Scholarlink

<p>The primary objective of the present study is to carry out the evaluation of the Global Forecast System (GFS) at T1534 (~ 12.5&#xa0;km) model forecasted precipitation skill, its variability and systematic biases at multiple temporal scales over the Indian summer monsoon (ISM) region. The study uses 19&#xa0;years (2000–2018) of GFS T1534 forecasted precipitation. The result shows that the operational short and medium range prediction system, i.e., GFS T1534 predicts reasonable global mean precipitation across all the forecast lead times. However, it exhibits systematic wet biases over the equatorial central Pacific (0.8 mmday<sup>−1</sup>), Atlantic (1.4 mmday<sup>−1</sup>), central Africa (1.2 mmday<sup>−1</sup>), and Maritime Continent (0.8 mmday<sup>−1</sup>). Additionally, dry precipitation bias noted over the western Pacific (−&#xa0;2.7 mmday<sup>−1</sup>) and equatorial Indian Ocean (EIO) (− 1.4 mmday<sup>−1</sup>) region. Over ISM domain, the precipitation is overestimated in Central India (CI), the Western Ghats (WGs), the Indo-Gangetic Plains, northeast India, and Himalayan foothills, while systematic underestimation occurs in the Bay of Bengal (BoB) and eastern EIO. Further, the precipitation probability distribution function (PDF) analysis reveals an overestimation of light-to-moderate rain and an underestimation of heavy rainfall. Seasonal differences in precipitation distribution and its PDF are evident in the present study, with improved forecasts in June. In addition, the synoptic-scale rainfall variability is underestimated, while intraseasonal oscillation (ISO) variability is overestimated indicating requirement for ocean coupling or better parameterization processes. The evaluation of extreme rainfall (90th and 95th percentiles) shows consistent underestimation of intensity, particularly at the 95th percentile (−&#xa0;7.3 mmday<sup>−1</sup>) than 90th percentile (−&#xa0;2.1 mmday<sup>−1</sup>). The diurnal cycle assessment shows the model captures broad phase patterns for Day-1 lead but premature early peaks (3&#xa0;h earlier than observation) in longer lead forecasts over the CI region. The precipitation skill score analysis suggests that the highest skill is noted at lower thresholds (2.5&#xa0;mm/day) and it diminishes at higher thresholds and becomes negligible (ETS value 0.005–0.02) beyond Day-3 for extreme rainfall. The results indicate further efforts are needed through improved model physics and parameterization for enhancing forecast accuracy.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Evaluation of GFS T1534 hindcast precipitation skill over the Indian summer monsoon region

  • 期刊论文

  • 发表日期：

    2026年2月6日

查看全文
  • Malay Ganai,
  • Snehlata Tirkey,
  • R. P. M. Krishna,
  • C. J. Johny,
  • Tanmoy Goswami,
  • Medha Deshpande,
  • P. Mukhopadhyay

摘要

The primary objective of the present study is to carry out the evaluation of the Global Forecast System (GFS) at T1534 (~ 12.5 km) model forecasted precipitation skill, its variability and systematic biases at multiple temporal scales over the Indian summer monsoon (ISM) region. The study uses 19 years (2000–2018) of GFS T1534 forecasted precipitation. The result shows that the operational short and medium range prediction system, i.e., GFS T1534 predicts reasonable global mean precipitation across all the forecast lead times. However, it exhibits systematic wet biases over the equatorial central Pacific (0.8 mmday−1), Atlantic (1.4 mmday−1), central Africa (1.2 mmday−1), and Maritime Continent (0.8 mmday−1). Additionally, dry precipitation bias noted over the western Pacific (− 2.7 mmday−1) and equatorial Indian Ocean (EIO) (− 1.4 mmday−1) region. Over ISM domain, the precipitation is overestimated in Central India (CI), the Western Ghats (WGs), the Indo-Gangetic Plains, northeast India, and Himalayan foothills, while systematic underestimation occurs in the Bay of Bengal (BoB) and eastern EIO. Further, the precipitation probability distribution function (PDF) analysis reveals an overestimation of light-to-moderate rain and an underestimation of heavy rainfall. Seasonal differences in precipitation distribution and its PDF are evident in the present study, with improved forecasts in June. In addition, the synoptic-scale rainfall variability is underestimated, while intraseasonal oscillation (ISO) variability is overestimated indicating requirement for ocean coupling or better parameterization processes. The evaluation of extreme rainfall (90th and 95th percentiles) shows consistent underestimation of intensity, particularly at the 95th percentile (− 7.3 mmday−1) than 90th percentile (− 2.1 mmday−1). The diurnal cycle assessment shows the model captures broad phase patterns for Day-1 lead but premature early peaks (3 h earlier than observation) in longer lead forecasts over the CI region. The precipitation skill score analysis suggests that the highest skill is noted at lower thresholds (2.5 mm/day) and it diminishes at higher thresholds and becomes negligible (ETS value 0.005–0.02) beyond Day-3 for extreme rainfall. The results indicate further efforts are needed through improved model physics and parameterization for enhancing forecast accuracy.