Precipitation is a key feature of the water cycle. Its quantification is crucial for various hydrometeorological applications. At the same time, precipitation is particularly variable in space and time, which makes quantitative estimation on high spatiotemporal resolutions difficult. Since simulations from numerical weather models cannot produce such high-quality quantification, the primary information sources are observations. These observations may suffice with regard to temporal resolution; however, the spatial coverage is generally limited. For hydrological applications such as runoff models, it is crucial to capture the spatial distribution, that is, the pattern of precipitation, accurately.Common observational data comprises dedicated sensors such as rain gauges and weather radars. Despite some limitations, for example, wind-related underestimation (Pollock et al., 2018) and limited spatial coverage, rain gauges can be considered very accurate precipitation sensors. Weather radars provide a high spatial resolution but have other disadvantages stemming from beam blockage, ground clutter, the measuring height above ground, and the sensitivity to the drop size distribution (Berne & Krajewski, 2013).

No abstract

Spatial structures of natural variables are often very complex due to the different physical chemical or biological processes which contributed to the emergence of the fields. These structures often show non‐Gaussian spatial dependence. Unfortunately, there are only a limited number of approaches that can explicitly consider non‐Gaussian behavior. In this contribution, a very flexible way of defining non‐Gaussian spatial dependence is presented. The approach is based on a kind of continuous deformation of fields with different Gaussian spatial dependence. Theoretical examples illustrate the methodology for a wide variety of non‐Gaussian structures. A real‐life example of groundwater quality parameters shows the practical applicability of the geostatistical model.

A meta-learning (ML) based time-varying channel estimation method with inter-carrier interference (ICI) cancellation is proposed for the high-speed mobile OFDM systems. To reduce the effect of the ICI caused by the Doppler shift on the accuracy of channel estimation, a improved transformation matrix is given to transform the transmitted signal with the comb-type pilot in the frequency domain into the one with the block-type pilot in the time domain, which can reduce the interference of the neighboring data symbols to the pilots. Since the ML network has ability to quickly adapt to the new channel scenario, it is employed to estimate the time-varying channel only by the received ICI-free pilots. To improve the practicability of the estimation model, the training target of the network is set as the channel estimation with high accuracy rather than the ideal channel state information. Simulation results show that the proposed method has high estimation accuracy and low computational complexity, and it is robust to the fast time-varying channel in the high-speed mobile scenarios.