Scholarlink

<p>The high-resolution rainfall forecast was generated using ARW-WRF for the NE monsoon season for 2022 and 2023. The model forecast was available for 12&#xa0;km, 4&#xa0;km, 1.33&#xa0;km and 0.44&#xa0;km. Rainfall is a crucial parameter in weather forecasting because it strongly influences agricultural activities. In the present study, the skill of kilometre- and sub-kilometre-scale rainfall forecasting has been investigated by validating against GPM, IMD-merged, and CHIRPS rainfall. In addition, the rainfall forecasts over Ananthapur and Sri Sathya Sai districts have been verified with AWS rainfall data and bias-correction techniques are demonstrated to improve the forecast accuracy. For a 1.33&#xa0;km resolution forecast over South Peninsular India during the OND of 2022 and 2023, the model consistently showed higher RMSE values over oceanic areas, notably over the southwest BoB, and exhibited a wet bias across much of the southern Peninsula land region. Moreover, wet bias tendency increased with the increase of forecast lead time, particularly over oceanic regions. It is also seen that the model tends to overestimate the frequency of light rain events and underestimate the frequency of heavy rain events with respect to GPM rainfall. For a 0.44&#xa0;km resolution forecast over South Andhra Pradesh during OND 2022 and 2023, a wet bias is observed over the interior parts and a dry bias is noticed over the coastal region. The forecast skill scores are calculated for all resolutions over the south Andhra Pradesh (12° to 16°N, 77° to 81°E). The probability of detection decreased with increasing resolution and false alarm ratio decreased with increasing resolution. The overall accuracy for 2022 is 74%, and that for 2023 is 70%. The Taylor diagram analysis over the south Andhra Pradesh indicates that 4&#xa0;km and 1.33&#xa0;km resolution forecasts shows the best overall agreement with observations in terms of spatial variability and RMSE, whereas 0.44&#xa0;km resolution forecast exhibits higher correlation but reduced spatial variability. The location-specific 0.44&#xa0;km forecast is assessed using AWS data over Anantapur and Sri Sathya Sai districts for OND 2022. The location-specific&#xa0;rainfall forecast for all lead times shows 75% accuracy. Quantile Mapping and Random Forest (RF) bias correction methods were applied to improve the rainfall forecasts and it was found that the RF based approach showed a lower RMSE (about 14% reduction) over the study region during the evaluation period.</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 Convection Permitting Rainfall Forecast Over South Peninsular India During Northeast Monsoon Season

  • 期刊论文

  • 发表日期：

    2026年6月30日

查看全文
  • Jyoti Bhate,
  • G. Pavani,
  • Amit Kesarkar

摘要

The high-resolution rainfall forecast was generated using ARW-WRF for the NE monsoon season for 2022 and 2023. The model forecast was available for 12 km, 4 km, 1.33 km and 0.44 km. Rainfall is a crucial parameter in weather forecasting because it strongly influences agricultural activities. In the present study, the skill of kilometre- and sub-kilometre-scale rainfall forecasting has been investigated by validating against GPM, IMD-merged, and CHIRPS rainfall. In addition, the rainfall forecasts over Ananthapur and Sri Sathya Sai districts have been verified with AWS rainfall data and bias-correction techniques are demonstrated to improve the forecast accuracy. For a 1.33 km resolution forecast over South Peninsular India during the OND of 2022 and 2023, the model consistently showed higher RMSE values over oceanic areas, notably over the southwest BoB, and exhibited a wet bias across much of the southern Peninsula land region. Moreover, wet bias tendency increased with the increase of forecast lead time, particularly over oceanic regions. It is also seen that the model tends to overestimate the frequency of light rain events and underestimate the frequency of heavy rain events with respect to GPM rainfall. For a 0.44 km resolution forecast over South Andhra Pradesh during OND 2022 and 2023, a wet bias is observed over the interior parts and a dry bias is noticed over the coastal region. The forecast skill scores are calculated for all resolutions over the south Andhra Pradesh (12° to 16°N, 77° to 81°E). The probability of detection decreased with increasing resolution and false alarm ratio decreased with increasing resolution. The overall accuracy for 2022 is 74%, and that for 2023 is 70%. The Taylor diagram analysis over the south Andhra Pradesh indicates that 4 km and 1.33 km resolution forecasts shows the best overall agreement with observations in terms of spatial variability and RMSE, whereas 0.44 km resolution forecast exhibits higher correlation but reduced spatial variability. The location-specific 0.44 km forecast is assessed using AWS data over Anantapur and Sri Sathya Sai districts for OND 2022. The location-specific rainfall forecast for all lead times shows 75% accuracy. Quantile Mapping and Random Forest (RF) bias correction methods were applied to improve the rainfall forecasts and it was found that the RF based approach showed a lower RMSE (about 14% reduction) over the study region during the evaluation period.