Experiment Data and Code -- Evaluating the Relationship between Endmember Variability and Temporal Unmixing Performance

[Paper title] Model-based Analysis of the Relationship between Endmember Variability and Temporal Unmixing PerformanceThis dataset includes high-resolution classification map (2m) and preprocessed Sentinel-2 reflectance data (20m) as intermediate data. And the final data directly for the experiment:1. abun_real.mat: includes the reference abundance of three classes;2. sentinel_image.mat: contains a cell array, each element is a Sentinel-2 three-dimensional array of a phase, arranged in date order;3. EMloc.mat: contains the endmember positions of three classes, the first column is the row coordinates, the second column is the column coordinates, and the row and column coordinates are generated based on the data size inside sentinel_image.mat. [MATLAB Demonstration Package] In addition to the experimental data, this repository provides a Demo Code Package to facilitate reproducibility and application of the proposed framework.Core Methodology: Provides the core functions for calculating the Endmember Variability (EMvar) measure.Representative Demo: Includes a complete workflow script using Mangrove as a case study under a fixed-dimensionality setting (10 features). This demo validates the linear relationship between EMvar and the SCM-based Confusion (Co) metric.Extensibility: By utilizing the provided scripts (sampling, unmixing, and modeling) and functions, users can easily extend the framework to variable-dimensional experiments or apply it to other datasets and scenarios. [Citation Note for Data and Code Usage] If you utilize the provided code or datasets in your research, please ensure to cite the following original foundational works: 1. For the Reference Classification Map and Data: Chen W, Shi C. Fine-scale mapping of Spartina alterniflora-invaded mangrove forests with multi-temporal WorldView-Sentinel-2 data fusion[J]. Remote Sensing of Environment, 2023, 295: 113690. 2. For the AAM Unmixing Algorithm: Heylen R, Zare A, Gader P, et al. Hyperspectral unmixing with endmember variability via alternating angle minimization[J]. IEEE Transactions on Geoscience and Remote Sensing, 2016, 54(8): 4983-4993. 3. For the SCM-based Confusion Metric: Silván-Cárdenas J L, Wang L. Sub-pixel confusion–uncertainty matrix for assessing soft classifications[J]. Remote Sensing of Environment, 2008, 112(3): 1081-1095.