Multi-resolution Filter for Spatio-temporal Satellite and Sensor Data

Marcin Jurek from The University of Texas at Austin presents a method for filtering massive, noisy, and incomplete spatio-temporal data, such as high-resolution environmental variables from satellites and aircraft. The proposed multi-resolution filter enables real-time inference and uncertainty quantification for linear Gaussian state-space models on large spatial grids. The work includes a simulation study, a real satellite-data application, and supplementary Python code for reproducing results.

Use Cases

• Real-time filtering of environmental sensor data based on the described linear Gaussian state-space models.
• Uncertainty quantification for incomplete spatial maps based on the probabilistic filtering method.
• Scalable inference on time-varying parameters using the described approximate Rao-Blackwellized particle filter.
• Performance benchmarking of filtering approaches for satellite data applications as demonstrated in the paper.

Strengths

• Includes a real satellite-data application demonstrating the method's performance.
• Supplementary materials provide Python code for reproducing the simulations.
• The method is described as highly scalable and fully probabilistic for filtering large spatial grids.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file size are unknown, which may limit suitability assessment.

Provenance

Source

The University of Texas at Austin

Collection Method

Likely involves simulated data and a real satellite-data application, though specific collection methods are not detailed.