Quantifying unpredictability: A multiple-model approach based on satellite imagery data from Mediterranean ponds

Area of Science:

• Ecology
• Environmental Science
• Remote Sensing

Background:

• Environmental fluctuations are crucial for understanding organismal evolution.
• Quantifying environmental predictability is vital but challenging due to data limitations.
• Long-term remote sensing data offers a solution for ecological studies.

Purpose of the Study:

• To develop a method for accurately estimating water-surface area from satellite images, addressing challenges in saline environments.
• To extract and quantify predictable and unpredictable components of environmental variation.
• To compare different modeling approaches for assessing environmental predictability.

Main Methods:

• Utilized 27 years of Landsat TM/ETM+ satellite data for environmental monitoring.
• Developed a novel image processing pipeline combining band ratios and a near-infrared filter to accurately measure water-surface area in saline ponds.
• Applied two distinct analytical approaches: Colwell's predictability metrics and General Additive Models (GAMs) to quantify variation components.

Main Results:

• Successfully developed a robust method for estimating water-surface area from satellite imagery, even in challenging saline conditions.
• Quantified a wide range of environmental predictability across 20 Mediterranean saline ponds and lakes.
• Observed that different modeling assumptions can lead to divergent predictability estimations, suggesting context-dependent interpretations.

Conclusions:

• The developed fluctuation analysis is broadly applicable to various ecosystems, aiding in the characterization of environmental predictability.
• Divergent predictability estimations may reflect varying impacts on different organism types.
• Selecting appropriate predictability metrics based on organismal a priori information is recommended for ecological studies.