Simple precip downscaling

[joelfiddes]

Just been thinking that there must be a way of using moisture information on pressure levels to downscale surface precip (TP). Particularly to identify elevation of maximum precipitation. I think the important principles are that

• Mid-elevation maxima occur due to orographic lifting and condensation:
• Air is forced upslope → condensation begins at cloud base, often above valley floor
• Condensation rate peaks somewhere in the mid-mountain as rising air cools and relative humidity reaches 100%
• At higher elevations, air is drier and/or already depleted → precipitation declines
• This is fundamentally dynamics-driven, not just moisture availability.

the plan:

1. Compute condensable moisture per layer: Multiply layer PW by RH fraction: ΔPW_i × (RH_i / 100).
   PW (Precipitable Water) = total water vapor in a vertical column above the layer, converted to equivalent depth of liquid water:
2. Compute layer fractions: Normalize condensable moisture across all layers to get each layer’s relative contribution.
3. Distribute precipitation: Multiply total ERA5 precipitation by layer fractions to get vertical profile; interpolate to any target elevation — naturally produces mid-elevation maxima, moisture depletion aloft, and regional/seasonal differences.

By using layer-wise moisture (ΔPW_i) weighted by RH, we capture where condensation is most likely, which is why mid-elevation maxima emerge.

I did some examples in various regimes in HMA:

Monsoon dominated (Summer precip dominates)

Westerly dominated (winter precip dominates)

[joelfiddes]

Elevation (m)
4000 | ░
3500 | ░
3000 | ░░ <- peak of layer PW → mid-elevation maximum
2500 | ░░░
2000 | ░░░░
1500 | ░░░ <- layer below has less moisture than peak
1000 | ░░
500 | ░
0 | ░ <- valley has low layer contribution; cumulative PW above would put max here

[joelfiddes]

[joelfiddes]

@ArcticSnow just tagging you in case you have any thoughts on this...i cant find much in literature but seems somehow obvious. Or am i getting my science wrong? Theres got to be something in here we can use?