environmental-modeling

Research project

🐢 Interface for Running NetLogo Simulations from R

Infrastructure and tooling for the NextGen Research DataStream (NRDS)

Numerical interpolation and approximation of water pollution data to construct continuous spatial pollution maps.

Heterogenous Graph Attention Transformer for high-resolution, distributed spatiotemporal flood prediction. Published at ACM SIGSPATIAL 2025.

A Python based kernel to perform spatial (environmental) impact assessment

Spatiotemporal Gaussian process modeling for environmental data: non-stationary PDE prior, deep kernels, multi-fidelity fusion, and A-optimal sampling.非稳态 PDE + 核深度学习 + 多保真 Co-Kriging + 主动采样的物理约束克里金方法，用于复杂时空环境建模与预测

Hydrological Model Continuum (HMC) – Open-source distributed hydrological model for simulation of water cycle processes, including runoff, soil moisture, snow, and routing.

Comparing LightGBM, a vanilla neural net, and a Physics-Informed NN for ETo estimation. The PINN embeds FAO-56 physics in its loss; reproducible synthetic data show physics helps most when measurements are scarce or noisy. Ready-to-run code.

Complementary repository with data and code for Wolf & Tollefsen, 2021.

Advanced framework combining 2D contaminant transport simulation (advection–diffusion) with machine learning to classify ecological risk (Low/Medium/High) and generate dashboards, GIFs, and reports.

Particle-based Python demo of synthetic plume transport through a variable wind field using advection, random-walk diffusion, eddy/shear flow, and density visualization.

Data-driven river water quality modeling using MATLAB and HSY algorithm with simulation and optimization techniques

Operational HAB forecasting for the San Jose Lagoon (in San Juan, PR) using weekly chlorophyll-a regression and bloom-risk classification.

Deterministic simulation framework for bio-stabilizing lunar regolith using spray dynamics, curing physics, and validated system models.

C++ systems development for environmental modeling, 3D visualization, and performance-critical applications. Combining computational efficiency with domain expertise in environmental science and data systems.

Reproducible hybrid mGFD-machine-learning defect correction for two-dimensional reactive contaminant transport, with trained PyTorch models, held-out validation, and an irregular-geometry audit.

Multivariate GLM/GAM pipeline modeling fish community structure responses to SST, wind, chlorophyll, and kelp biomass forcing in the California Current ecosystem

Code and processed data accompanying the MSc thesis Integrating Dynamical Systems and Machine Learning for Modeling and Predicting Decay Across Wood Treatments, conducted at DTU in collaboration with the Danish Technological Institute.

Bilingual web application for exploring validated one-dimensional reactive-transport simulations and comparing analytical, finite-volume, and neural solutions.