Calculation of the aeolian sediment flux-density profile based on estimation of the kernel density

- It is the first time that a nonparametric method is used for the aeolian sediment flux-density profile.
- The importance of bandwidth selection is introduced with an example.
- In this nonparametric method, Gaussian kernel is better than exponential kernel for our purposes.
- The percentage of the total sediment mass flux in an arbitrary height interval of height can be achieved.
- This method may lead to the development of a suite of related analytical and modeling methods.

Aeolian sediment flux is an important issue of aeolian research. Parametric estimation is a traditional method in which aeolian sediment flux is estimated based on parameterization of a chosen equation. This method is simple, but has some limitations; specifically, it requires a priori assumptions about the density distribution that may not be correct. In this study, we applied a popular and extensively used, data-driven, nonparametric method called kernel-density estimation to calculate the aeolian sediment flux-density profile. Nonparametric methods make no prior assumption about the form of the density distribution to be estimated; instead, the aim is to obtain an empirical estimate from the data that can provably converge on the true density that would be obtained using an infinite sample size. Through the calculation of aeolian sediment flux based on kernel-density estimation, we determined that the key point in this method is not selection of the kernel function, but rather the selection of the optimal bandwidth, which is a difficult task. The results of our calculations showed that the method is both computationally feasible and acceptably accurate. Equally significantly, the idea of applying nonparametric methods to the calculation of aeolian sediment fluxes may lead to the development of a suite of other related analytical and modeling methods.