MCP-Based Reconstruction Mechanism |
The MCP-based reconstruction mechanism reconstructs a single target data column using data columns of the same type from other datasets. It aims to generate synthetic data for the vacant time steps in which pattern-based reconstruction is impossible.
Windographer prefers pattern-based reconstruction, which uses reference data from within the target dataset, over MCP-based reconstruction, which uses reference data from other datasets, so it employs MCP-based reconstruction only for the vacant time steps.
Whether the MCP-based mechanism succeeds at reconstructing all vacant time steps depends on the availability of the source data, and how strongly the source data columns correlate with the target data column.
The procedure steps through the sensors from the other datasets in descending order of their strength of correlation with the target sensor (as measured by R2). For each of those source sensors, in each of the target dataset's vacant time steps in which the source sensor contains valid data, it synthesizes a target value using the appropriate MCP algorithm. For direction sensors, the appropriate MCP algorithm is the Constant Offset algorithm that Windographer always uses for MCP of direction data. For speed, speed SD, and temperature data, the appropriate MCP algorithm is the one the user has chosen in the Reconstruct Across Datasets window.
The algorithm stops when it has reconstructed all vacant time steps in the target data column, or when it reaches the end of the source sensors, or when the correlation with the next source sensor falls below the minimum R2 specified in the settings.
Whenever the MCP-based reconstruction mechanism reconstructs a speed value, it also reconstructs the associated speed SD value using the same source sensor, and assuming the turbulence intensity at the target sensor equals that at the source sensor in that time step.
See also
Pattern-based reconstruction mechanism
Reconstruct Across Datasets window