Sensitivity to the three configuration parameters
Three parameters were set by design judgement, not formal optimisation: the DL blend weight (α = 0.70), the GPD tail threshold (80th percentile), and the density saturation count (2). A Bali sweep varies each around its default while holding the others fixed. The result is both reassuring and a little humbling - the daily correlation barely moves.
The shaded band is the full sweep envelope; every curve stays inside it, and each red star marks the operating default. Hover any point for its full metric set. There is no sharp optimum to miss - the response surface is flat.
The full sweep
All fifteen settings, against CPC-UNI daily at the native window, per-pixel spatial median over the Bali subdomain then dekad-averaged. The operating default of each parameter is shaded. Pearson r holds in [
Why this is the expected result, not a lucky one
The parameters shape how the marginal correction is applied, but none of them add day-to-day timing information. Pearson r stays pinned close to the raw satellite value (the r ≈ 0.35 ceiling against CPC-UNI), so tuning α, the GPD threshold, or the saturation count moves skill only within sampling noise. Formal optimisation would not lift the ceiling - that needs the calendar-window fix or a different class of method.
The one place a parameter matters more than this chart suggests is spatial: the density saturation count barely moves the spatial-median r, but it decides where the CNN is allowed to act. Stability in the spatial-median number is not the same as stability in the map. The thesis flags all three as reasonable-but-unoptimised settings, and this sweep supports only a narrow claim: varying one parameter at a time, over the 80 land pixels of the Bali subdomain, daily r does not respond across the range tested. It is not a joint optimisation, and it says nothing about the parameters at full-domain scale.