Research on the implementation of GPS radio occultation one-dimensional bending angle operator in four-dimensional variational data assimilation system

Abstract

GPS radio occultation data are implemented in the numerical weather prediction as a kind of very important observation data because of its excellent characteristics, which makes a remarkable positive effect in the system. In this article, we design an implement method and a quality control strategy of one-dimensional bending angle operator in four-dimensional variational data assimilation system, and we also describe the steps of implementing one-dimensional bending angle forward operator and its linear and adjoint operators in four-dimensional variational data assimilation system. The author chose Reger’s three term expression and its best average coefficients are chosen to calculate refractivity values. In this article, we introduce an observation error model of one-dimensional bending angle operator through the statistics of GPS radio occultation observation data of August, 2012. A comparison of experiment of the analysis field between in the cases with and without GPS radio occultation data is made. It is concluded that GPS radio occultation data plays a positive role in the global areas, especially in the low and middle latitude areas higher than 10 km. This conclusion is the foundation of the following-up GPS radio occultation data assimilation verification experiments.

Impact of GNSS radio occultation bending angle data assimilation in YH4DVAR system

“…Without considering the horizontal gradient in the through path of radio, the equation of 1D GNSS bending angle operator described in Ref. [9] is given as Eq. ( 1) with the assumption that the atmosphere distribution is spherical symmetric…”

“…With the designed observation error model in Ref. [9], the analysis fields have been generated again for the period from 1 August 2012 to 31 August 2012. A statistical test has been done for the fractional differences of the standard deviations of bending angle between GNSS RO and model simulations.…”

“…The experiments are performed for the period 1 August 2012-31 August 2012 using YH4DVAR data assimilation system with the observation operator of one-dimensional bending angle implemented in it. [9] Figure 5 The bias values of geopotential height analysis fields in the Northern Hemisphere exhibit little impact at the surface or at the low troposphere, decrease from the height of 500 hPa until the top level of the model. The values' decreasing denotes the impact of GNSS RO observations is increasing.…”

“…The one-dimensional bending angle operator was simultaneously implemented in YH4DVAR. [6][7][8][9] The bending angle data assimilation is performed on observation levels with geopotential height considered in four-dimensional variational data assimilation system rather than three-dimensional variational data assimilation system.…”

New data assimilation system DNDAS for high-dimensional models

“…Variational principles for various differential equations have been investigated for centuries. [14] Currently, the most popular data assimilation methods are 3/4D-Var [15][16][17][18][19][20][21][22] and Kalman filter. [23][24][25] The aim of data assimilation is to provide accurate initial state information to the discrete partial differential equations controlling the motions of atmosphere or ocean.…”