Retrieval And Assimilation Of Boundary Layer Height Based On Ground-based Lidar Data

The atmosphere boundary layer(ABL)is the lowers part of the troposphere that is directly influenced by Earth's surface and responds to surface forcing over a short period of time,it is closely related to human activities and the whole ecosystem.As an important length scale characterizing the boundary layer,the height of ABL(ABLH)is an important input parameter for the medium-and small-scale numerical weather/air-pollution prediction models as well as for the climate models,accurate estimation of which is critically necessary.Presently,the ABLH can't be directly measured,the estimation of which mainly relies on the measurements of vertical profiles of some suitable variables such as potential temperature,specific humidity,wind speed and the aerosol concentration.As a powerful active remote sensing instrument,aerosol lidar is widely used in ABLH retrieval,although many algorithms have been proposed and developed to improve the accuracy of ABLH retrieval in last few decades,accurately retrieving the ABLH based on lidar data still remains a challenge especially under complex scattered-signal structures caused by cloud layers and residual aerosol layers.In addition,presently a great number of aerosol lidars have been deployed and have established networks in many countries and regions,providing a great number of valuable observations with high spatial-temporal resolution,assimilating the data into numerical models will undoubtedly extend its application value.The lidar data are unconventional observations,when can't be directly assimilated,it may be an effective but challenge way to assimilate the lidar-retrieved ABLH into the numerical models.This study will explore the approach to improve the accuracy of ABLH retrieval from ground-based lidar,and based on the correlation analysis between ABLH and atmospheric variables,the ABLH assimilation study will be preliminarily carried out.Based on the Micro-Pulse lidar(MPL)data over SACOL(Semi-Arid Climate Observatory and Laboratory)site in Lanzhou University,the study firstly gives a comparison of presently available techniques for the lidar retrieval of ABLH with case analysis.Then,under both clear-sky and multiple layer(cloud,residual layers)conditions,the curve fitting and wavelet covariance transform methods are used to retrieve the ABLH over SACOL,the daytime ABL development process and forcing mechanisms are also analyzed under different conditions.In clear-sky condition,two algorithms are directly used to retrieve ABLH,the interference of the local noise signal is eliminated through considering temporal continuity.The results show that lidar-retrieved result is close to the thermal ABLH determined by the gradient of potential temperature based on vertical profiles of atmosphere provided by a nearby radiosonde site,the correlation coefficient between the resulted ABLH on 41 clear-sky cases based on two instruments exceeds 0.9.The lidar-retrieved ABLH could clearly reflect the diurnal development process of the daytime ABL,which is mainly forced by the thermodynamic turbulence,the diurnal variation of the ABLH basically responds to the variation of the surface buoyancy production of the surface turbulent kinetic energy.Meanwhile,the temperature inversion structure severely limits the development of daytime ABL.In conditions of cloud and residual layers,a new objective technique is proposed in this study,presenting as dynamically determining a top limiter altitude according to the location of cloud and residual layers to reduce the influence of multiple layers.The multiple layers are identified according to the characteristics that the strong positive and negative gradients of scattered signal occur in pairs at the top and bottom boundaries of the layers.Meanwhile,the scattering of the lasers by the cloud(or residual layer)is usually much stronger than(or close to)the scattering by aerosol in the ABL.In addition,for cloud layers,whether they overlay the ABL or are decoupled from the ABL are classified according to the vertical continuity of the signal below the cloud base,specifically,whether there is strong negative signal gradient(less than-2)reflecting the information of the ABL top.When the cloud is above(or overlay)the ABL top,the cloud base(or the altitude where a positive signal gradient first occurs above the cloud upper edge)is determined as the top limit height;for the residual layer,the top limiter is the altitude where the greatest positive gradient occurs below the RL top.Below the top limiter altitude,the ABLH is determined by the wavelet transform method and the curve fitting method.Similar to under clear-sky situation,the temporal continuity for the results is also considered to eliminate the local noise.Comparison between the lidar-retrieved and radiosonde-determined ABLH shows that the upper limiter determined in this study effectively eliminates the interference of the cloud layers on lidar-retrieval of ABLH,the correlation coefficient between twomeasurements-determined ABLH on 20 cloudy cases exceeds 0.7.In terms of diurinal variation,it can be seen that the interferences produced by residual layers on lidar-retrieval ABLH is significantly reduced,meanwhile,the daytime ABL development processes are clearly presented and could be reasonably explained by the space-time distribution of equivalent potential temperature and the diurnal variation of surface turbulence kinetic energy(shear generation and buoyancy production).In addition,in cloudy conditions,the ABL development may be affected by the radiation forcing caused by the cloud layers,the development mechanisms of the ABL may be more complex than in clear sky condition.The major difficulty for ABLH assimilation is the determination of complicated nonlinear observation operator,the ensemble Kalman filter(EnKF)assimilation method resolves the difficulty in implementation of adjoint matrix in variational assimilation methods.Different from the steady invariant background covariance used in the 3D-VAR,EnKF calculates the flow-and location-dependent error covariance based on ensemble forecast members.A simplified version of EnKF is used in this study,the ensemble square root filter(EnSRF),which eliminates the sampling error related to the estimation of the observation error covariance in the EnKF.Before assimilation,the correlation between ABLH and atmospheric variables,as well as the spatial distribution of the correlation coefficients are statistically analyzed.The results show that the ABLH significantly correlates to potential temperature(or perturbation potential temperature),specific humidity(or water vapor mixing ratio)and the perturbation potential height,but weakly correlates to three-dimensional wind fields.In the vertical direction,the correlation coefficients between ABLH and physical variables is nearly constant below the ABLH,and the correlation is stronger than that in the free atmosphere,there are obvious transition zones near the ABL top.In the horizontal direction,the strength and properties of the correlation between ABLH and model variables are related to the type of surface.The higher correlation coefficients occur over regions with the same surface type.Finally,based on correlation statistics,using the Weather Research and Forecast(WRF)model and EnSRF assimilation technique,a preliminary exploration of the ABLH assimilation is carried out through the observation system simulation experiments(OSSEs).This study mainly concentrates on the selection of suitable analysis variables,the defining of the influence radii including both vertical and horizontal localization.The results show that daytime ABLH assimilation cloud significantly improve the simulation and forecast of the model variables including perturbation potential temperature(ptprt),water vapor mixing ratio(qv)and potential height(ph)only within the ABL,while hardly no effect is produced on three variables above the ABLH;the daytime ABLH assimilation has no obvious effect on the vertical velocity but can cause the horizontal wind fields to deviate from their "truth" at all vertical model levels.When the directly updated analysis variables only including ptprt,qv and ph,and the vertical influence radius is defined the same as the ABLH,the daytime ABLH assimilation cloud enhance the vertical turbulent mixing intensity,producing a warmer and drier ABL.Meanwhile,inducing stronger entrainment process and the enhanced downward momentum flux transport over ABL top,finally,the wind fields within the ABL is correspondingly strengthen and close to the its “truth”(except near the boundaries of the study area).Therefore,during daytime ABLH assimilation process,the directly updated analysis variables can only contain ptprt,qv and ph,the vertical localization radius should be determined the same as or slightly higher than the ABLH.The height of nocturnal stable ABL is generally far lower than daytime thermaldynamical ABLH,the results show that the influence of nocturnal ABLH assimilation on simulation and forecast of model variables is much less than daytime.When the nocturnal ABLH is assimilated,only slight improvement will be made to the horizontal wind fields,and the vertical influence radius should be the same as or a little higher than the ABLH.Thus,other model variables such as ptprt,qv and ph could be slightly improved due to the dynamic adjustment during model integration process.In the aspect of horizontal localization,the results show that the assimilation of ABLH over land surface could cause negative effect on ptprt and qv over the lake area,therefore,the influence of surface type should be considered during ABLH assimilation,the ABLH over land(or water)only affects the model variables over land(or water).Setting different horizontal localization radii,the results show that the ABLH assimilation cloud significantly improve the simulation and forecast of ptprt,qv and ph over the model grids within a certain horizontal range.When there is only one observation point,three variables cloud be effectively updated within a region of 20dx(approximately 60 km).When there are several observation points,the scale of horizontal influence radius is related to the sparsity of observation stations.The main function of localization is to reduce the influence of spurious correlation over long distance.Theoretically,there should be an optimal horizontal localization scale.However,because of the small experimental simulation area with uniform underlying surface,the optimal horizontal localization scale does not appear in the sensitivity test.The experimental results in this study show that,when the boundary layer is assimilated over the region with uniform underlying surface type,the horizontal influence radius should be 1-2 times of the distance between observation points when both the assimilation influence and computing cost are taken into account.