Evaporation demand and the canopy
Two thirds of the Leaf River's rain goes back to the air. This session follows the first place where that happens in dCREST: a small store of water held on leaves and branches. It explains how the day's evaporative demand is set, why the canopy is served first, and why calibration gave the Leaf River a canopy far larger than real leaves could hold.
You will be able to
- compute the day's evaporative demand from PET and ke;
- apply the canopy's two rules, fill first and evaporate first, by hand;
- say what demand is passed on to the soil;
- explain why a calibrated canopy can be much larger than a physical one.
The day's evaporative demand
Session 1 introduced potential evapotranspiration, PET: how much water the atmosphere could take up in a day from a surface that never dries out. The Leaf River's PET comes from a formula that uses only temperature and the sunlight reaching the top of the atmosphere (set by latitude and date), so it is an approximation. dCREST therefore scales it by a calibrated multiplier ke:
D is the day's evaporative demand in mm/day. For the Leaf River ke = 0.955, so the model uses almost exactly the formula's PET. This demand is then met in turn: from the canopy, then from the water reaching the ground that day, then from the soil, and finally from the aquifer (Session 6). Whatever an earlier source supplies is no longer demanded from the later ones.
The canopy store
When rain falls on a forest, the first few millimetres wet leaves, needles and bark instead of reaching the ground. This interception evaporates again quickly, because wet leaves are exposed to sun and wind. dCREST represents it with a small store, the canopy, holding W mm of water. Its capacity is a fraction ft of the soil capacity wm:
Tying the canopy to the soil capacity makes ft a pure number between 0 and 0.95, the same kind of quantity in every catchment. Each day the canopy follows two rules, in this order.
1. Fill first. Water arriving from the snow step, \(L\) (rain plus melt), tops up the canopy before anything reaches the soil:
2. Evaporate first. The day's demand is taken from the canopy before anywhere else:
The water \(L\) that remains after filling moves on to the soil (Session 5), together with the unmet demand \(D - E_c\).
fill = torch.minimum(liquid_mm, torch.clamp(capacity_mm - wt, min=0.0))
wt = wt + fill
liquid = liquid_mm - fill
demand = ke * pet_mm
et_canopy = torch.minimum(demand, wt)
wt = torch.minimum(torch.clamp(wt - et_canopy, min=0.0), capacity_mm)
pet_soil = torch.clamp(demand - et_canopy, min=0.0) / torch.clamp(ke, min=1e-3)
return liquid, pet_soil, wt, et_canopy
One detail looks odd: the leftover demand is divided by ke. The soil step multiplies by ke again, so the soil sees exactly \(D - E_c\). The division keeps the two kernels' interfaces in the same units as PET. The clamps at zero and the cap at the capacity can never actually bind; they are guards.
Two real days
With the Leaf River's parameters the canopy capacity is \(C = 0.1349 \times 235.2 = 31.7\) mm.
| Quantity | 27 Aug 1992, tropical downpour | 15 Aug 1993, hot and dry |
|---|---|---|
| water arriving \(L\) (rain + melt) | 127.55 | 0.00 |
| canopy before, W | 27.35 | 0.00 |
| fill \(F = \min(L, C - W)\) | 4.37 | 0.00 |
| PET → demand \(D = \mathit{ke}\cdot\mathit{PET}\) | 4.46 → 4.26 | 5.05 → 4.82 |
| canopy evaporation \(E_c = \min(D, W)\) | 4.26 | 0.00 |
| canopy after | 27.46 | 0.00 |
| water passed to the soil | 123.18 | 0.00 |
| demand passed to the soil \(D - E_c\) | 0.00 | 4.82 |
On the storm day the canopy was already nearly full from earlier rain, took only the last 4.4 mm, and then met the whole demand itself, so the soil had nothing to evaporate. On the dry day the canopy had emptied in the week since the last rain (8 August), and the full demand of 4.82 mm passed to the soil.
Why is the Leaf River's canopy so big?
Real forest canopies hold about 1 to 3 mm of water. The calibrated Leaf River canopy holds 32 mm, and over 1989–1999 it supplies 74% of all the model's evaporation. That is not a description of Mississippi pine forests. It is calibration using a store for a job it does well.
The canopy evaporates at the full demand until it is empty, while the soil, as Session 6 shows, evaporates more slowly as it dries. A large canopy therefore gives the model a way to return water to the air quickly after each rain, at the full rate, before it can run off. That is what calibration on the 1999–2008 decade chose for this warm, humid basin, where most rain is evaporated again within days or weeks. (It has a price: in the wetter test decade a smaller canopy would have matched the river's volume better, as the widget below shows.) In the model, the canopy is best read as "water that will evaporate soon", whether it really sits on leaves, in puddles, in leaf litter or in the topmost centimetres of soil.
This is the first example of a theme that runs through the course. A calibrated parameter is the value that makes the model's structure reproduce the data, which is not necessarily the value one would measure in the field. Session 13 discusses which dCREST parameters behave consistently enough to be compared across catchments.
Try different canopies
The widget reruns the whole model with your ft and ke, from 1 October 1986, and shows a window of 60 days. The scores at the top are for the whole test decade, 1989–1999.
In Python
Check your understanding
The canopy holds 10 mm of a 30 mm capacity. 4 mm of rain falls and PET is 3 mm with ke = 1. What evaporates from the canopy, and what reaches the soil?
Fill: min(4, 30 − 10) = 4 mm, so W = 14 mm and nothing reaches the soil. Demand 3 mm is met by the canopy: Ec = 3 mm, W = 11 mm. Water to soil 0 mm, demand to soil 0 mm.
What does the order "fill first, evaporate first" mean for a light shower on a hot day?
The shower is caught by the canopy and evaporated the same day; the soil sees neither the water nor, if the canopy holds enough, any demand.
Why might calibration make the canopy much larger than real interception storage?
Because the canopy evaporates at the full demand rate, a large canopy lets the model return rain to the air quickly, which fits a basin where most rain evaporates within days. The parameter is effective, not physical.
What to remember
- Demand is \(D = \mathit{ke}\cdot\mathit{PET}\); it is met first by the canopy, then by the water reaching the ground, then by the soil, then by the aquifer.
- The canopy (capacity \(C = \mathit{ft}\cdot\mathit{wm}\)) fills first and evaporates first.
- The soil receives the water left after filling, and the demand left after canopy evaporation.
- The Leaf River's calibrated canopy (32 mm) is an effective "evaporates soon" store.