Environmental Modelling & Software最新文献

Hydrological streamline delineation is critical for effective environmental management, influencing agriculture sustainability, river dynamics, watershed planning, and more. This study develops a novel approach to combining transfer learning with convolutional neural networks that capitalize on ImageNet pre-trained models to improve the accuracy and transferability of streamline delineation. We evaluate the performance of eleven ImageNet pre-trained models and a baseline model using datasets from Rowan County, NC, and Covington River, VA in the USA. Our results demonstrate that when models are adapted to a new area, the fine-tuned ImageNet pre-trained model exhibits superior predictive accuracy, markedly higher than the models trained from scratch or those only fine-tuned on the same area. Moreover, the ImageNet model achieves better smoothness and connectivity between classified streamline channels. These findings underline the effectiveness of transfer learning in enhancing the delineation of hydrological streamlines across varied geographies, offering a scalable solution for accurate and efficient environmental modelling.

Runoff generation in humid and semi-arid regions are usually dominated by saturation-excess mechanism and infiltration-excess mechanism, respectively. However, both mechanisms can co-exist in semi-humid regions. Therefore, we proposed a unified runoff generation scheme to represent the single and mixed runoff generation processes, making it applicable for different hydrometeorological conditions. The saturation-excess runoff generation scheme of the Xin'anjiang model was integrated with a modified Horton infiltration scheme at the grid cell scale. By substituting the original runoff generation scheme of the grid-Xin'anjiang (GXAJ) with the integrated scheme, we developed a new distributed hydrological model called grid-Xin'anjiang-infiltration-excess (GXAJ-IE) model. GXAJ-IE was tested in four watersheds and compared with two benchmark models with single runoff generation mechanism. The results indicate that GXAJ-IE model has higher flexibility and robustness in reproducing flood hydrographs under different rainfall conditions in semi-humid watersheds and has comparable performances with the benchmark models in humid and semi-arid regions.

Floods pose a significant threat to communities and infrastructure, necessitating timely predictions for effective disaster management. Conventional hydrodynamic models often encounter limitations in data requirements and computational efficiency. To overcome these constraints, we propose a novel multi-model ensemble framework integrating the flood extent and depth models for fluvial flood mapping. Various flood conditioning factors, such as terrain elevation and slope, flow direction, distance from the river, and latitude-longitude, were selected as model inputs, considering their relevance. The proposed framework was evaluated for predictive, extrapolative, and generalization capabilities. Results indicate that the proposed model successfully captures flood dynamics across a wide range of streamflow values, including unforeseen events, making it a valuable tool for predicting flood extent and depth. Overall, our approach offers a promising alternative to conventional hydrodynamic models, providing robustness, computational efficiency, scalability, automation, and integration with existing tools for flood inundation mapping tasks.

GeoGOBLIN, a novel environmental impact and land balance assessment tool, builds upon the GOBLIN biophysical land use emissions model for Ireland, offering enhanced resolution and system-level detail. It combines remotely sensed data and national agricultural census data to model climate change, air quality, and eutrophication emissions at the catchment level. Integration of the CBM-CFS3 forest carbon modelling framework (utilised in Ireland's National Inventory Report) increases alignment with national emissions reporting. GeoGOBLIN disaggregates emissions by life cycle assessment impact categories, making it a valuable tool for policymakers and researchers evaluating environmental and economic trade-offs associated with land-use scenarios. Illustrative scenarios demonstrate GeoGOBLIN's ability to assess the multifaceted impacts of alternative land uses, supporting informed decision-making for sustainable land use, food production, and a circular bioeconomy.

The approach of defining landscape patterns based on climate indices is applied in a case study area in North-Central Bulgaria. The results proved the strong interrelation between the climate indices and the elevation, enabling the implementation of a regression model. The results of the regression are used to define threshold values for delineation of all potential contours based on climate indices. The GIS modelling enables the integration of the results from different indices for the delineation of landscape contours of five potential landscape types. Four of them are validated with higher precision, proving the approach's applicability. One of the main capacities of the proposed approach is the opportunity for reconstructing climax vegetation and furthermore, the climax ecosystems that form the matrix of the potential landscapes. This can significantly contribute to assessing ecosystems and their services for restoration measures and implementing nature-based solutions.

Challenges in forest management are increasing due to climate change and its associated risks. Considering the needs and demands of various stakeholders leads to more complex decision-making. The increasing amount and quality of available geographic, forest and individual tree data, the combination of this data, and the use of forest growth simulators make it possible to support forest managers in this decision-making process. Our aim was to develop a strong visualization instrument that can be used in both forest planning and stakeholder communication. We present a solution based on a game engine, where data from multiple sources (terrain data, satellite imagery, tree data) is combined into a virtual environment. The user can move freely inside this virtual forest, look at the forest from arbitrary perspectives, and observe its development over the years under different management scenarios. We demonstrate the usefulness of this approach with a study region in Switzerland.

Process-based soil-crop models are widely used in agronomic research. They are major tools for evaluating climate change impact on crop production. Multi-model simulation studies show a wide diversity of results among models, implying that simulation results are very uncertain. A major path to improving simulation results is to propose improved calibration practices that are widely applicable. This study proposes an innovative generic calibration protocol. The two major innovations concern the treatment of multiple output variables and the choice of parameters to estimate, both of which are based on standard statistical procedure adapted to the particularities of soil-crop models. The protocol performed well in a challenging artificial-data test. The protocol is formulated so as to be applicable to a wide range of models and data sets. If widely adopted, it could substantially reduce model error and inter-model variability, and thus increase confidence in soil-crop model simulations.

This study presents a machine learning framework for multi-step-ahead prediction of groundwater levels in agricultural regions heavily reliant on groundwater irrigation. The framework utilizes a comprehensive set of predictive factors, including meteorological, hydrological, and human activity data. An optimal combination of input variables and their temporal delays was determined using a novel selection method. To address overfitting, a mathematical model for hyperparameter optimization was developed, leveraging sample subset cross-validation and an improved differential evolution algorithm. Numerical experiments on the YingGuo region in the Huaihe River Basin demonstrated that the hyperparameter optimization resulted in an 11.6%–38.5% increase in the Nash-Sutcliffe Efficiency (NSE) indicator. Additionally, fine-tuned temporal scales, from monthly to five-day resolution, significantly improved predictive performance, with NSE increasing from 0.629 to 0.952 (33.9% enhancement). However, longer forecasting horizons led to a 29.4% reduction in NSE. The study also implemented a multi-core parallel computing framework, which achieved a 15.35-fold improvement in computational efficiency while maintaining predictive precision. The integration of external factors enhanced the predictive performance across various observation wells. These findings contribute to a better understanding of groundwater dynamics and highlight the potential of machine learning models in improving groundwater depth predictions in agricultural regions with high reliance on groundwater irrigation.

In this paper, a time-explicit Finite-Volume method is adopted to solve the 2-D shallow water equations on an unstructured triangular mesh, using a two-stage Runge-Kutta integrator and a monotone MUSCL model to achieve second-order accuracy in time and space, respectively. A multi-GPU model is presented that uses the Message Passing Interface (MPI) with OpenACC and uses the METIS library to produce the domain decomposition. A CUDA-aware MPI library (GPUDirect) and overlapped MPI communication with computation are used to improve parallel performance. Two benchmark tests with wet and dry downstream beds are used to test the code's accuracy. Good results were achieved compared to the numerical simulations of published studies. Compared with the multi-CPU version of a 6-core CPU, maximum speedups of 56.18 and 331.51 were obtained using a single GPU and 8 GPUs, respectively. Higher mesh resolution enhances acceleration performance, and the model is applicable to other environmental modeling activities.

The accuracy of soil erosion models in agroecosystems with heterogeneous field conditions is challenging due to uncertainties from soil water fluxes and crop growth. In this study, we coupled two modeling methods (Freebairn and Rose) to represent soil erosion with a process-based crop and runoff models within the SIMPLACE framework. Their accuracy was compared to a statistical model developed using 16 erosion plots (each of 625 cm²) within the same field. Uncertainty analysis showed that runoff and slope angle were the most critical components for predicting sediment yield in both models, followed by soil erodibility in the Freebairn model and entrainment efficiency in the Rose model. However, due to plot size constraints, slope-length effects were not examined. The Freebairn model had a slightly higher accuracy (RMSE = 0.69 t ha⁻¹ d⁻¹) of sediment yield predictions than the Rose model (RMSE = 0.83 t ha⁻¹ d⁻¹). Both models are effective for predicting soil loss with appropriate parameter values.