Achieving accurate and low-latency spectrum sensing on resource-constrained devices is essential but very difficult. Traditional In-phase and Quadrature (I/Q)-based and the ShortTime Fourier Transform (STFT)-based methods fail to balance the computational overhead and classification accuracy. In this paper, we propose a novel framework -Spectrum Painting (SP)- which enables on-device signal classification with low latency and high accuracy. We design new signal processing methods to compress spectrograms while keeping global signal features and augmenting the salient features of small objects. SP achieves high-accuracy signal classification, assisted further by our proposed Dual-channel Convolutional Neural Network (DualCNN). We collect diverse datasets to evaluate the proposed SP, including synthesized data, and testbed data (from up to 18 commodity devices) obtained from real-world environments in the wild and office settings. Experimental results of SP running on Raspberry Pi 4B show a great reduction in latency up to 20 × while maintaining a 95% accuracy. Furthermore, SP demonstrates superior performance within both the centralized learning architecture and the Federated Learning (FL) architecture. For example, the challenging cross-environment evaluation of the SP in the iid-FL scenario yields a substantial accuracy improvement, on average from 24.6% to 83.8%.

Distance measurement using frequency sweeping interferometry is an absolute distance measurement technique that allows for high accuracy over long distances. Notwithstanding, the measurement accuracy is affected by laser sweeping nonlinearity and limited sweeping range. In this work, an optimized post-processing linearization method is demonstrated to realize high-accuracy arbitrary distance measurement using a laser with small modulation range. The interference signal is sparsely resampled to eliminate the influence of the sweeping nonlinearity, and the absolute distance is obtained by analyzing the phase of the resampled signal. In the measurement system, a high-finesse Fabry-Pérot cavity placed in vacuum is used as the measurement reference, so the effect of dispersion mismatch is negligible. Moreover, the distance measurement result is determined by the linear fit of the phase of each resampled point. Therefore, the influence of target vibration and other external random noise can be partially eliminated, and the reliability of the result is high. In the experiment, the sweeping range of the laser source is only 88 GHz. Comparing with a fringe-counting interferometer, the standard deviation of the residual errors is 34 µm within a distance of 6.7 m.

Management of large hydrologic datasets including storage, structuring, clustering, indexing, and query is one of the crucial challenges in the era of big data. This research originates from a specific problem: time series extraction at specific locations takes a long time when a large multidimensional (MD) dataset is stored in the NetCDF classic or the 64-bit offset format. The essence of this issue lies in the contiguous storage structure adopted by NetCDF. In this research, NetCDF file-based solutions and a MD array database management system applying a chunked storage structure are benchmarked to determine the best solution for storing and querying large MD hydrologic datasets. Expert consultancy was conducted to establish benchmark sets, with the HydroNET-4 system being utilized to provide the benchmark environment. In the final benchmark tests, the effect of data storage configurations, elaborating chunk size, dimension order (spatio-temporal clustering) and compression on the query performance, is explored. Results indicate that for big hydrologic MD data management, the properly chunked NetCDF-4 solution without compression is, in general, more efficient than the SciDB DBMS. However, benefits of a DBMS should not be neglected, for example, the integration with other data types, smart caching strategies, transaction support, scalability, and out-of-The-box support for parallelization.

Management of large hydrologic datasets including storage, structuring, indexing and query is one of the crucial challenges in the era of big data. This research originates from a specific data query problem: time series extraction at specific locations takes a long time when a large multidimensional dataset is stored in non-chunked NetCDF classic or 64-bit offset format. The essence of this issue lies in the contiguous storage structure adopted by NetCDF. In this research, NetCDF file based solutions and a multidimensional (MD) array database management system (DBMS) applying chunked storage structure are benchmarked to determine the best solution for storing and querying large hydrologic datasets. To achieve this, expert consultancy was conducted to establish benchmark sets. To guarantee a fair benchmark test environment, HydroNET-4 system was utilized and adapters for NetCDF files and SciDB were developed to manage and query data. In final benchmark tests, effect of data storage configurations such as chunk size and compression on query performance is also explored. Results indicate that SciDB arrays utilizing small chunk sizes show favorable performance. However with current implementation of SciDB, large numbers of small chunks cause huge overload of main memory which constraints SciDB's scalability. Compression of SciDB can either have negative or no effect on query performance, while it causes significant query degradation to NetCDF-4 solution. The research illustrates that for big hydrologic array data management, the properly chunked NetCDF-4 solution without compression is in general more efficient than the SciDB DBMS. So under current big data environment, traditionally adopted file-based hydroinformatic solutions can still be applicable after proper updating.

The hydrodynamic characteristics of a delta or estuary are mainly governed by discharges of rivers and water level at the sea (or lake) boundaries. A joint probability approach is widely applied to quantify the high water level frequency in deltas. In the approach the relevant hydrodynamic loading variables, namely the astronomical tides, the wind induced storm surge and the river flows, are jointly investigated. The joint probability distribution is used to generate a large number of scenarios of boundary conditions which can drive a deterministic model to derive the water levels at locations of interest. The resulting water levels as well as their associated joint probabilities can be inverted to the high water level frequency curve. However, in the joint probability distribution, marginal distributions may contain large statistical uncertainties due to their relevant parameters being estimated from a limited length of data. In the case of the Rhine Delta, a nonparametric bootstrap method is applied to quantify the statistical uncertainties in three critical marginal distributions: wind induced storm surge peak level, wind induced storm surge duration and River Rhine discharge. The uncertainties are incorporated into the marginal distributions with a Monte Carlo integration method. Further the uncertainty-incorporated marginal distributions are used for the high water level frequency assessment. Compared to previous studies, water levels for given return periods are much higher. The uncertainty differs in each marginal distribution and its impact on the high water level frequency curve also varies.