Estimation of wet variable’s background error information for regional model

In order to improve the humidity analysis of the WRF model, a new multivariable balance constraint scheme has been introduced to estimate wet variable's background error information from a series of historical forecasts. The new scheme consists of three critical procedures: physical transformation, vertical transformation and horizontal transformation. By removing the balanced part associated with other control variables, the unbalanced part of relative humidity is used as the new wet control variable. Statistical results show that relative humidity's background error structure appears an obviously localized characterization, which has a large negative value on model high level in the vertical direction and a stable characteristic length scales about 20km in the horizontal direction.