Abstract

The present study highlights the role of high-resolution land data assimilation in improving the prediction of the radiation fog and near-surface meteorological variables. The performance of the Weather Research and Forecasting (WRF) model coupled with the High-Resolution Land Data Assimilation S... Show moreThe present study highlights the role of high-resolution land data assimilation in improving the prediction of the radiation fog and near-surface meteorological variables. The performance of the Weather Research and Forecasting (WRF) model coupled with the High-Resolution Land Data Assimilation System (HRLDAS) is evaluated for a dense fog event that occurred on 24-25 January 2018 using detailed observations from the Winter Fog EXperiment (WiFEX) over the Delhi (India) region. The Noah-MP Land-Surface Model (LSM) based HRLDAS framework was executed in uncoupled mode to develop fine-grid soil states for soil moisture (SM) and soil temperature (ST) covering the Indo Gangetic Plain (IGP) region at 2 km horizontal grid resolution. The quality of soil states (SM/ST) from the HRLDAS and reanalysis (CNTL) dataset was verified with observed SM/ST at the Indira Gandhi International (IGI) airport, New Delhi, during 2017-18 winter months (December-January). It was found that the soil state in the CNTL dataset is moist, despite the actual soil condition being dryer at the observation site. HRLDAS simulated SM reasonably agrees with observations at IGI by reducing wet mean bias by about 56%. Subsequently, four sensitive experiments were carried out using Noah-MP (NM) and Pleim-Xiu (PX) land-surface parameterisations in the WRF model initialised with CNTL and HRLDAS soil states. We found that the bias in micro-meteorological variables (T2, RH2, WS10) and Turbulent Kinetic Energy (TKE) during the fog event was significantly improved with Pleim-Xiu (PX) land-surface parameterisations and HRLDAS soil states (HRLDAS_PX). Statistical performance of the micro-meteorological variables (T2, RH2 and WS10) exhibited low variance (0.13 degrees C, 0.10% and 0.75 m s-1), high correlation (0.93, 0.92 and 0.82) and high index of agreement (0.92, 0.87 and 0.78). As a result, the error in fog onset timing was notably reduced to 02 h, and the vertical representation of fog was skillfully demonstrated in the HRLDAS_PX simulation. The sensitivity experiments revealed that there is a need to revisit soil states to improve the skill of the fog forecast. Show less