Sampling rate-corrected analysis of irregularly sampled time series.

Area of Science:

• Geosciences
• Data Science
• Complex Systems Analysis

Background:

• Irregularly sampled time series present analytical challenges due to variable sampling rates.
• Standard time series analysis methods can introduce biases when sampling resolution changes abruptly.
• Edit distance, used for comparing time series segments, is sensitive to local sampling rates.

Purpose of the Study:

• To address biases in time series analysis arising from irregular sampling.
• To improve the accuracy of recurrence quantification analysis (RQA) for nonlinear time series with varying sampling rates.
• To identify true regime shifts and environmental changes without artifacts from sampling variations.

Main Methods:

• Developed a constrained randomization approach using sampling-rate-constrained (SRC) surrogates.
• Generated time series and time axis surrogates that preserve local sampling rate characteristics.
• Applied the correction scheme to both synthetic data and a real-world speleothem proxy record.

Main Results:

• Demonstrated that transformation costs in edit distance are non-trivially related to sampling rates.
• Showed that abrupt sampling rate shifts can create spurious transitions in RQA.
• Identified a spurious transition in speleothem data solely due to sampling rate changes.
• Uncovered periods of reduced rainfall predictability linked to El Niño-Southern Oscillation and tropical cyclones.

Conclusions:

• The proposed SRC surrogate method effectively corrects biases in RQA caused by irregular sampling.
• Accurate analysis of irregularly sampled paleoclimate data is crucial for understanding past environmental dynamics.
• The findings highlight the importance of accounting for sampling variations in time series analysis, particularly in climate research.