical techniques in which the initial conditions and boundary conditions are always provided by the GCMs. Numerous parameterization schemes have been coupled within the WRF Model to resolve processes occurring within a grid box. Currently, the WRF Model has become widely used for forecasting regional climate (Caldwell et al. 2009; Pérez et al. 2014). In this study, the WRF Model, version 3.5, published on 18 April 2013, was selected to perform the dynamic downscaling simulations. The study domain was chosen to avoid the Qinghai–Tibetan Plateau to reduce the influence of complex topography. Furthermore, grid nudging was applied to avoid the difficulties RCMs might have in the representation of large-scale features. Previous research has indicated that a grid ratio of 3:1 for nesting is best for mesh refinement (Pérez et al. 2014). Thus, the domain configuration in this study comprised two domains: a 30-km-resolution parent domain with 56 × 34 grid points (D01) and a 10-km-resolution one-way nested domain with 91 × 61 grid points (D02), centered at 32.1°N, 110.2°E (Fig. 3). The time steps for D01 and D02 were 180 and 60 s, respectively, and all of the domains were discretized with 38 vertical eta levels.

Fig. 3.
Fig. 3.

Configuration of the WRF Model nested domains and the regional topography of the Hanjiang basin. The outer domain is D01 with a 30-km grid spacing, and the inner domain is D02 with a 10-km grid spacing.

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

All the atmospheric settings of the model were kept unchanged except for the adopted convective scheme. The physical parameterization schemes used for all the simulations were as follows: Noah land surface parameterization (Ek et al. 2003), Ferrier scheme for microphysics parameterization (Lang et al. 2003), Yonsei University scheme for planetary boundary layer parameterization (Hong et al. 2006), Dudhia scheme for shortwave parameterization (Dudhia 1989), and RRTM for longwave radiation parameterization (Mlawer et al. 1997).

b. Convective parameterizations

Convective parameterization has significant influence on precipitation simulations. The WRF Model, version 3.5, provides 11 convective schemes, each of which employs different trigger and closure assumptions, making them suitable for different convective regimes (Liang et al. 2004, 2007). In this study, four convective schemes were used independently to generate predictions of seasonal precipitation over the Hanjiang basin. The four convection schemes selected in this paper have different trigger functions and different considerations on the mechanism of convections. The comparison of four convection schemes provides information on the seasonal precipitation forecast uncertainties of the Hanjiang basin.

The Kain–Fritsch (KF) scheme (Kain 2004) uses low-level vertical motion as a trigger function and convective available potential energy removal as the closure; thus, it can provide better simulation of the convective processes associated with the late afternoon thermodynamic vertical motion induced by heating at the lower boundary. The Betts–Miller–Janjić (BMJ) scheme (Fonseca et al. 2015; Janjić 2000) is an adjustment scheme where the essential principle lies in the relaxation of the temperature and humidity profiles toward reference thermodynamic profiles. The Grell–Freitas (GF) convective parameterization scheme tries to smooth the transition to cloud-resolving scales, as proposed by Arakawa (2004). The new simplified Arakawa–Schubert (NSAS) scheme is a new mass flux scheme with deep–shallow components and momentum transport. In this scheme, the penetrative convection is simulated based on Grell (1993) with a saturated downdraft, and the cloud ensemble is reduced to a single cloud type with detrainment only from its top (Krishnamurti and Sanjay 2003).

c. Integration scheme

In this study, the CFSv2 retrospective forecasts were used to provide initial and lateral boundary conditions for the WRF Model during 2001–09. The lateral boundary conditions and sea surface temperature for the WRF were updated every six hours. Zhong et al. (2007) suggested that the length of spinup time for seasonal predictions should not be less than one month. Therefore, the first month integration was selected as the spinup time, and the remaining three months were selected as forecast periods. Forecasts initialized from 1 to 4 February for the March–May (MAM) prediction, 1 to 16 May for the June–August (JJA) prediction, 1 to 12 August for the September–November (SON) prediction, and 1 to 20 November for the December–February (DJF) prediction were taken as corresponding to the spring, summer, autumn, and winter periods, respectively.

d. Evaluation methods

Both the CFSv2 and WRF results were regridded into the CPC global daily precipitation dataset using a bilinear interpolation method. For each 0.5° grid box, the observed and the forecast precipitation were accumulated over daily, monthly, and seasonal time intervals. Subbasin-averaged precipitation was calculated to provide additional information for the forecast skills over different regions. Meanwhile, the multiyear average differences between the predictions and the corresponding observations of each month and each season were calculated to assess whether there was any systematic bias in the prediction, and if so, how the bias was distributed spatially.

Because of the skewness of the precipitation distribution, traditional correlation coefficients might be inadequate for the evaluation of precipitation prediction skill. In this study, the temporal Spearman’s rank correlation coefficient was calculated for each grid point to examine the model’s ability in predicting the direction of the seasonal anomaly each year. The centered anomaly correlation coefficient (ACC) was used to qualify the similarity of the WRF simulations and the observations in terms of the spatial patterns of the climatic anomalies.

The Kolmogorov–Smirnov (K–S) test was used to compare the probability distributions of the observed and the forecast gridded daily precipitation. The D statistics in the K–S test, which are the absolute maximum distance between the CDFs of the two samples, were calculated as well. However, the K–S test was more sensitive near the center of the distribution than at the tails, so other methods were also needed to evaluate extreme precipitation. The quantile (Q–Q) plot and the 99th percentile of the CFSv2 and WRF Model predicted daily precipitation were used to compare the forecast and observed extreme daily precipitation.

In addition, the convective available potential energy (CAPE) and the convective inhibition (CIN), which indicated the instability of the troposphere, were calculated to compare the differences among the four selected convective schemes (Moncrieff and Miller 1976).

4. Results

a. Seasonal precipitation

Figure 4 illustrates the multiyear mean seasonal precipitation bias from the CFSv2 for the period 2001–09. In general, a wet bias can be found in spring, and a dry bias is observed in summer. The multiyear mean autumn precipitation is slightly lower than the observations. The CFSv2 shows the highest prediction skill in winter, when the wet bias is always <0.5 mm day−1. No obvious regional differences are found for all seasons, indicating that the CFSv2 is unable to capture the spatial variation of precipitation because of its low resolution.

Fig. 4.
Fig. 4.

The mean biases of CFSv2 forecast 2001–09 (a) spring, (b) summer, (c) autumn, and (d) winter precipitation when compared with the observations.

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

Figure 5 shows the biases of the 9-yr mean seasonal precipitation between the WRF Model coupled with the different convective schemes and the CPC observations. The results are similar to those of the CFSv2. Overall, the autumn and winter precipitation forecast by the WRF Model shows a slight bias, while a larger systematic bias is found in spring and summer. Overestimation of spring precipitation by the KF scheme is clearly shown in Fig. 5. The KF forecast bias is >1.5 mm day−1 in most of the area. The BMJ, GF, and NSAS schemes also overestimate spring precipitation but less so than the KF scheme and the CFSv2 (bias ranging from 0.5 to 1.25 mm day−1). Dry biases are observed (ranging from 1 to 1.5 mm day−1) in the simulations of summer precipitation using the BMJ, GF, and NSAS schemes over Hanjiang basin. However, the forecast summer precipitation by the WRF Model coupled with the KF scheme shows considerable improvement relative to the CFSv2. Slight negative biases are found (approximately −0.5 mm day−1) over the middle of the Hanjiang basin, and slight positive biases (approximately 0.5 mm day−1) are found over eastern areas. Dry biases in the autumn precipitation have been removed by the WRF Model in the upper basin, but slight dry biases remain in the middle and lower basins. The winter precipitation simulations show smaller differences than in spring and summer between the four convective schemes and the CFSv2; however, wet biases range from 0.25 mm day−1 in eastern areas to 0.75 mm day−1 in western areas in winter.

Fig. 5.
Fig. 5.

Mean (a) spring, (b) summer, (c) autumn, and (d) winter precipitation biases of the WRF Model coupled with the KF–Noah, the BMJ–Noah, the GF–Noah, and the NSAS–Noah convective schemes during the period of 2001–09.

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

The temporal Spearman’s rank correlation coefficient (TCC) between the CFSv2 and the observations from 2001 to 2009 shows clearly that precipitation prediction has a negative correlation to the observations almost everywhere in spring, autumn, and winter. This suggests that the CFSv2 is unable to predict the correct direction of the precipitation anomaly in these seasons (Fig. 6). Nevertheless, the prediction in summer is much more accurate than that in the other seasons. The correlation between the prediction and observation in summer is positive in most areas of the upper basin (TCC reaches values of approximately 0.4); however, the correlation is still negative in the middle and lower basins in summer.

Fig. 6.
Fig. 6.

Spearman’s correlation coefficients of (a) spring, (b) summer, (c) autumn, and (d) winter precipitation between the CFSv2 and observations during the period of 2001–09.

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

The WRF simulations and observations at each grid point from 2001 to 2009 are different from those of the CFSv2 (Fig. 7). In summer, higher prediction skill is observed when compared with the CFSv2, and the TCC values in the upper basin range from 0.4 to 0.6. The BMJ scheme shows the lowest prediction skill in summer, and the TCC values are positive in western areas and negative in eastern areas. In winter, the TCC values among the four convective schemes are similar. The values in western parts of the Hanjiang basin are between 0.2 and 0.3, which are a little higher than that in the CFSv2. In terms of the correlation, no improvement is observed in spring and autumn precipitation when comparing the WRF Model with the CFSv2.

Fig. 7.
Fig. 7.

As in Fig. 6, but between WRF Model coupled with the four convection schemes and the observations.

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

The ACC is calculated for each year during 2001–09 to evaluate the spatial prediction capability of the CFSv2 and the WRF Model (Fig. 8). The results indicate that great uncertainty remains in the prediction of the spatial distribution of seasonal precipitation. The ACC values vary drastically from year to year and from one scheme to the other. In spring, the ACC values are negative in most of the years for both the CFSv2 and the WRF Model. However, there are some improvements in 2005 and 2007, where the ACC values of the WRF Model are nearly twice those of the CFSv2. Among these years, the WRF Model always shows the highest prediction skill in 2003, 2005, and 2008, with the highest ACCs of 0.38, 0.34, and 0.38 when using the GF, KF, and NSAS schemes, respectively. Conversely, the CFSv2 always shows higher prediction skill for autumn precipitation when compared with the WRF Model, except in 2004. Little difference exists between the CFSv2 and the WRF Model for winter precipitation simulations, except in 2007 and 2009.

Fig. 8.
Fig. 8.

ACCs between the forecast and observed (top left) spring, (top right) summer, (bottom left) autumn, and (bottom right) winter precipitation over Hanjiang basin for the CFSv2 (label OBS) and the WRF Model coupled with the KF–Noah, the BMJ–Noah, the GF–Noah, and the NSAS convective schemes during the period of 2001–09.

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

When taking the subbasin-averaged precipitation into consideration, the WRF Model and the CFSv2 forecast precipitation are always larger than the observations in spring (Fig. 9). The KF scheme always predicts the highest amount of precipitation, while in the CFSv2, it always shows the lowest amount, indicating that the troposphere is more unstable in the WRF Model. The results are similar in summer but with a lower bias when compared with spring precipitation. Furthermore, the summer precipitation in the upper and middle basin in 2005 and 2007 is much higher than the climate norm, while the CFSv2 shows lower forecast skills than the WRF Model in these two years. This implies that the WRF Model is more capable of serving additional useful precipitation information than the CFSv2, especially in heavier precipitation.

Fig. 9.
Fig. 9.

The differences between the forecast and observed upper, middle, and lower subbasin-averaged seasonal precipitation for the CFSv2 and the WRF Model coupled with the KF–Noah, BMJ–Noah, CFS, NSAS–Noah, and GF–Noah convective schemes during the period of 2001–09. The dashed lines denote the climatic normal precipitation during the period of 1981–2010.

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

b. Monthly precipitation

The high temporal resolutions of the CFSv2 and WRF Model make it possible to assess their prediction skills with different lead times. In this study, monthly precipitation biases and correlation coefficients have been calculated for each grid point in each season to permit deeper analysis of the prediction ability at one-, two-, and three-month lead times.

Figure 10 shows that the bias varies with lead time both in the CFSv2 and in the WRF Model. Generally, the WRF Model coupled with the KF scheme shows an improvement for all months relative to the CFSv2. With a lead time of three months, the dry bias has been removed from the CFSv2 simulation in May, and the wet bias in the WRF Model is about 0.5–1.5 mm day−1. A greater improvement is found in the WRF Model for June with a lead time of one month. The bias varies from 0.0 to 0.5 mm day−1 in most parts of the Hanjiang basin. Although a dry bias still exists in July and August (with a value of approximately −1.0 mm day−1) in the WRF Model, it is much smaller than that in the CFSv2, especially in the middle and lower basins. Similar results are observed in autumn when the dry biases are removed for most parts of the Hanjiang basin. Precipitation in October shows a slight wet bias of about 1 mm day−1 in the WRF Model, while the biases in the other months of autumn are about ±0.25 mm day−1. The forecast precipitation from December to February by the WRF Model is almost the same as that in the CFSv2.

Fig. 10.
Fig. 10.

Mean monthly precipitation biases of (top) the CFSv2 and (bottom) the WRF Model coupled with the KF–Noah convective schemes during the period of 2001–09 at different lead times (LT).

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

Similar results were observed when the WRF Model was coupled with the GF, NSAS, and BMJ convective schemes (not shown). Improvement has been observed irrespective of which convective scheme was adopted. Precipitation biases in May are about ±0.25 mm day−1 in the WRF Model, much smaller than that in March and April. Slight dry biases are observed in June when using the GF, NSAS, and BMJ convective parameterizations, while dry biases remain in June and August. The biases of the WRF forecast precipitation using these three convective schemes in autumn and winter are almost the same as the results of the KF scheme, which were about ±0.5 mm day−1.

The spatial distribution of the TCC values shows greater regional characteristics when compared with the biases (Fig. 11). The forecast upper-basin precipitation always has stronger correlation with the observations when compared with the middle and lower basins, especially in several months in spring and summer. The correlation is strongest in the upper basin in May with a lead time of three months. Meanwhile, the WRF Model always shows a higher prediction skill in the upper basin in summer, where the TCC values vary from 0.4 to 0.8. However, the upper basin shows no higher prediction skills than the middle and lower basins in December and January.

Fig. 11.
Fig. 11.

As in Fig. 10, but for Spearman’s correlation coefficient with the observations.

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

Figure 12 illustrates the ACC for monthly precipitation from 2001 to 2009 over the Hanjiang basin. The results are quite different from the seasonal precipitation simulations in which the WRF Model shows no notable improvement relative to the CFSv2. The ACC values of the CFSv2 are almost the same as the highest value of the WRF Model in March and May in most years, except 2004. The CFSv2 forecast monthly precipitation also shows higher prediction skill in June. Both the CFSv2 and the WRF Model show high prediction skill for precipitation in July. The prediction skills are relatively low in autumn with large negative ACC values. In winter, the ACC values have no remarkable differences between the WRF Model and the CFSv2, where large positive ACC values are always observed in February.

Fig. 12.
Fig. 12.

As in Fig. 8, but for monthly precipitation.

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

The monthly averaged precipitation for the upper basin shows similar results to the seasonal precipitation simulations (Fig. 13). Precipitation in April and May contributes most to the uncertainties in spring precipitation forecasts. Meanwhile, the largest uncertainties exist in the summer months, and the differences between the lowest and highest average precipitation could be up to 6 mm day−1 in July. The source of the uncertainty in the autumn precipitation simulation derives mainly from September and October. Moreover, monthly precipitation forecast by the CFSv2 is slightly different from that by the WRF Model. Large uncertainties are also observed in the middle and lower basins in summer, with higher uncertainties in the lower basin, especially in July (not shown).

Fig. 13.
Fig. 13.

As in Fig. 9, but for the upper basin-averaged monthly precipitation.

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

c. Daily precipitation characteristics

The observed and forecast daily precipitation are evaluated through the K–S test to see whether they follow the same distribution. The p values are always lower than 0.001 for all seasons, indicating that the predicted daily precipitation comes from the same distribution of the observation. However, the D statistic, which is the absolute maximum distance between the CDFs of the two samples, shows great diversity (Table 2). It is obvious that the cumulated distribution function of the WRF Model is much more correlated to the observation in spring and winter when compared with the CFSv2. However, the result of summer is contrary to that of the spring and winter. The CFSv2 shows higher prediction skill where the D statistic value is almost 0, while the WRF Model is always higher than 0.168. This suggests that the CFSv2 is more capable of forecasting the mean values of daily precipitation, for the K–S test is more sensitive near the center of the distribution than at the tails.

Table 2.

The D statistics of the CFSv2 and the WRF Model coupled with the KF, the BMJ, the GF, and the NSAS convective schemes forecast daily precipitation in spring, summer, autumn, and winter during the period of 2001–09.

Table 2.

In addition, the daily precipitation of the CFSv2 and WRF Models at each grid point are compared with the gridded observations in the Q–Q plot in Fig. 14. The performances of the CFSv2 and WRF Model vary greatly among the seasons. The WRF Model always shows a higher daily precipitation than the observations in spring, irrespective of the convective scheme used. Meanwhile, the WRF Model fits well with the observations for high daily precipitation in summer, indicating that the WRF Model is able to add valuable information for extreme precipitation events. In autumn, the WRF Model overestimates the highest quantiles, while the observed highest quantiles are underestimated by the CFSv2. The winter daily precipitation quantiles are overestimated by all models, although the WRF Model coupled with the NSAS scheme fits the observations the best.

Fig. 14.
Fig. 14.

Q–Q plots of the forecast and observed daily precipitation in (top left) spring, (top right) summer, (bottom left) autumn, and (bottom right) winter; the forecast daily precipitation is derived from the CFSv2 and the WRF Model coupled with the KF–Noah, BMJ–Noah, GF–Noah, and NSAS–NOAA convective schemes during the period of 2001–09.

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

To avoid the influence of the precipitation magnitude and distribution on the evaluation, the 99th percentiles of the CFSv2 and WRF Model predicted daily precipitation have been compared with that of the observation for each grid point. The CFSv2 predicted percentiles are much closer to the observation in spring and winter, while the WRF Model percentiles are higher than the observation, especially when the KF and BMJ cumulus convection schemes are used (not shown). However, the CFSv2 substantially underestimates the summer precipitation percentiles, and the WRF Model coupled with the KF and BMJ schemes show higher consistency with the observation than the CFSv2 (Fig. 15). The WRF Model coupled with GF scheme shows the highest prediction skills when compared with the CFSv2 and other convective schemes when the autumn daily precipitation is taken into consideration (not shown).

Fig. 15.
Fig. 15.

A comparison of the forecast 99th percentile of daily precipitation and the observations in summer, where the forecasts are from the CFSv2 and from the WRF Model coupled with the KF–Noah, BMJ–Noah, GF–Noah, and NSAS–Noah convective schemes during the period of 2001–09.

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

5. Discussion and conclusions

a. Discussion

To have a deeper understanding of the results from the WRF Model, the mean differences between the predicted seasonal precipitation from the CFSv2 and the WRF Model have been calculated for each grid point (Fig. 16). Small differences have been observed in autumn and winter with a value of ±0.5 mm day−1 for all subbasins. However, the summer precipitation shows great differentiation between the CFSv2 and the WRF Model. The predicted summer precipitation from the WRF Model is always higher than that from the CFSv2 in middle and lower subbasins no matter what convective scheme is used. On the contrary, precipitation predicted by the WRF Model is lower than that by the CFSv2 in the upper subbasin unless the KF scheme is used. The Spearman correlation of the CFSv2 and the WRF Model is strongest in autumn and winter precipitation predictions, in which the TCC values are within the range 0.8–1.0 (Fig. 17). The WRF Model predicted summer precipitation in the upper subbasin also shows a high correlation with the CFSv2, but the TCC values in the middle and lower subbasins are much lower, indicating that the convective schemes have a greater influence on precipitation in these areas.

Fig. 16.
Fig. 16.

Mean (a) spring, (b) summer, (c) autumn, and (d) winter precipitation biases of the WRF Model coupled with the KF–Noah, the BMJ–Noah, the GF–Noah, and the NSAS–Noah convective schemes relative to the CFSv2 during the period of 2001–09.

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

Fig. 17.
Fig. 17.

Spearman’s correlation coefficients of (a) spring, (b) summer, (c) autumn, and (d) winter precipitation between the WRF Model coupled with the KF–Noah, the BMJ–Noah, the GF–Noah, and the NSAS–Noah convective schemes and the CFSv2 during the period of 2001–09.

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

In terms of the temporal correlation coefficient between the forecast and observed precipitation, a slight improvement is observed in summer. However, the fall, winter, and spring precipitation is not improved when applying convection schemes. This result is similar to those of Dirmeyer et al. (2012), who found that the increase of horizontal resolution has little impact on the timing of precipitation even if the convection is parameterized. So far, the El Niño–Southern Oscillation (ENSO) variability still contributes most to the seasonal forecast skills. The Hanjiang basin is located in the middle of the Yangtze River basin. Lang et al. (2014) calculated the correlation coefficient between observed seasonal precipitation and Oceanic Niño Index over 17 hydroclimatic regions in China. The autumn and winter precipitation has a negative correlation with the Oceanic Niño Index with a value lower than −0.2, which may contribute to the low forecast skills in these seasons. The correlation coefficient in spring and summer is higher than 0.3. However, the spring predictability barrier for ENSO predictions may result in the low forecast skills of spring precipitation (Webster and Yang 1992).

The spatial anomaly correlation coefficient shows the greatest uncertainty when compared with the bias and temporal correlation coefficient. The ACC values can be positive under some schemes but negative under others for the same year and the same month. The averaged CAPE and CIN in summer are calculated and presented in Fig. 18. It is clear that the CAPE values of the BMJ and NSAS schemes are almost 100 J kg−1 lower than the KF and GF schemes in the upper basin and lower basin. Meanwhile, the CIN values of the KF scheme in the lower basin are lower than the three other schemes. These results indicate that the KF scheme always produces higher convective unstable atmosphere and leads to higher precipitation intensity. To have a more detailed analysis, the CAPE and CIN values are compared specifically in 2007, when the summer precipitation in each subbasin is almost the same (Fig. 19). The CAPE values of the KF and GF schemes are always higher than the two other schemes, particularly in the upper and lower basins. Additionally, the CIN values of the KF and GF schemes are much lower than that of the NSAS scheme in the upper basin, leading to higher ACC values.

Fig. 18.
Fig. 18.

A comparison of average summer (a) CAPE and (b) CIN between the KF–Noah, the BMJ–Noah, the GF–Noah, and the NSAS–Noah convective schemes during the period of 2001–09.

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

Fig. 19.
Fig. 19.

As in Fig. 18, but for 2007.

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

Another interesting finding is that the upper basin always shows better agreement with the observations than the middle and lower basins, especially in summer. To find out the reasons, the climatological distribution of convective precipitation is calculated based on the NCEP–DOE Reanalysis-2 data (http://www.esrl.noaa.gov/psd/). The results indicate that the convective precipitation in the upper basin is always lower than that in the middle and lower basins for all seasons, which means that the MCSs are much more active in the middle and lower basin than in the upper basin (Fig. 20). The higher forecast skills over the upper basin may be mainly due to the lower active MCSs, and this also gives evidence that the high-resolution WRF Model along with convective schemes could better resolve mesoscale weather systems than the CFSv2. This can also be used to explain the differences of daily precipitation characteristics between the CFSv2 and the WRF Model. These findings agree with the conclusions of Yuan et al. (2014), who found that higher model resolution might contribute to higher prediction skill because of the greater capability of resolving deep convection in summer.

Fig. 20.
Fig. 20.

The climatology of (a) spring, (b) summer, (c) autumn, and (d) winter convective precipitation over Hanjiang basin during the period of 1981–2010.

Citation: Journal of Applied Meteorology and Climatology 56, 5; 10.1175/JAMC-D-16-0268.1

b. Conclusions

This study investigated the downscaling capability of the WRF Model when coupled with different convective schemes over the Hanjiang basin at a horizontal resolution of 10 km. On average, the WRF Model outperformed the CFSv2 in all seasons on the seasonal time scale. It was apparent that the dry biases in summer and autumn in CFSv2 were removed in the WRF Model. In particular, the WRF Model coupled with the KF scheme performed better than the three other schemes. Moreover, the convective schemes showed greater influence on precipitation in summer than in the other seasons.

Although the WRF Model showed smaller biases in monthly precipitation, no time-dependent characteristics were observed for all months, which suggests that a convection-permitting scale is needed to improve the timing of precipitation forecasts (Brisson et al. 2016).

In terms of spatial anomaly distribution, improvement was observed only in summer, during which the ACC values of the WRF Model were almost twice as high as the CFSv2 in 2003 and 2005. The KF scheme always showed higher prediction skill in summer in comparison with the three other schemes.

The prediction skill for the upper basin was always higher than that for the middle and lower basins, especially in summer. Furthermore, daily precipitation quantiles forecast by the WRF Model were closer to the observations in summer, which indicates that the WRF Model is able to provide more accurate extreme precipitation than the CFSv2.

The accuracy of the climate forecast has been improved greatly with the improvement of CGCMs during the last two decades. The CFSv2 transcends almost all aspects of the CFSv1, including the data assimilation system and forecast model components. However, seasonal precipitation forecasts remain a challenging problem. Bauer et al. (2015) pointed out that convection parameterizations would remain a challenge for climate modeling because of the limitation of resolutions. Meanwhile, other parameterization schemes, including land surface schemes and radiation schemes, have been shown to influence precipitation simulations, suggesting that further studies on such parameterizations should be conducted to improve the prediction skill of seasonal precipitation.

Acknowledgments

This work is supported by the Natural Science Foundation of Jiangsu Province of China (Grant BK20131368), the National Natural Science Foundation of China (Grant 51579065), the Special Public Sector Research Program of Ministry of Water Resources (Grants 201301040 and 201401008) and the Program for New Century Excellent Talents in University (Grant NCET-12-0842).

REFERENCES

  • Arakawa, A., 2004: The cumulus parameterization problem: Past, present, and future. J. Climate, 17, 24932525, doi:10.1175/1520-0442(2004)017<2493:RATCPP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Argüeso, D., J. M. Hidalgo-Muñoz, S. R. Gámiz-Fortis, M. J. Esteban-Parra, and Y. Castro-Díez, 2012: Evaluation of WRF mean and extreme precipitation over Spain: Present climate (1970–99). J. Climate, 25, 48834897, doi:10.1175/JCLI-D-11-00276.1.

    • Search Google Scholar
    • Export Citation
  • Barnston, A. G., and M. K. Tippett, 2013: Predictions of Nino3.4 SST in CFSv1 and CFSv2: A diagnostic comparison. Climate Dyn., 41, 16151633, doi:10.1007/s00382-013-1845-2.

    • Search Google Scholar
    • Export Citation
  • Bauer, P., A. Thorpe, and G. Brunet, 2015: The quiet revolution of numerical weather prediction. Nature, 525, 4755, doi:10.1038/nature14956.

    • Search Google Scholar
    • Export Citation
  • Brisson, E., K. Van Weverberg, M. Demuzere, A. Devis, S. Saeed, and M. Stengel, and N. P. M. van Lipzig, 2016: How well can a convection-permitting climate model reproduce decadal statistics of precipitation, temperature and cloud characteristics? Climate Dyn., 47, 30433061, doi:10.1007/s00382-016-3012-z.

    • Search Google Scholar
    • Export Citation
  • Caldwell, P., H.-N. S. Chin, D. C. Bader, and G. Bala, 2009: Evaluation of a WRF dynamical downscaling simulation over California. Climatic Change, 95, 499521, doi:10.1007/s10584-009-9583-5.

    • Search Google Scholar
    • Export Citation
  • Chen, H., S. L. Guo, C. Y. Xu, and V. P. Singh, 2007: Historical temporal trends of hydroclimatic variables and runoff response to climate variability and their relevance in water resource management in the Hanjiang basin. J. Hydrol., 344, 171184, doi:10.1016/j.jhydrol.2007.06.034.

    • Search Google Scholar
    • Export Citation
  • Cretat, J., B. Pohl, Y. Richard, and P. Drobinski, 2012: Uncertainties in simulating regional climate of southern Africa: Sensitivity to physical parameterizations using WRF. Climate Dyn., 38, 613634, doi:10.1007/s00382-011-1055-8.

    • Search Google Scholar
    • Export Citation
  • Cui, J.-X., S.-Z. Xu, R.-Q. Wang, and H.-Y. Zhang, 2007: A study on meteorological and hydrological characteristics of autumnal flood in the Hanjiang River in 2005. Torrential Rain Disasters, 26, 289294.

    • Search Google Scholar
    • Export Citation
  • Di Luca, A., R. D. Elia, and R. Laprise, 2012: Potential for added value in precipitation simulated by high-resolution nested regional climate models and observations. Climate Dyn., 38, 12291247, doi:10.1007/s00382-011-1068-3.

    • Search Google Scholar
    • Export Citation
  • Dirmeyer, P. A., and Coauthors, 2012: Simulating the diurnal cycle of rainfall in global climate models: Resolution versus parameterization. Climate Dyn., 39, 399418, doi:10.1007/s00382-011-1127-9.

    • Search Google Scholar
    • Export Citation
  • Dudhia, J., 1989: Numerical study of convection observed during the Winter Monsoon Experiment using a mesoscale two-dimensional model. J. Atmos. Sci., 46, 30773107, doi:10.1175/1520-0469(1989)046<3077:NSOCOD>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Ek, M. B., K. E. Mitchell, Y. Lin, E. Rogers, P. Grunmann, V. Koren, G. Gayno, and J. D. Tarpley, 2003: Implementation of Noah land surface model advances in the National Centers for Environmental Prediction operational mesoscale Eta model. J. Geophys. Res., 108, 8851, doi:10.1029/2002JD003296.

    • Search Google Scholar
    • Export Citation
  • Fonseca, R. M., T. Zhang, and K. T. Yong, 2015: Improved simulation of precipitation in the tropics using a modified BMJ scheme in the WRF Model. Geosci. Model Dev., 8, 29152928, doi:10.5194/gmd-8-2915-2015.

    • Search Google Scholar
    • Export Citation
  • Grell, G. A., 1993: Prognostic evaluation of assumptions used by cumulus parameterizations. Mon. Wea. Rev., 121, 764787, doi:10.1175/1520-0493(1993)121<0764:PEOAUB>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Hong, S. Y., Y. Noh, and J. Dudhia, 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 23182341, doi:10.1175/MWR3199.1.

    • Search Google Scholar
    • Export Citation
  • Hu, Z.-Z., A. Kumar, B. Huang, W. Wang, J. Zhu, and C. Wen, 2013: Prediction skill of monthly SST in the North Atlantic Ocean in NCEP Climate Forecast System version 2. Climate Dyn., 40, 27452759, doi:10.1007/s00382-012-1431-z.

    • Search Google Scholar
    • Export Citation
  • Janjić, Z. I., 2000: Comments on “Development and evaluation of a convection scheme for use in climate models.” J. Atmos. Sci., 57, 3686, doi:10.1175/1520-0469(2000)057<3686:CODAEO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Jiang, X. W., S. Yang, Y. Q. Li, A. Kumar, X. W. Liu, Z. Y. Zuo, and B. Jha, 2013: Seasonal-to-interannual prediction of the Asian summer monsoon in the NCEP Climate Forecast System version 2. J. Climate, 26, 37083727, doi:10.1175/JCLI-D-12-00437.1.

    • Search Google Scholar
    • Export Citation
  • Kain, J. S., 2004: The Kain–Fritsch convective parameterization: An update. J. Appl. Meteor., 43, 170181, doi:10.1175/1520-0450(2004)043<0170:TKCPAU>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Krishnamurti, T. N., and J. Sanjay, 2003: A new approach to the cumulus parameterization issue. Tellus, 55A, 275300, doi:10.1034/j.1600-0870.2003.00021.x.

    • Search Google Scholar
    • Export Citation
  • Lang, S., W. K. Tao, J. Simpson, and B. Ferrier, 2003: Modeling of convective–stratiform precipitation processes: Sensitivity to partition methods. J. Appl. Meteor., 42, 505527, doi:10.1175/1520-0450(2003)042<0505:MOCSPP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Lang, Y., and Coauthors, 2014: Evaluating skill of seasonal precipitation and temperature predictions of NCEP CFSv2 forecasts over 17 hydroclimatic regions in China. J. Hydrometeor., 15, 15461559, doi:10.1175/JHM-D-13-0208.1.

    • Search Google Scholar
    • Export Citation
  • Liang, X. Z., L. Li, A. Dai, and K. E. Kunkel, 2004: Regional climate model simulation of summer precipitation diurnal cycle over the United States. Geophys. Res. Lett., 31, L24208, doi:10.1029/2004GL021054.

    • Search Google Scholar
    • Export Citation
  • Liang, X. Z., M. Xu, K. E. Kunkel, G. A. Grell, and J. S. Kain, 2007: Regional climate model simulation of U.S.–Mexico summer precipitation using the optimal ensemble of two cumulus parameterizations. J. Climate, 20, 52015207, doi:10.1175/JCLI4306.1.

    • Search Google Scholar
    • Export Citation
  • Liu, H., D.-L. Zhang, and B. Wang, 2008: Daily to submonthly weather and climate characteristics of the summer 1998 extreme rainfall over the Yangtze River basin. J. Geophys. Res., 113, D22101, doi:10.1029/2008JD010072.

    • Search Google Scholar
    • Export Citation
  • Luo, L., W. Tang, Z. Lin, and E. F. Wood, 2013: Evaluation of summer temperature and precipitation predictions from NCEP CFSv2 retrospective forecast over China. Climate Dyn., 41, 22132230, doi:10.1007/s00382-013-1927-1.

    • Search Google Scholar
    • Export Citation
  • Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A. Clough, 1997: Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102, 16 66316 682, doi:10.1029/97JD00237.

    • Search Google Scholar
    • Export Citation
  • Moncrieff, M. W., and M. J. Miller, 1976: The dynamics and simulation of tropical cumulonimbus and squall lines. Quart. J. Roy. Meteor. Soc., 102, 373394, doi:10.1002/qj.49710243208.

    • Search Google Scholar
    • Export Citation
  • Park, D. S. R., C. H. Ho, and J. H. Kim, 2014: Growing threat of intense tropical cyclones to East Asia over the period 1977–2010. Environ. Res. Lett., 9, 014008, doi:10.1088/1748-9326/9/1/014008.

    • Search Google Scholar
    • Export Citation
  • Pei, L. S., N. Moore, S. Y. Zhong, L. F. Luo, D. W. Hyndman, W. E. Heilman, and Z. Q. Gao, 2014: WRF Model sensitivity to land surface model and cumulus parameterization under short-term climate extremes over the southern Great Plains of the United States. J. Climate, 27, 77037724, doi:10.1175/JCLI-D-14-00015.1.

    • Search Google Scholar
    • Export Citation
  • Pérez, J. C., J. P. Díaz, A. González, J. Expósito, F. Rivera-López, and D. Taima, 2014: Evaluation of WRF parameterizations for dynamical downscaling in the Canary Islands. J. Climate, 27, 56115631, doi:10.1175/JCLI-D-13-00458.1.

    • Search Google Scholar
    • Export Citation
  • Pohl, B., and H. Douville, 2011: Diagnosing GCM errors over West Africa using relaxation experiments. Part II: Intraseasonal variability and African easterly waves. Climate Dyn., 37, 13131334, doi:10.1007/s00382-011-1106-1.

    • Search Google Scholar
    • Export Citation
  • Saha, S., and Coauthors, 2010: The NCEP Climate Forecast System Reanalysis. Bull. Amer. Meteor. Soc., 91, 10151057, doi:10.1175/2010BAMS3001.1.

    • Search Google Scholar
    • Export Citation
  • Saha, S., and Coauthors, 2014: The NCEP Climate Forecast System version 2. J. Climate, 27, 21852208, doi:10.1175/JCLI-D-12-00823.1.

  • Stéfanon, M., P. Drobinski, F. D’Andrea, C. Lebeaupin-Brossier, and S. Bastin, 2013: Soil moisture-temperature feedbacks at meso-scale during summer heat waves over western Europe. Climate Dyn., 42, 13091324, doi:10.1007/s00382-013-1794-9.

    • Search Google Scholar
    • Export Citation
  • Wang, Z. W., P. M. Zhai, and H. T. Zhang, 2003: Variation of drought over northern China during 1950–2000. J. Geogr. Sci., 13, 480487, doi:10.1007/BF02837887.

    • Search Google Scholar
    • Export Citation
  • Webster, P. J., and S. Yang, 1992: Monsoon and ENSO: Selectively interactive systems. Quart. J. Roy. Meteor. Soc., 118, 877926, doi:10.1002/qj.49711850705.

    • Search Google Scholar
    • Export Citation
  • Xie, P. P., A. Yatagai, M. Y. Chen, T. Hayasaka, Y. Fukushima, C. M. Liu, and S. Yang, 2007: A gauge-based analysis of daily precipitation over East Asia. J. Hydrometeor., 8, 607626, doi:10.1175/JHM583.1.

    • Search Google Scholar
    • Export Citation
  • Xue, Y. K., Z. Janjic, J. Dudhia, R. Vasic, and F. De Sales, 2014: A review on regional dynamical downscaling in intraseasonal to seasonal simulation/prediction and major factors that affect downscaling ability. Atmos. Res., 147–148, 6885, doi:10.1016/j.atmosres.2014.05.001.

    • Search Google Scholar
    • Export Citation
  • Yuan, C., T. Tozuka, W. A. Landman, and T. Yamagata, 2014: Dynamical seasonal prediction of southern African summer precipitation. Climate Dyn., 42, 33573374, doi:10.1007/s00382-013-1923-5.

    • Search Google Scholar
    • Export Citation
  • Zhong, Z., Y. Hu, J. Min, and H. Xu, 2007: Numerical experiments on the spin-up time for seasonal-scale regional climate modeling. Acta Meteor. Sin., 21, 409419.

    • Search Google Scholar
    • Export Citation
Save