Aerosols and clouds are important components of the earth-atmosphere system with intricate physical, chemical, and optical properties. They have a significant influence on the atmospheric environment, climate change, and human health. Observing and studying the properties of aerosols and clouds are of great significance to gain insight into these issues. Currently, remote sensing technologies and methods are widely developed to observe aerosol and cloud properties, such as optical depth, extinction coefficient, and particle size distribution.

Lidar is one of the most useful active remote sensing tools due to its ability to detect the vertical distribution of the atmosphere. Among various types of lidars, ground-based Mie scattering lidar is the most popular one for cloud and aerosol detection with strong echo signals, simple system structure, and easy implementation. The development of Mie lidar began in the 1960s, and later multi-wavelength and polarization techniques were developed to more comprehensively detect scattering properties and particle sizes of aerosols and clouds. Nowadays, many lidar networks have been established for regional and global atmospheric environmental monitoring.

As the ground-based Mie lidar is becoming widespread, accurately retrieving their data is urgently required. However, retrieval is still facing many challenges that lead to large uncertainties. First, the correction of the overlap factor is crucial because the near-surface atmospheric information is often the most concerned. Second, the identification and extraction of cloud and aerosol layers from noisy lidar signals are essential for subsequent optical parameter retrieval and atmospheric research. Finally, data retrieval is a key step in lidar signal processing as it reveals the optical properties of aerosols and clouds. Hence, we mainly review the research progress in overlap factor correction, layer detection, and signal retrieval for the ground-based Mie lidar to guide future research and application.

The key challenges in Mie scattering lidar data processing include overlap factor correction, layer detection, and signal retrieval (Fig. 1). For overlap factor, the correction methods can be divided into experimental and theoretical methods. The experimental methods do not depend on the lidar system parameters but require the assumption of a uniform atmospheric distribution. The theoretical methods include analytical methods and ray tracing methods, which can guide the design of the lidar system. In addition, the overlap factor effect can be reduced more effectively by adjusting and improving lidar systems, such as dual field-of-view lidar and CCD side-scattering lidar.

For layer detection, the slope-based method can be directly applied to the raw lidar signal but is very sensitive to the noise. The threshold-based method is relatively more robust and commonly used to produce standard products (Fig. 4). However, tenuous layers may be missed because their signal intensity does not consistently exceed the threshold. The hypothesis test method based on the Bernoulli distribution decides whether the signal is a layer or not based on the estimated probability of it belonging to a layer. Studies have shown that its detection performance is superior to the threshold-based methods (Fig. 5).

For signal retrieval, the Fernald method is the most widely used but requires two parameters: the lidar ratio and the boundary value. The boundary value will directly affect the retrieval accuracy (Fig. 6) and can be determined by the fixed scattering ratio method, single-component fitting method, two-component fitting method, and joint observation method. Among them, the two-component fitting method can independently distinguish the contribution of atmospheric molecules, with excellent applicability and high accuracy. Furthermore, an incorrect lidar ratio will cause the overall retrieval deviation (Fig. 7). Methods for determining the lidar ratio mainly include the empirical method, aerosol optical depth (AOD) constraint method, and joint observation method. The popular AOD constraint method can obtain the lidar ratio mean value accurately but lacks its vertical profile distribution. The joint observation method using multiple vertical observations can provide a lidar ratio profile, but there are very few simultaneous vertical observations. In addition, many signal denoising algorithms have also been developed, but it is still a problem to evaluate their performance due to the lack of accurate observations as references.

Key issues such as overlap factor correction, layer detection, and signal retrieval still exist in ground-based Mie scattering lidar data processing. The development of new technologies such as dual field-of-view lidar and CCD side-scattering lidar provides more possibilities for low-overlap observation. The hypothesis test method can avoid one-size-fits-all empirical judgments and detect layers more accurately than other methods. In the retrieval, accurate boundary value selection requires avoiding simple assumptions and separating aerosol and molecule contributions. In addition, with the development of other vertical observations, the acquisition of lidar ratio profiles has become easier, which largely improves the retrieval accuracy of ground-based Mie scattering lidar.