Figure 7

Using predictions from the functional concurrent model, these heat maps show how the non-linear effects of environmental variables on LRM-50 vary over time in mission (ranging from 0 to 1 representing the proportion of mission time elapsed). The same functional concurrent model, which was fit on the entire observed data, was used for each panel. For each environmental variable, predictions were made at a regular grid of time points between 0 and 1, and at all observed values of that environmental variable. All other variables were held at their average (continuous) or reference (categorical) value. The marginal distribution of the environmental variable observations is displayed as a rug plot (orange lines) above the x-axis. We find that better predicted performance (i.e., lower LRM-50, indicated by lighter yellow regions) is generally associated with lower radiation dose, moderate to higher temperatures, higher CO₂, and moderate and lower O₂.