Relative Root Mean Square Error Research Articles

Enhancing river discharge inversion for the midstream Jialing River (China) using the Google Earth Engine platform

ABSTRACT This study focused on the Jinxi section of the Jialing River, utilizing data from Landsat 8, Sentinel-2, and MODIS accessed via the Google Earth Engine (GEE) platform. The objective was to estimate runoff by applying three models: the improved Manning's formula (Model 1), the relationship fitting method (Model 2), and the C/M signal method (Model 3). The models were evaluated based on their accuracy in runoff inversion, the influence of hydraulic parameters, and their suitability for medium-sized rivers. The results indicated that all three models performed well in simulating runoff, with Nash–Sutcliffe Efficiency coefficients exceeding 0.90. The root mean square error (RMSE) for the improved Manning's formula, the relationship fitting method, and the C/M signal method were 50.2, 117.1, and 69.5 m³/s, respectively, corresponding to relative RMSE (RRMSE) of 4.71, 16.15, and 5.88%. It was observed that both the improved Manning's formula and the C/M signal method generally underestimated flow, while the relationship fitting method tended to overestimate it. Overall, the improved Manning's formula and the C/M signal method outperformed the relationship fitting method in terms of accuracy and applicability.

Automated estimation of individualized organ-specific dose and noise from clinical CT scans

Objective. Radiation dose and diagnostic image quality are opposing constraints in x-ray computed tomography (CT). Conventional methods do not fully account for organ-level radiation dose and noise when considering radiation risk and clinical task. In this work, we develop a pipeline to generate individualized organ-specific dose and noise at desired dose levels from clinical CT scans.Approach. To estimate organ-specific dose and noise, we compute dose maps, noise maps at desired dose levels and organ segmentations. In our pipeline, dose maps are generated using Monte Carlo simulation. The noise map is obtained by scaling the inserted noise in synthetic low-dose emulation in order to avoid anatomical structures, where the scaling coefficients are empirically calibrated. Organ segmentations are generated by a deep learning-based method (TotalSegmentator). The proposed noise model is evaluated on a clinical dataset of 12 CT scans, a phantom dataset of 3 uniform phantom scans, and a cross-site dataset of 26 scans. The accuracy of deep learning-based segmentations for organ-level dose and noise estimates was tested using a dataset of 41 cases with expert segmentations of six organs: lungs, liver, kidneys, bladder, spleen, and pancreas.Main results. The empirical noise model performs well, with an average RMSE approximately 1.5 HU and an average relative RMSE approximately 5% across different dose levels. The segmentation from TotalSegmentator yielded a mean Dice score of 0.8597 across the six organs (max = 0.9315 in liver, min = 0.6855 in pancreas). The resulting error in organ-level dose and noise estimation was less than 2% for most organs.Significance. The proposed pipeline can output individualized organ-specific dose and noise estimates accurately for personalized protocol evaluation and optimization. It is fully automated and can be scalable to large clinical datasets. This pipeline can be used to optimize image quality for specific organs and thus clinical tasks, without adversely affecting overall radiation dose.

3D LiDAR SLAM for natural guided navigation of automated guided vehicles in unstructured scenes

Abstract This paper proposes a 3D Light Detection and Ranging (LiDAR) simultaneous localization and mapping system for indoor unstructured scenes to improve the robustness and accuracy of automated guided vehicles. First, in the point cloud preprocessing stage, a curve voxel clustering method considering rarity point clouds is proposed, which effectively distinguishes adjacent objects while handling outliers, and achieves accurate clustering of rarity point clouds. Next, in the feature extraction stage, a curvature evaluation method based on principal component analysis is used to ensure the quality of feature extraction through the invariance of viewpoint and distance. Finally, a two-stage degeneration compensation strategy based on a recursive noise estimation adaptive Kalman filter is proposed, which significantly improves pose estimation accuracy by integrating LiDAR and an inertial measurement unit. Experimental results show that the proposed method achieves outstanding performance on the KITTI dataset, with average relative translation error and rotation error RMSE maintained at 0.7922% and 0.7513 deg/100 m, respectively. Additionally, the method achieves over 40% accuracy improvement in real indoor environments compared to LeGO-LOAM.

Refinement of Atomic Polarizabilities for a Polarizable Gaussian Multipole Force Field with Simultaneous Considerations of Both Molecular Polarizability Tensors and In-Solution Electrostatic Potentials.

Atomic polarizabilities are considered to be fundamental parameters in polarizable molecular mechanical force fields that play pivotal roles in determining model transferability across different electrostatic environments. In an earlier work, the atomic polarizabilities were obtained by fitting them to the B3LYP/aug-cc-pvtz molecular polarizability tensors of mainly small molecules. Taking advantage of the recent PCMRESPPOL method, we refine the atomic polarizabilities for condensed-phase simulations using a polarizable Gaussian Multipole (pGM) force field. Departing from earlier works, in this work, we incorporated polarizability tensors of a large number of dimers and electrostatic potentials (ESPs) in multiple solvents. We calculated 1565 × 4 ESPs of small molecule monomers and dimers of noble gas and small molecules and 4742 × 4 ESPs of small molecule dimers in four solvents (diethyl ether, ε = 4.24, dichloroethane, ε = 10.13, acetone, ε = 20.49, and water, ε = 78.36). For the gas-phase polarizability tensors, we supplemented the molecule set that was used in our earlier work by adding both the 4252 monomer and dimer sets studied by Shaw and co-workers and the 7211 small molecule monomers listed in the QM7b database to a combined total of 13,523 molecular polarizability tensors of monomers and dimers. The QM7b polarizability set was obtained from quantum-machine.org and was calculated at the LR-CCSD/d-aug-cc-pVDZ level of theory. All other polarizability tensors and all ESPs were calculated at the ωB97X-D/aug-cc-pVTZ level of theory. The atomic polarizabilities were developed using all polarizability tensors and the 1565 × 4 ESPs of small molecule monomers and were then assessed by comparing them to the 4742 × 4 ab initio ESPs of small molecule dimers. The predicted dimer ESPs had an average relative root-mean-square error (RRMSE) of 9.30%, which was only slightly larger than the average fitting RRMSE of 9.15% of the monomer ESPs. The transferability of the polarizability set was further evaluated by comparing the ESPs calculated using parameters developed in another dielectric environment for both tetrapeptide and DES monomer data sets. It was observed that the polarizabilities of this work retained or slightly improved the transferability over the one discussed in earlier work even though the number of parameters in the present set is about half of that in the earlier set. Excluding the gas-phase data, for the DES monomer set, the average transfer RRMSEs were 16.25% and 10.83% for pGM-ind and pGM-perm methods, respectively, comparable to the average fitting RRMSEs of 16.03% and 10.54%; for tetrapeptides, the average transfer RRMSEs were 5.62% and 3.95% for pGM-ind and pGM-perm methods, respectively, slightly larger than 5.41% and 3.61% of the fitting RRMSEs. Therefore, we conclude that the pGM methods with updated polarizabilities achieved remarkable transferability from monomer to dimer and from one solvent to another.

Assessment of Pine Tree Crown Delineation Algorithms on UAV Data: From K-Means Clustering to CNN Segmentation

The crown area is a critical metric for evaluating tree growth and supporting various ecological and forestry analyses. This study compares three approaches, i.e., unsupervised clustering, region-based, and deep learning, to estimate the crown area of Pinus eldarica Medw. using UAV-acquired RGB imagery (2 cm ground sampling distance) and high-density point clouds (1.27 points/cm3). The first approach applied unsupervised clustering techniques, such as Mean-shift and K-means, to directly estimate crown areas, bypassing tree top detection. The second employed a region-based approach, using Template Matching and Local Maxima (LM) for tree top identification, followed by Marker-Controlled Watershed (MCW) and Seeded Region Growing for crown delineation. The third approach utilized a Convolutional Neural Network (CNN) that integrated Digital Surface Model layers with the Visible Atmospheric Resistance Index for enhanced segmentation. The results were compared against field measurements and manual digitization. The findings reveal that CNN and MCW with LM were the most effective, particularly for small and large trees, though performance decreased for medium-sized crowns. CNN provided the most accurate results overall, with a relative root mean square error (RRMSE) of 8.85%, a Nash–Sutcliffe Efficiency (NSE) of 0.97, and a bias score (BS) of 1.00. The CNN crown area estimates showed strong correlations (R2 = 0.83, 0.62, and 0.94 for small, medium, and large trees, respectively) with manually digitized references. This study underscores the value of advanced CNN techniques for precise crown area and shape estimation, highlighting the need for future research to refine algorithms for improved handling of crown size variability.

A Flow-Speed Model for Motorways in England: Analysis Under Various Weather Conditions

This work proposes a single regime speed–flow model to fit the speed–flow relationship on the M25, London’s main motorway which is recurrently congested, especially near Heathrow Airport. The proposed model had a better performance compared with the existing classic models. A whole year’s field data on various sites of the M25 motorway were collected by the National Highways MIDAS (Motorway Incident Detection and Automatic Signalling) system and analysed. The proposed model was fitter on both four-lane and lane-by-lane conditions than the existing models, in terms of lower relative error and RMSE values and higher R2 values (minimum R2 = 0.79), which means the proposed model captured the speed–flow relationship better. In addition, the proposed model was used to fit traffic characteristics under different weather conditions and decided the threshold of the CM algorithms using the Gaussian function. The results showed that both free-flow speed and road capacity were significantly reduced by up to 7% and 11%, respectively, under different rainfall conditions, and that congestion management should be triggered in advance on rainy days. Further analysis of extensive data over a wider space and time is required to test the transferability of these findings to other contexts.

A framework for upscaling aboveground biomass from an individual tree to landscape level and qualifying the multiscale spatial uncertainties for secondary forests

ABSTRACT Secondary forests, a typical forest type in the sub-frigid zone of Northeast China, have significant potential for carbon sequestration. Accurate estimation of the Aboveground Biomass (AGB) of secondary forests and assessment of multiscale uncertainties are crucial for promoting Reduced Emissions from Deforestation and Degradation. This study developed a novel framework to upscale the AGB estimation from the tree to the landscape level and assessed multiscale uncertainties based on multi-platform laser scanning data and Unmanned Aerial Vehicle (UAV) hyperspectral images. The framework included two stages: (1) quantifying multiple uncertainties (uncertainties of individual tree crown delineation, individual tree parameters estimation, and tree species classification) in individual tree-based AGB estimation using Monte Carlo simulations; (2) upscaling the plot to the landscape level estimated AGB using the Nonlinear Simultaneous Equation (NSE) with error-in-variables and quantifying the uncertainties of model residuals, model parameters, and model independent variables. The findings revealed a high accuracy from tree to plot AGB estimation (R2: 0.75, Root Mean Square Error (RMSE): 6.65 Mg/ha, relative RMSE (rRMSE): 5.40%), with the total and relative uncertainties of 16.85 Mg/ha and 16.29%, respectively, with the highest uncertainty (9.73 Mg/ha) observed in tree species classification. The AGB estimation using NSE achieved an R2 of 0.69, with an RMSE of 9.91 Mg/ha and an rRMSE of 10.43% from the plot to landscape level; and the uncertainties caused by model parameters, independent variables, and residuals were 5.52 Mg/ha, 14.56 Mg/ha, and 25.25 Mg/ha, respectively, accounting for 3.46%, 24.09%, and 72.45% of the total uncertainty. This study develops a framework for large-scale AGB estimation of mixed forests based on the individual tree approach and uncertainty quantification of multiscale estimates and provides a foundation for precise forestry, sustainable forest management, and carbon neutrality.

Improving Forest Canopy Height Mapping in Wuyishan National Park Through Calibration of ZiYuan-3 Stereo Imagery Using Limited Unmanned Aerial Vehicle LiDAR Data

Forest canopy height (FCH) is a critical parameter for forest management and ecosystem modeling, but there is a lack of accurate FCH distribution in large areas. To address this issue, this study selected Wuyishan National Park in China as a case study to explore the calibration method for mapping FCH in a complex subtropical mountainous region based on ZiYuan-3 (ZY3) stereo imagery and limited Unmanned Aerial Vehicle (UAV) LiDAR data. Pearson’s correlation analysis, Categorical Boosting (CatBoost) feature importance analysis, and causal effect analysis were used to examine major factors causing extraction errors of digital surface model (DSM) data from ZY3 stereo imagery. Different machine learning algorithms were compared and used to calibrate the DSM and FCH results. The results indicate that the DSM extraction accuracy based on ZY3 stereo imagery is primarily influenced by slope aspect, elevation, and vegetation characteristics. These influences were particularly notable in areas with a complex topography and dense vegetation coverage. A Bayesian-optimized CatBoost model with directly calibrating the original FCH (the difference between the DSM from ZY3 and high-precision digital elevation model (DEM) data) demonstrated the best prediction performance. This model produced the FCH map at a 4 m spatial resolution, the root mean square error (RMSE) was reduced from 6.47 m based on initial stereo imagery to 3.99 m after calibration, and the relative RMSE (rRMSE) was reduced from 36.52% to 22.53%. The study demonstrates the feasibility of using ZY3 imagery for regional forest canopy height mapping and confirms the superior performance of using the CatBoost algorithm in enhancing FCH calibration accuracy. These findings provide valuable insights into the multidimensional impacts of key environmental factors on FCH extraction, supporting precise forest monitoring and carbon stock assessment in complex terrains in subtropical regions.

Effect of Patient Positioning on CT Number Accuracy: A Phantom Study Comparing Energy Integrating and Deep Silicon Photon Counting Detector CT.

Patient positioning during clinical practice can be challenging, and mispositioning leads to a change in CT number. CT number fluctuation was assessed in single-energy (SE) EID, dual-energy (DE) EID, and deep silicon photon-counting detector (PCD) CT over water-equivalent diameter (WED) with different mispositions. A phantom containing five clinically relevant inserts (Mercury Phantom, Gammex) was scanned on a clinical EID CT and a deep silicon PCD CT prototype at vertical positions of 0, 4, 8, and 12 cm. EID scans used 120 kV and rapid kV-switching DE techniques. CT number was calculated for air, water, polystyrene, iodine 10 mg/mL, and bone. Ideal CT numbers were calculated using the NIST XCOM database toolkit. Comparison of measured to ideal CT number utilized relative root mean square error (RMSE). Trends in CT number versus WED were compared using linear regression and statistical comparisons to test for differences in slope. No statistical difference of CT number with mispositioning was seen between acquisition modes. CT number fluctuation was larger due to WED than mispositioning for all material inserts. Water, iodine, and bone, for deep silicon PCD CT had statistically significant (P < 0.05) smaller slopes compared to EIDof CT number over WED for all tested mispositions. The accuracy of deep silicon PCD CT was higher than either SE or DE EID CT for all materials at all mispositions except for polystyrene. WED (ie, patient size) contributes to CT number fluctuation more than mispositioning. The change in CT number was significantly smaller, and CT number accuracy was higher for deep silicon PCD CT in this phantom study.

An in‐silico approach exploring sorghum source:sink balance across sorghum hybrids: How many leaves are enough?

AbstractPrevious literature documented an imbalance for sorghum [Sorghum bicolor (L.) Moench] between source (leaves) and sink (grains), favoring the source. Therefore, reducing leaf number, anticipating maturity, and placing the dry‐down with more favorable environment might be advantageous for producers to fit another crop in the rotation. The aims of this study were to (1) evaluate via in‐silico the effects of leaf removal during the grain filling and (2) explore those impacts using a field dataset for sorghum yield. For the first objective, the APSIM (Agricultural Production Systems Simulator) sorghum model was tested with four hybrids across 12 locations in the United States (2015–2023) resulting in an RRMSE (relative root mean squared error) of 25% for yield. As a second step, an APSIM defoliation module was developed using field data of one site‐year, demonstrating an RRMSE of 17% for yield. As a last step, the model was used to simulate the effect of sequential defoliations on yield, across 38 years of weather data (1984–2022), without showing any yield penalties when removing up to four leaves after flowering. Leaf area removal after flowering indicated a positive imbalance in source:sink ratio (i.e., source excess). For the second objective, a field dataset from 21 sorghum hybrids with different attainable leaf numbers and cycle duration did not result in significant yield differences. Early maturity hybrids with fewer leaves give farmers the opportunity to intensify crop sequences. Less focus in sorghum improvement for early relative to late maturing hybrids has been reported; therefore, there is still ample room for future yield gains.

Retrieval of Chinese fir tree parameters under different understory conditions with the integration of handheld and airborne Lidar data

ABSTRACT Accurate extraction of tree parameters is vital for the calculation of high-quality forest volume or biomass. The Light Detection and Ranging (Lidar) technology with its ability to acquire three-dimensional forest stand structures is an important data source for extracting tree parameters. However, complex forest stand structures due to dense canopy and abundant understory vegetation can seriously affect the extraction performance of tree parameters. This study selected Chinese fir plantations to examine how different forest stand structure conditions affected tree parameter extraction based on the integration of handheld laser scanning (HLS) and airborne laser scanning (ALS) data. This study also explored the ability of developing models using extracted tree height and crown width, or directly using the calculated variables from point clouds to estimate the diameter at breast height (DBH) and stem volume. The methods mainly consisted of two parts: (1) Selecting appropriate height to identify seed points for tree segmentation, and subsequently extracting tree parameters; (2) Taking tree height, crown width, and tree-level point cloud metrics as independent variables to establish regression and random forest models for estimating tree DBHs and stem volumes. Results showed that (1) tree DBHs were accurately extracted with relative root mean square error (RMSEr) of 7.1%−10.5% and stem volume with an RMSEr of 12.3%−16.8% under different understory conditions; (2) As stand structure became complex, direct extraction of tree DBHs from HLS data became a challenging task. However, developing a DBH estimation model using tree height and crown width proved feasible; (3) In the complex-understory condition, the volume estimation model utilizing point cloud variables achieved good performance with RMSEr of 28.0%. This research provides new insights for extracting tree parameters using HLS and ALS data, offering a potential replacement for field measurements of tree DBH, height, and stem volume for Chinese fir plantations.

Improving Solar Radiation Forecasting in Cloudy Conditions by Integrating Satellite Observations

Solar radiation forecasting is the basis of building a robust solar power system. Most ground-based forecasting methods are unable to consider the impact of cloud changes on future solar radiation. To alleviate this limitation, this study develops a hybrid network which relies on a convolutional neural network to extract cloud motion patterns from time series of satellite observations and a long short-term memory neural network to establish the relationship between future solar radiation and cloud information, as well as antecedent measurements. We carefully select the optimal scales to consider the spatial and temporal correlations of solar radiation and design test experiments at ten stations to check the model performance in various climate zones. The results demonstrate that the solar radiation forecasting accuracy is considerably improved, particularly in cloudy conditions, compared with purely ground-based models. The maximum magnitude of improvements reaches up to 50 W/m2 (15%) in terms of the (relative) root mean squared error (RMSE) for 1 h ahead forecasts. The network achieves superior forecasts with correlation coefficients varying from 0.96 at 1 h ahead to 0.85 at 6 h ahead. Forecast errors are related to cloud regimes, of which the cloud amount leads to a maximum relative RMSE difference of about 50% with an additional 5% from cloud variability. This study ascertains that multi-source data fusion contributes to a better simulation of cloud impacts and a combination of different deep learning techniques enables more reliable forecasts of solar radiation. In addition, multi-step forecasts with a low latency make the advance planning and management of solar energy possible in practical applications.

Effects of Turbulence Modeling on the Simulation of Wind Flow over Typical Complex Terrains

The correct prediction of the wind speed and turbulence levels over complex terrain is essential for accurately assessing wind turbine wake recovery, power production, safety, and wind farm design. In this paper, two modified RANS turbulence models are proposed, which are innovative variants of the conventional SST k-ω model and the linear Reynolds stress model (RSM) featuring optimized closure constants. Then, these two modified models and their origin models are applied to compare and analyze wind flows from a 3D hill wind tunnel experiment and two field measurements over typical complex terrain, including Askervein hill and Bolund island, with the aim of analyzing the sensitivity of wind flows to different RANS turbulence models. The study focuses on analyzing the effects of different turbulence models on the self-sustainability of wind speed and turbulent kinetic energy upstream of the computational domain and on the accuracy of wind flow prediction over complex terrain. The results show that our modified RSM model shows better agreement with the available experimental data on the upstream and leeward sides of all simulated hills. The wind speed on the leeward slope is particularly sensitive to the turbulence model, with a maximum difference in the relative root mean square error (RRMSE) that can reach 11% among the four models. The accuracy of the turbulent kinetic energy depends on the self-sustainability of the upstream turbulent kinetic energy and the predictive ability of the turbulence model for separated flows, and the maximum difference in the RRMSE of the four models can reach 47%. In addition, the advantages and disadvantages of the tested models are discussed to provide guidance for model selection during wind flow simulations in complex terrain.

Few-time-points time-integrated activity coefficients calculation using non-linear mixed-effects modeling: Proof of concept for [111In]In-DOTA-TATE in kidneys

PurposeThe purpose of this study is to investigate the accuracy of few-time-points (FTP) time-integrated activity coefficients (TIACs) in peptide-receptor radionuclide therapy (PRRT) using non-linear mixed-effects (NLME) modeling. MethodsBiokinetic data of [111In]In-DOTA-TATE in kidneys at T-1 = (2.9 ± 0.6) h, T-2 = (4.6 ± 0.4) h, T-3 = (22.8 ± 1.6) h, T-4 = (46.7 ± 1.7) h, and T-5 = (70.9 ± 1.0) h after injection were obtained from eight patients using planar imaging. The Sum-Of-Exponentials (SOE) function with four parameters was used, which was selected as the best model for the renal biokinetic data of [111In]In-DOTA-TATE. The parameters of the SOE function were fitted to the all-time-point data in the NLME framework to derive reference (rTIACs). FTP fits, which consist of all combinations of time points, are done to calculate the estimated TIACs (eTIACs). The accuracy of the FTP-NLME TIACs calculations was assessed by calculating the relative deviations (RDs) and relative root-mean-square errors (RMSEs) between the eTIACs and rTIACs. ResultsThe lowest (mean ± SD) of RDs for the single-, two-, three-, four-time point FTPs were (0 ± 8) % (T-4), (1 ± 6) % (T-3 and T-4), (3 ± 5) % (T-2, T-3 and T-4), and (0 ± 2) % (T-2, T-3, T-4, and T-5), respectively. The lowest RMSEs for the one-, two-, three-, and four-time point FTPs were 8 % (T-4), 6 % (T-3 and T-4), 5 % (T-2, T-3 and T-4), and 2 % (T-2, T-3, T-4, and T-5), respectively. ConclusionOur results showed that FTP-NLME in an example of [111In]In-DOTA-TATE could lead to a high accuracy of eTIAC across various time points, when incorporating time point T-4 = (46.7 ± 1.7) h.

Eficácia da ferramenta de progressão de arquivos climáticos CCWorldWeatherGen

Abstract This paper aimed to analyze the effectiveness of the CCWorldWeatherGen tool, focusing on climate change in São Paulo, São Paulo State, Brazil. For this, dry-bulb temperature, relative humidity, global solar radiation, and wind speed data from the test reference year weather file (1954) and the CCWorldWeatherGen file for the 2020 period (representing the 2011-2040 period) were compared with observational data collected between 2011 and 2023 by the Meteorological Station of the Institute of Astronomy, Geophysics, and Atmospheric Sciences of the University of São Paulo. The accuracy of variables predicted using weather files was evaluated using five statistical measures of error. Annual relative root mean square error (RRMSE) values for dry-bulb temperature, relative humidity, global solar radiation, and wind speed in the morphed weather file were 17.04% (good), 17.95% (good), 31.57% (poor), and 224.44% (poor), respectively. It is concluded that CCWorldWeatherGen is suitable for generating future weather files with complete information, mainly for its practicality. However, this approach requires caution, as sequences depend on the consistency of the weather file used as a basis.

Concepts Towards Nation-Wide Individual Tree Data and Virtual Forests

Individual tree data could offer potential uses for both forestry and landscape visualization but has not yet been realized on a large scale. Relying on 5 points/m2 Finnish national laser scanning, we present the design and implementation of a system for producing, storing, distributing, querying, and viewing individual tree data, both in a web browser and in a game engine-mediated interactive 3D visualization, “virtual forest”. In our experiment, 3896 km2 of airborne laser scanning point clouds were processed for individual tree detection, resulting in over 100 million trees detected, but the developed technical infrastructure allows for containing 10+ billion trees (a rough number of log-sized trees in Finland) to be visualized in the same system. About 92% of trees wider than 20 cm in diameter at breast height (corresponding to industrial log-size trees) were detected using national laser scanning data. Obtained relative RMSE for height, diameter, volume, and biomass (stored above-ground carbon) at individual tree levels were 4.5%, 16.9%, 30.2%, and 29.0%, respectively. The obtained RMSE and bias are low enough for operational forestry and add value over current area-based inventories. By combining the single-tree data with open GIS datasets, a 3D virtual forest was produced automatically. A comparison against georeferenced panoramic images was performed to assess the verisimilitude of the virtual scenes, with the best results obtained from sparse grown forests on sites with clear landmarks. Both the online viewer and 3D virtual forest can be used for improved decision-making in multifunctional forestry. Based on the work, individual tree inventory is expected to become operational in Finland in 2026 as part of the third national laser scanning program.

Assessment of wind energy potential along the Egyptian Mediterranean Coast

BackgroundCurrently, there is no wind park operating along the Egyptian Mediterranean Coast. Therefore, this study aims to find suitable locations for such projects. Wind data from five coastal meteorological stations were used. These are Marsa Matruh (MM), Ras El-Tin (RE), Abu Qir (AQ), Port Said (PS), and Arish (Ar), in that order from west to east. The wind regime dataset, comprising velocity and direction measurements at a 10-m elevation, was collected from January 2007 to December 2022 (16 years), with a complete record of all data points. The Weibull distribution function, along with its different parameters, was used to characterize wind energy along the Egyptian Mediterranean Coast. The coefficient of determination (R2), root mean squared error (RMSE), and relative root mean squared error (RRMSE) for the Weibull parameters, along with the relative percentage errors (RPE) for the wind power density were calculated to assess the concordance between outcomes derived from observed data and those predicted by the Weibull function.ResultsResults revealed that the dominant wind direction along the Egyptian Mediterranean Coast was the NNW to N wind, except at Ar where the dominant wind was S. The wind velocity range of 4–6 m/s dominated RE, AQ, and PS. At MM and Ar, this was reduced to 2–4 m/s. The analysis of wind power density outlined significant insights into the potential for wind energy generation in the region. The overall analysis showed that AQ and PS were potentially the most suitable locations for wind energy projects. However, the high variability at the AQ site required robust system designs to manage the fluctuating wind conditions. PS might be more suitable for projects prioritizing stability and consistency over maximum energy output. Although Arish, characterized by its lower wind power density, may be less conducive for large-scale wind energy projects, it could still be viable for smaller installations or when integrated with other renewable energy sources.ConclusionsThe different statistical indices reflected good model fitting, displaying the reliability of the Weibull distribution as a tool for preliminary wind resource assessment along the Egyptian Mediterranean Coast and facilitating accurate predictions of wind power availability.

A novel framework for multi-layer soil moisture estimation with high spatio-temporal resolution based on data fusion and automated machine learning

High spatiotemporal resolution monitoring of multi-layer soil moisture (SM) is crucial for optimizing agricultural water management and precision irrigation strategy. However, achieving high temporal resolution at a 30 m spatial scale remains challenging given the confine of current satellite sensors. To overcome this, we developed an innovative framework synergizing multi-source remote sensing data, reanalysis data, auxiliary information (topography and soil texture), and ground-based SM observation. Initially, we generated seamless 30 m resolution metrics, including the normalized difference vegetation index (NDVI), land surface temperature (LST), and surface albedo, by employing the modified neighborhood similar pixel interpolator (MNSPI) in conjunction with the enhanced spatial and temporal adaptive reflectance fusion model (ESTARFM). These variables, combined with reanalysis data, auxiliary data, and ground-based SM observations, were input into an Automated Machine Learning (AutoML) workflow to estimate SM at 0–20, 20–40, and 40–60 cm soil layers. Validation conducted in the People's Victory Canal irrigation area revealed depth-dependent prediction accuracy, with Pearson correlation coefficient (R) values of 0.806, 0.772, and 0.680, root mean square errors (RMSEs) of 0.038, 0.047, and 0.054 cm³/cm³, and relative root mean square errors (RRMSEs) of 16.170 %, 20.346 %, and 22.689 % for the 0–20, 20–40, and 40–60 cm soil layers, respectively. This framework shows significant potential for enhancing water resources management at the field scale by providing accurate, high-resolution SM estimates across multiple depths.

Estimating forest litter fuel load by integrating remotely sensed foliage phenology and modeled litter decomposition

Litter on the forest floor, or from a fire perspective the litter fuel load (LFL), is a key driver of the occurrence and spread of surface fires and an important regulator of forest fire behavior. High-quality spatiotemporal LFL data are essential for modeling fire behavior and assessing fire risk in forest ecosystems. Traditionally, LFL is estimated from ground-based measurements, but they are difficult to implement on large spatial scales. While remote sensing techniques have the advantage of large-scale observation, they encounter challenges in retrieving LFL because forest canopies generally block signals from the forest floor. Here we present a new method based on modeled litter accumulation to estimate LFL dynamics, integrating litterfall influx from the forest canopy and decomposition outflux through a mass balance approach. Annual litterfall was estimated based on seasonal changes in foliage fuel load which are retrieved from Landsat imagery and a radiative transfer model, while the decomposition rate was derived from meteorological data. Litterfall and decomposition were quantified over the past 20 years with the difference between the two being LFL accumulating over time. We validated the estimated LFL using 105 ground-based measurements in Liangshan Yi Autonomous Prefecture, China, and this validation demonstrated a reasonably strong performance for estimating LFL (R2 = 0.67, root mean squared error (RMSE) = 2.56 Mg ha−1, relative RMSE = 31.61 %). Our method integrates remote sensing-based foliage phenology with the ecological process of LFL accumulation, enabling large-scale LFL monitoring for forest fire risk assessments.

Explainable artificial intelligence-machine learning models to estimate overall scores in tertiary preparatory general science course

Educational data mining is valuable for uncovering latent relationships in educational settings, particularly for predicting students' academic performance. This study introduces an interpretable hybrid model, optimised through Tree-structured Parzen Estimation (TPE) and Support Vector Regression (SVR), to predict overall scores (OT) utilising five assignments and one examination mark as predictors. Neural Network-based, Tree-Based, Ensemble-Based, and Boosting-based methods are evaluated against the hybrid TPE-optimised SVR model for forecasting final examination grades among 492 students enrolled in the TPP7155 (General Science) course at the University of Southern Queensland, Australia, during the 2020-2021 academic year. Additionally, Local Interpretable Model-agnostic Explanations (LIME) and SHapley Additive explanations (SHAP) techniques are employed to elucidate the inner workings of these prediction models. The findings highlight the superior performance of the proposed model, exhibiting the lowest Root Mean Squared Error (RMSE) and Relative Root Mean Squared Error (RRMSE), as well as the highest Willmott's index (WI), Legates–McCabe index (LM), and Nash–Sutcliffe Efficiency (NS). With assignment and examination marks identified as pivotal predictors of OT. SHAP and LIME analyses reveal the examination score (ET) as the most influential feature, impacting predicted OT by an average of ±4.93. Conversely, Assignment 1 emerges as the least informative feature, contributing merely ±0.64 to OT predictions. This research underscores the efficacy of the proposed interpretable hybrid TPE-optimised SVR model in discerning relationships among continuous learning variables, thereby empowering educators with early intervention capabilities and enhancing their ability to anticipate student performance prior to course completion.