Incorporation of Surface Observations in the Land Data Assimilation System and Application to Mesoscale Simulation of Pre-monsoon Thunderstorms

Abstract

In the present study, local insitu observations are utilized in the Land Data Assimilation System (LDAS) to generate high-resolution (4 km and hourly) soil moisture (SM) and soil temperature (ST) data over India. Further, the impact of the LDAS-derived SM and ST initialization on simulation of pre-monsoon (March–May) thunderstorms over the Gangetic West Bengal region are assessed. The high-resolution (4 km and hourly) land surface conditions such as SM and ST data are prepared for the period from 2010 to 2013 using the LDAS forced with various insitu observations from Automatic Weather Stations (AWS), Meteorological Aviation Reports (METAR), and micro-meteorological tower observations over India. Four thunderstorm (TS) events during the pre-monsoon season of 2010 are considered for the numerical experiments using climatological SM and ST at coarser resolution (CNTR) and LDAS-generated high-resolution SM and ST (WLDAS) for initializing the Weather Research and Forecasting (WRF) Model. The efficacy of the LDAS-generated SM and ST data are verified against micro-meteorological tower observations at Kharagpur, West Bengal, and the results indicate the magnitude and variation in the data product are close to observations. The initialization of high-resolution LDAS-generated SM and ST in the WRF improves the simulation of near-surface air temperature, humidity, and pressure at Kharagapur. Most importantly, the location and timing of the storm are relatively better simulated in the WLDAS than CNTR over the Kolkata region. The study encourages the use of more local observations in the generation of high-resolution SM and ST data for application in the simulation of pre-monsoon TSs.

Appendix

1.1 Barnes Objective Analysis

The precipitation is interpolated using the Barnes successive correction method (Barnes 1973; Koch et al. 1983). The Barnes scheme assumes the desired scalar field at the grid point is the sum of the known prior information or the background and weighted sum of the observational increment inside a certain radius of influence. The weight (W_n) at each station within radius R is a function of distance (X_n, Y_n) to grid point (i, j) and expressed as:

$${W}_{n}=\mathrm{exp} \left(-\frac{{{d}_{n}}^{2}}{{R}^{2}}\right),$$

(1)

where d is the distance between station observation and grid point. For the grid point inside the radius of influence of station observation, the first guess is calculated as:

$${g}_{1}(i,j)=\frac{\sum_{n=1}^{N}{W}_{n}S({x}_{n},{y}_{n})}{\sum_{n=1}^{N}{W}_{n}},$$

(2)

where S$\left({x}_{n, }{y}_{n}\right)$ is the observation value within the radius of influence. The first guess is added with a correction term given by

$$g_{{2~}} \left( {i,j} \right) = g_{1} \left( {i,j} \right) + \frac{{\sum\nolimits_{{n = 1}}^{N} {W^{\prime}_{n} \left[ {S\left( {x_{n} ,y_{n} } \right) - S^{\prime}\left( {x_{n} ,y_{n} } \right)} \right]} }}{{\sum\nolimits_{{n = 1}}^{N} {W^{\prime}_{n} } }},$$

(3)

where ${s}^{{\prime}}\left({x}_{n, }{y}_{n}\right)$ is the first guess obtained at observation point using an inverse distance interpolation technique. The W′ is the corrected weight given by

$${{W}^{{\prime}}}_{n}=\mathrm{exp}\left(- \frac{{{d}_{n}}^{2}}{\gamma {R}^{2}}\right),$$

(4)

where γ (gamma) is the convergence parameter and can vary between 0 and 1.