@Kai_zp The 1716 cases are for different geometries of an urban canyon, changing street width, building height and orientation, so the SVF is changing accordingly. My results are pretty much presented as yours; the average of all grid points (the number of which also depends on the canyon geometry), the average of the simulation period (15 hours of daytime). In total I had 1716 cases with 1716 SVF and UTCI average values. In my first comment I was trying to say I almost had the same situation like yours but ended up with a different correlation.
By Dragonfly, I mean Urban Weather Generator, which takes a rural weather file and, based on some properties (surface roughness, materials thermal properties, avg buildings heights, glazing ratio, etc) generates a weather file that represents the thermal conditions of an urban area. Intrinsically, it only morphes two variables in the weather file, which are the air temperature (Ta) and relative humidity (RH), so nothing with ENVI-met.