This work explores the development of predictive models for the noise of a contra-rotating propeller (CRP) drone operating under real flight conditions. Using supervised machine learning techniques, empirical models are established to predict the time-dependent behavior of the overall sound pressure level (OSPL) and its A-weighted counterpart (OASPL) during ascent, hover, and descent manoeuvres. Experimental array-based acoustic data and comprehensive telemetry recorded during outdoor flight tests are employed to train and optimize the models, and subsequently assess their performance on unseen data. A set of meaningful telemetry-based observables is used as candidate input features for predicting both acoustic signals. These observables reflect the underlying flight dynamics and the environmentally induced perturbations to the vehicle dynamics. Statistical methods are employed to quantify the importance of each telemetry variable in predicting the acoustic signals, indicating that only a small set of variables is needed to achieve high predictive accuracy for both OSPL and OASPL using a reduced set of variables. Comprehensive evaluation of the models is conducted on unseen and highly variable acoustic data. The results reveal that the proposed models can accurately capture the time-dependent evolution of OSPL and OASPL during the studied manoeuvres. These findings provide a useful preliminary predictive tool for noise prediction during real UAV operations, as well as for noise-based trajectory optimization of single-UAV operations.

Highlights: What are the main findings? The trained ANN algorithm simultaneously retrieves soil moisture and vegetation optical depth from CYGNSS observations, showing strong agreement with reference satellite products (SMAP SM: R = 0.83, RMSE = 0.063 m ³/m ³; SMOS VOD: R = 0.89, RMSE = 0.088). ANN-derived VOD shows strong correlation with independent vegetation indicators—biomass (R~0.77), canopy height (R~0.95), Leaf Area Index (R = 0.96), and vegetation water content (R~0.90)—confirming reliable sensitivity to vegetation structure. What are the implications of the main findings? The combination of GNSS-R data with environmental variables enables reliable dual retrieval of soil moisture and vegetation optical depth, serving as a cost-effective, higher-resolution alternative/complement to SMAP and SMOS. Joint retrieval of SM and VOD enables improved characterization of land–atmosphere interactions, supporting hydrological, ecological, and climate applications. Accurate estimation of soil moisture (SM) and vegetation optical depth (VOD) is essential for understanding land–atmosphere interactions, climate dynamics, and ecosystem processes. While passive microwave missions such as SMAP and SMOS provide reliable global SM and VOD products, they are limited by coarse spatial resolution and infrequent revisit times. Global Navigation Satellite System Reflectometry (GNSS-R) observations, particularly from the Cyclone GNSS (CYGNSS) mission, offer an improved spatiotemporal sampling rate. This study presents a deep learning framework based on an artificial neural network (ANN) for the simultaneous retrieval of SM and VOD from CYGNSS observations across the contiguous United States (CONUS). Ancillary input features, including specular point latitude and longitude (for spatial context), CYGNSS reflectivity and incidence angle (for surface signal characterization), total precipitation and soil temperature (for hydrological context), and soil clay content and surface roughness (for soil properties), are used to improve the estimates. Results demonstrate strong agreement between the predicted and reference values (SMAP SM and SMOS VOD), achieving correlation coefficients of R = 0.83 and 0.89 and RMSE values of 0.063 m ³/m ³ and 0.088 for SM and VOD, respectively. Temporal analyses show that the ANN accurately reproduces both seasonal and daily variations in SMAP SM and SMOS VOD (R ≈ 0.89). Moreover, the predicted SM and VOD maps show strong agreement with the reference SM and VOD maps (R ≈ 0.93). Additionally, ANN-derived VOD demonstrates strong consistency with above-ground biomass (R ≈ 0.77), canopy height (R ≈ 0.95), leaf area index (R = 96), and vegetation water content (R ≈ 0.90). These results demonstrate the generalizability of the approach and its applicability to broader environmental sensing tasks.

This study presents a comprehensive mathematical framework for modeling Bluetooth Low Energy (BLE) beacon Received Signal Strength Indicator (RSSI) measurements to improve the estimation of RSSI-distance model parameters through the integration of stochastic and functional modeling. Although such approaches are well established in geodesy, they have been only sparsely applied to indoor positioning, particularly for characterizing RSSI dependencies arising from multipath propagation. Experiments were conducted using two types of BLE beacons in a controlled static outdoor environment to isolate signal behavior and facilitate reliable RSSI–distance modeling. Outliers were mitigated using a median filter based on local neighborhood comparisons. Stochastic modeling was then performed using the autocorrelation function (ACF) and partial autocorrelation function (PACF), revealing that RSSI noise exhibits short-memory, time-correlated behavior that is well represented by a second-order autoregressive AR(2) model. Measurement weights were estimated using Least Squares Variance Component Estimation (LS-VCE), followed by functional modeling for the estimation of environmental and beacon transmit-power parameters together with their uncertainties. The estimated candidate AR models were further compared using their maximized log-likelihood values, which confirmed that the AR(2) model provides the most appropriate representation of the RSSI noise process. The reconstructed RSSI achieved root mean square errors (RMSEs) of 1.12dBm and 0.50dBm for the IT002 and IT008 beacons, respectively, at a distance of 1m, increasing to 2.19dBm and 2.40dBm at 10m. These errors correspond to relative distance uncertainties of approximately 6–13% at 1m and 15–25% at 10m, demonstrating the effectiveness of the proposed framework for accurate RSSI modeling and parameter estimation.

This paper presents the implementation of the single-layer least-squares-based deep learning (LSBDL) model, optimized using the steepest descent method. As a showcase, the work numerically validates LSBDL’s performance in complex non-linear applications, such as surface fitting. LSBDL is proposed as a transparent deep learning solution, uniquely merging the theoretical robustness and quality control capabilities of the least squares (LS) method with the flexibility of deep learning (DL) models. Unlike conventional black-box DL architectures, the LSBDL framework naturally provides statistical quality assessment metrics, including the covariance matrix of estimated parameters and precision of predicted outcomes. This enables seamless model mis-specification and outlier detection using established reliability theory. The key focus of this study is the model’s demonstrated efficiency, accuracy, and performance in complex non-linear applications. In a complex surface fitting application, the implemented LSBDL model achieved a root mean square error (RMSE) of 0.0021, which is significantly lower than the simulated noise level. Furthermore, the estimated LS residuals are consistent with the simulated (and also estimated) standard deviation of σ = 0.01. The implemented model offers an effective, statistically grounded, and numerically efficient solution for handling complex non-linear problems, particularly those involving heterogeneous and correlated observations. All hyperparameters, initialization steps, optimization, and validation procedures are thoroughly discussed. The Matlab and Python code is freely available at: https://github.com/tud-dasaa/lsbdl.v1.

Given the importance of predicting the total ionospheric electron content (TEC), many studies have attempted to predict its spatiotemporal nature. In this study, a patch-based convolutional neural network with long short-term memory (CONVLSTM) (with patch sizes of 5 and 15), simple CONVLSTM, and 3D-U-Net models were used to predict the spatiotemporal nature of the next day's TEC data (next 12 samples). The proposed models use the spatiotemporal nature of the previous day's TEC data (previous 12 samples) along with temporal data such as AP, KP, DST, SN, and F10.7 to predict the next day's TEC data. The results showed that the 3D-U-Net model and then the model with patch size 5 had a higher generalization ability than the classical CONVLSTM architectures, while reducing the RMSE and MAE. The execution time of the program in the 3D-U-Net model has been significantly reduced compared to other models and it has also been able to better extract the microstructural features of TEC maps.

The increase in flight volumes in the aviation industry has significant socioeconomic implications that affect different aspects of our communities and economies. Although it has great economic benefits, it also causes annoyance and disturbance to communities living near airports. The latter requires understanding and prediction of the varying noise levels generated by various aircraft types. Noise assessment on a fleet level is traditionally achieved by using prediction models such as the DOC29. Such models need to be validated using real measurements. For Amsterdam Schiphol Airport, the so-called NOMOS (Noise Monitoring System) with 39 measurement stations is used for this purpose. We analyze the time series of these stations, collecting annual data for the period from 2006 to 2023. The main objective is to determine how the aircraft-generated noise at these stations can be assigned to 13 different aircraft types, taking into account the different noise levels produced by each aircraft type. This is performed by time series analysis of individual stations and the averaged time series over all stations. The results from two least-squares methods, namely unconstrained least squares (LS) and a proposed bounded least squares subject to weighted constraints (BLS + WC), are compared. The constraints are based on certification data as prior information in the least squares method, which is expected to enhance the model's performance. Based on the above two least squares methods, predictions are performed for 2022 and 2023. The results clearly demonstrate the superiority of the BLS + WC over the LS method. We further extend our analysis to predict noise levels for a hypothetical future year with more newer aircraft models. The results indicate a substantial reduction in the noise level compared to 2023. These findings can thus underscore the effectiveness of the proposed method in outperforming the LS and highlight the model's capability to forecast the impact of fleet modernization on noise reduction.

Seabed backscatter data acquired by the multibeam echosounder (MBES) have been identified as a valuable indicator of sediment properties and benthic community characteristics. However, developing robust change detection models with MBES backscatter remains challenging due to the high costs and limited spatial coverage of seabed ground truth data. Lack of absolute backscatter calibration also hinders the comparison between repeated MBES measurements. To mitigate these issues, we propose an unsupervised method to detect seabed changes by fitting a Gaussian Mixture Model to the backscatter difference between two datasets. A relative calibration is conducted based on a stable reference area to eliminate the impact of possible drifts in echosounder characteristics on the backscatter difference. We then model the unchanged class as a zero-mean Gaussian distribution, with its variance constrained by the backscatter uncertainty estimated from the reference area. By processing each incident angle individually, the angular range with the greatest ability for seabed change detection can also be investigated. We demonstrate the effectiveness of the proposed method through two case studies in the Dutch North Sea. The detected changes reveal seasonal and temporal variations in benthic communities, such as sand mason worms, and are consistent with the sediment movement in one of the study areas. This research highlights the value of MBES backscatter data for seabed change detection and provides a cost-effective solution for seabed habitat monitoring with acoustic measurements.

A-check maintenance scheduling is a complex and critical undertaking for airlines requiring an efficient allocation of resources. Current state-of-the-art focuses primarily on long-term A-check planning, typically targeting a longer scheduling horizon while foregoing individual task planning. This paper introduces a novel integrated approach for A-check scheduling at a seasonal level for an airline fleet, which accounts for both repetitive and one-off maintenance tasks. The A-check maintenance scheduling problem is formulated as a mixed-integer linear program (MILP), which optimises for minimal interval waste and timely initiation of one-off activities. Furthermore, we explore the scalability and flexibility of this problem by proposing three distinct distributed architectures. Subsets of maintenance tasks are scheduled by individual components, guided by a genetic algorithm (GA) acting as a global optimiser, with each architecture managing shared resources differently. We demonstrate our method with a case study from a major European airline using recent data of a fleet of wide-bodied passenger aircraft. While our MILP baseline produces comparable results to real-world schedules within minutes, the distributed architectures, despite their potential for scalability, generally underperform compared to the central planner. We analyse the degradation of solution quality across these distributed architectures, providing insights into their design limitations and the inherent indivisibility of the problem. We propose that our central MILP-based scheduler can be used by airlines as a decision-support tool for A-check task planning at the seasonal level.

The study of long-term GNSS time series provides valuable insights for researchers in the field of earth sciences. Understanding the trends in these time series is particularly important for geodynamic researchers focused on earth crust movements. Functional and stochastic models play a crucial role in estimating trend values within time series data. Various methods are available to estimate variance components in GNSS time series. The least squares variance component estimation (LS-VCE) method stands out as one of the most effective approaches for this purpose. We introduce an innovative method, which streamlines calculations and simplifies equations, and therefore significantly boosting the processing speed for diagonal(ized) cofactor matrices. The method can be applied to the GNSS time series of linear stochastic models consisting of white noise, flicker noise and random walk noise. Moreover, unlike the conventional approaches, our method experiences high computational efficiency even with an increase in the number of colored noise components in time series data. For GNSS time series, this variable transformation has been applied to both univariate and multivariate modes, preserving the optimal properties of LS-VCE. We conducted simulations on daily time series spanning 5, 10, 15, and 20 years, employing two general and fast modes with one and two colored noise components plus white noise. The computation time for estimating variance components was compared between the two modes, revealing a notable decrease in processing time with the fast mode.

Indoor positioning systems based on Bluetooth technology have gained significant attention with the widespread adoption of Bluetooth Low Energy (BLE). The rapid growth of BLE-enabled devices—now exceeding eight billion worldwide—has enabled the development of cost-effective, location-aware applications. The precision of BLE indoor positioning systems depends on several factors, among which the quality of received signal strength indicator (RSSI) measurements and the spatial deployment of sensors are most critical. Existing research has largely focused on improving RSSI accuracy through techniques such as multichannel measurements, outlier detection, Kalman filtering, and regression modeling. In this work, we examine two primary components that govern positioning precision: 1) the uncertainty in the RSSI-distance model parameters, including the path-loss exponent, reference power, and raw RSSI values; and 2) the geometry of sensor deployment, which directly affects estimation precision through the geometric dilution of precision (GDoP). We demonstrate that optimizing sensor placement using centroidal Voronoi tessellation (CVT) reduces GDoP and substantially improves positioning precision. Comparative experiments across two deployment scenarios, one based on CVT, confirm that CVT-based sensor configurations yield significantly higher precision in BLE RSSI-based indoor positioning.

Satellite-derived bathymetry (SDB) provides a cost-effective solution for coastal mapping, but challenges remain in model interpretability and uncertainty quantification. This study investigates the applicability of the least-squares-based deep learning (LSBDL) framework for SDB, leveraging its hybrid structure that integrates neural networks with the available least-squares theory to enhance model transparency. ICESat-2 photon-counting LiDAR was used to train depth estimation from Sentinel-2 multispectral imagery over an approximately 30 km × 30 km region of near-coastal bathymetry at Anegada, British Virgin Islands. ICESat-2 provided high-precision depth information, of which 80% were used for training and the remainder for validation. LBSDL depth estimation achieved a root-mean-square error (RMSE) of 2.74 m, representing around 10% of the maximum observed depth, with the best performance in the 2–15 m depth range. These findings demonstrate the potential of LSBDL for interpretable and reliable bathymetric mapping, highlighting ICESat-2 as a globally accessible training and validation source and advancing SDB capabilities for data-sparse coastal regions.

Accurately predicting Earth’s rotation rate, as represented by Length of Day (LOD) variations, is essential for applications such as satellite navigation, climate studies, geophysical research, and disaster prevention. However, predicting LOD is challenging due to its sensitivity to various geophysical and meteorological factors. Current methods, including statistical approaches, often struggle with short-term forecasting accuracy. In this study, we use Monte Carlo Singular Spectrum Analysis (MCSSA) to distinguish between deterministic and non-deterministic components within the LOD time series. The deterministic components are extended using the SSA prediction algorithm. To enhance robustness, we refine Allen and Smith’s methodology (testing significance of eigenmodes against an autoregressive (AR) (1) noise null hypothesis) by integrating an autoregressive moving average (ARMA) model to account for noise, providing valuable insights into the non-deterministic behaviors present in the series. We comprehensively evaluate our methodology through a comparative analysis. For long-term prediction (365 days), we compare our method against the combined LS and autoregressive (AR) method. For short-term prediction (next 10 days), we compare it against the results of the second Earth Orientation Parameters Prediction Comparison Campaign (second EOP PCC). Using the IERS 20 C04 time series, our hybrid model demonstrates a superior long-term prediction accuracy with a mean absolute error (MAE) of 0.201 ms/day on the 365th day. Additionally, the short-term prediction performance is comparable to the second EOP PCC results. These results illustrate that the proposed method efficiently predicts LOD, showing significant improvement in long-term accuracy and robustness in short-term forecasting.