Environmental Modelling & Software最新文献

Attaining resource security in the water, energy, food, and ecosystem (WEFE) sectors, the WEFE nexus, is paramount. This necessitates the use of quantitative modelling, which presents many challenges, as this is a complex system acting at the intersection of the physical- and social sciences. However, as WEFE data is becoming more widely available, data-driven methods of modelling this system are becoming increasingly viable. Here, we discuss two main problems in WEFE nexus modelling: system identification and control. System identification uses Machine Learning algorithms to obtain dynamical models from data and have shown promise in many disciplines with similar characteristics as the nexus. Meanwhile, control algorithms manipulate a system to achieve objectives and are becoming instrumental in shaping nexus policy. Despite the promise of these algorithms, data-driven modelling is a vast and daunting field, and so here we provide an introductory overview of this field, with emphasis on nexus applications.

The present study demonstrates the capability of an inversion modelling scheme so-called the “triangle” to retrieve spatiotemporal estimates of surface energy fluxes and soil surface moisture (SSM) at high resolution using ASTER satellite imagery synergistically with SimSphere land biosphere model. In addition, as a further objective of this study is to examine the use of the technique for retrieving the Evaporative (EF) and the Non-Evaporative (NEF) Fractions as representations of the daytime average fluxes. The applicability of the investigated technique, is demonstrated for sixteen calendar days of year 2011 using in-situ data acquired from nine CarboEurope sites representing a variety of climatic, topographic and environmental conditions. Results indicated a close agreement between all the inverted parameters and the corresponding in-situ data. SSM predicted maps showed a small bias of 0.08 vol vol⁻¹, a scatter of 0.18 vol vol⁻¹ and a RMSD of 0.19 vol vol⁻¹. The predicted LE fluxes showed a relatively low overall agreement (RMSD = 65.10 Wm^-2), whereas for H flux reported RMSD was 85.02 Wm^-2. The results also confirmed the ability of the investigated technique to provide meaningful estimates of the NEF and EF. All in all, the present study findings were at least comparable, or of higher accuracy, to those reported in other similar verification studies of the “triangle” using both high resolution (airborne) and low resolution (satellite) data. To our knowledge, this study represents the first comprehensive evaluation of the performance of this particular methodological implementation at a European setting combining the SimSphere SVAT model and ASTER EO datasets.

A general approach for predicting indicator and pathogen decay in surface waters was developed using Bayesian hierarchical modeling, a persistence database, and a two-parameter model form. The resulting hierarchical regression describes general persistence behaviors with target-level intercepts and population-level coefficients. Uncertainty factors calculated with the approach suggest fecal indicator bacteria (FIB) and pathogenic bacteria persist similarly in surface waters, but median virus and protozoa persistence metrics may be 2–3 times greater than FIB in similar conditions. The two-parameter model underpinning the approach was used to identify drivers of these differences. Virus decay rates were shown to taper off more quickly than FIB, whereas protozoa were associated with longer initial periods of minimal decay. Despite the low accuracy of the hierarchical model compared to models fit to individual datasets, this approach addresses a critical gap for water management decision-making as site-specific and pathogen-specific persistence data are uncommon in water monitoring practices.

Underwater fish segmentation technology serves as a crucial foundation for extracting aquatic biological information. However, due to intricate and fluctuating underwater environments, existing segmentation models fail to precisely focus on key image regions. Based on this, the paper developed an underwater fish segmentation model, Receptive Field Expansion Model(RFEM), by enhancing soft attention performance (More attention is directed to fish regions when processing fish pixels). This paper tests ten different attention mechanisms and selects the attention mechanism with better performance indicators to improve it and form an RFEM model. This paper uses two underwater fish data sets to verify the proposed model. The experimental results show the segmentation mean intersection-over-union ratio (MIoU) of RFEM based on dilation convolution reached 88.37%, and the mCPA reached 93.83%, Accuracy reached 96.08%, and F1-score reached 93.74%. It can provide solid technical support for intelligent monitoring such as body length measurement, weight estimation of underwater fish.

Model results can have far-reaching societal implications, requiring fit-for-purpose models. However, model output is resulting from a particular path chosen with each modelling decision. We interviewed fourteen modellers in the Dutch water management sector in order to study how decision support hydrodynamic modellers make modelling decisions. An inductive-content analysis was performed. We identified eight motivation-categories. Individual and team considerations mostly motivate modelling decisions. We identified patterns between the motivation-categories and their occurrence across modelling steps. Modelling decisions during model implementation were found to be more in the modeller’s direct sphere of influence, while decisions concerning model structure and data selection more outside of it. So, even though modellers can leave their fingerprint, their sphere of influence and thus their fingerprint’s clarity is bound by institutionalised predefined decisions. Thus, models and their results are shaped within a broader sphere than the modeller’s alone, requiring a broader consideration of organisations and standards.

Water level monitoring in lakes and reservoirs is essential for effective water resource management, especially in remote areas where traditional ground sensors are costly and difficult to maintain. Remote sensing offers an alternative, but improving the quality, resolution, and accuracy of satellite data remains crucial. This paper introduces MoRLa (Measurement of Reservoir Level from Altimetry), a data filtering procedure designed to enhance satellite altimetry retrievals. MoRLa increases the acceptance of satellite observations and improves the quality of water level estimates by using physical characteristics of water bodies to exclude non-conforming measurements. Unlike previous studies with static masks, MoRLa employs a dynamic filter adaptable to actual water levels at specific times. Tested on reservoirs in the Korean Peninsula, including the Hwang-Gang dam, MoRLa shows significant improvements in water level measurements using Cryosat-2, ICESat-2, and Sentinel-3A and B satellites.

Recent research in evolutionary multi-objective optimization (EMO) highlights the concept of “Innovization”, which identifies essential patterns in high-quality, non-dominated solutions. This study introduces a novel method to pinpoint influential Best Management Practices (BMPs) in the Chesapeake Bay Watershed, optimizing the trade-off solution process. This approach, though innovative, demands considerable expertise and involves generating multiple solutions for expert analysis to detect commonly used BMPs. We devised three re-optimization strategies from these findings using an innovized BMP list, efficiently producing high-quality solutions. We also implemented transfer learning to adapt these strategies for new counties, demonstrating effectiveness in four West Virginia counties by reducing decision variables by 3% to 33% and achieving similar reductions in four additional counties. This showcases the potential of combining innovization with transfer learning to simplify complex optimization challenges, emphasizing its significant applicability in real-world settings.

This study presents a novel Coupled Human And Natural Systems (CHANS) modelling framework that integrates a hydrodynamic model with an agent-based model at the memory level within a multi-GPU computing environment. This two-way coupled model captures real-time interactions between human activities and flood dynamics, with a focus on the deployment of temporary flood defences during the 2015 Desmond flood in Carlisle, UK. The findings reveal that temporary defences can significantly reduce flood inundation by 30% with early warnings and 15% through real-time decision-making, leading to financial savings of £30 million and £15 million, respectively. The study further explores the decision-making process for effective emergency flood management, emphasising the importance of early warnings and resources optimisation. The new CHANS model provides a valuable tool for testing and optimising emergency flood management strategies, highlighting the necessity of directly incorporating human activities into flood risk management.

The evaluation of algal bloom forecasting models typically relies on error metrics that quantify the forecasting performance over the whole test set as a single number. Furthermore, the comparison with simple baseline methods is often omitted. To address this, we introduce a novel framework for Model performance Analysis and Visualization of time series forecasting (MAVts). MAVts incorporates novel algorithms for the automatic identification and visualization of time series periods of interest where the forecasting models are evaluated and compared with simple baseline methods. The application of MAVts on evaluating algal bloom forecasting models composed of sophisticated machine learning (ML) methods, reveals that in 85% of experiments a single error metric is not enough and only in 12.5% of experiments a ML model outperforms all baselines on all metrics and periods of interest. Thus, MAVts emerges as a valuable tool for analyzing and comparing ML models, advancing environmental management and protection.

Accurately forecasting ozone levels that exceed specific thresholds is pivotal for mitigating adverse effects on both the environment and public health. However, predicting such ozone exceedances remains challenging due to the infrequent occurrence of high-concentration ozone data. This research, leveraging data from 57 German monitoring stations from 1999 to 2018, introduces an Enhanced Extreme Instance Augmentation Random Forest (EEIA-RF) approach that significantly improves the prediction of days when the maximum daily 8-hour average ozone concentrations exceed $120\mu {\mathrm{g/m}}^3$. A pre-trained machine learning model is used to generate additional high-concentration data, which, combined with selectively reduced low-concentration data, forms a new dataset for training a refined model. This method achieved an improvement of at least 8% in the accuracy of predicting days with ozone exceedances across Germany. Our experiment underscores the approach’s value in enhancing atmospheric modeling and supporting public health advisories and environmental policy-making related to ozone pollution.