NPA Surveillance and Monitoring
Our methodology for assessing Natural Protected Areas (NPAs) in Mexico is structured into three levels of analysis:
Level 1: Satellite Imagery Processing
At the first level, high-resolution satellite imagery from the Harmonized Sentinel-2 MSI: MultiSpectral Instrument, Level-2A is utilized to generate a collection of GeoTIFF images. These images are processed using various band combinations appropriate for true color visualization, vegetation analysis, urban area detection, and water presence identification. By selecting images with low cloud coverage and minimal null pixels, bi-weekly composite images are produced. This ensures high-quality visual data suitable for generating time-lapse sequences that effectively capture temporal changes within the NPAs. There is a high attrition rate as low quality images are not used.
Level 2: Quantitative Image Analysis
At the second level, the focus shifts to quantitative analysis of the timestamped images to assess pixel composition for key environmental indicators. Vegetation health is evaluated using the Normalized Difference Vegetation Index (NDVI), while water presence is quantified for each image to construct detailed time series datasets. To address data gaps caused by cloud interference or sensor anomalies, monthly controlled interpolation methods are applied, ensuring the continuity and reliability of the data series. Subsequent trend analysis of these time series enables the detection of significant changes in vegetation cover and hydrological features over time, providing insights into the ecological dynamics within the NPAs.
Level 3: Advanced Machine Learning Applications
The third level involves the application of advanced machine learning models to corroborate and enhance the findings from the initial analyses. Models such as Dynamic World and various land use classification algorithms are implemented to generate detailed land cover maps and to detect complex patterns and anomalies in the environmental data. These models leverage the spectral and temporal richness of the satellite imagery to classify land usage with high accuracy and to monitor changes with greater sensitivity. By integrating machine learning approaches, the methodology provides a robust framework for comprehensive environmental assessment, facilitating informed decision-making for conservation and management efforts within Mexico's NPAs.
