An Improved DINEOF Algorithm for Filling Missing Values in Spatio-Temporal Sea Surface Temperature Data

Area of Science:

• Data science
• Geospatial analysis
• Time series analysis

Background:

• Missing data in spatio-temporal datasets pose challenges for analysis.
• Existing methods like DINEOF and I-DINEOF have limitations in accuracy and efficiency.
• Accurate reconstruction of missing values is crucial for environmental and climate studies.

Purpose of the Study:

• To introduce an improved Data INterpolating Empirical Orthogonal Functions (DINEOF) algorithm, termed VE-DINEOF.
• To enhance the accuracy and efficiency of reconstructing missing values in spatio-temporal datasets.
• To compare the performance of VE-DINEOF against existing DINEOF and I-DINEOF algorithms.

Main Methods:

• Development of the Variable EOF-DINEOF (VE-DINEOF) algorithm, which utilizes variable optimal EOFs for reconstruction.
• Iterative reconstruction process is streamlined, with convergence checked only once.
• Validation using a sea surface temperature (SST) dataset, comparing VE-DINEOF with DINEOF and I-DINEOF.

Main Results:

• The VE-DINEOF algorithm demonstrated significantly enhanced reconstruction accuracy compared to DINEOF and I-DINEOF.
• VE-DINEOF achieved a reduction in computational time.
• Key performance metrics including Pearson correlation coefficient, signal-to-noise ratio, RMSE, and MAD confirmed VE-DINEOF's superiority.

Conclusions:

• The VE-DINEOF algorithm offers a more accurate and efficient solution for imputing missing values in spatio-temporal data.
• This advancement is particularly beneficial for environmental datasets like SST, improving climate modeling and analysis.
• VE-DINEOF represents a significant improvement over previous DINEOF-based methods for data interpolation.