Penman-Monteith Explained for Reservoir Managers

If you manage a reservoir, you have probably been handed a “Penman-Monteith” number and asked to trust it. This article explains what that equation actually computes, what it needs from you, and — importantly — why applying the standard crop version straight to open water will mislead you.

What the equation does

Penman-Monteith combines two physical drivers of evaporation into one expression: the energy available to vaporize water (mainly net radiation) and the atmosphere’s capacity to carry that vapor away (a function of the vapor-pressure deficit and wind). The FAO-56 reference form (Allen et al. 1998) is written as:

$ET_0 = \frac{0.408\,\Delta\,(R_n - G) + \gamma\,\frac{900}{T+273}\,u_2\,(e_s - e_a)}{\Delta + \gamma\,(1 + 0.34\,u_2)}$

The first term in the numerator is the energy contribution; the second is the aerodynamic (wind + deficit) contribution. That structure is why Penman-Monteith is robust across climates: when radiation dominates (calm, sunny lakes) the first term carries the estimate; when wind and dry air dominate (exposed industrial ponds) the second does. The underlying $(e_s - e_a)$ deficit is the same physics described in what is evaporation.

The inputs you need

To run it you need, at the relevant time step:

• Net radiation ($R_n$) and soil/water heat flux ($G$)
• Air temperature ($T$)
• Wind speed at 2 m ($u_2$)
• Saturation and actual vapor pressure ($e_s$, $e_a$) — derived from temperature and relative humidity
• The slope of the vapor-pressure curve ($\Delta$) and the psychrometric constant ($\gamma$), both computed from temperature and pressure

In practice these come from a weather station or a gridded weather dataset. Garbage in, garbage out applies sharply here: a poorly sited anemometer or a humidity sensor reading the wrong micro-climate will skew the answer more than the choice of equation.

Why crop-ET over-predicts open water

Here is the trap. The FAO-56 form computes reference evapotranspiration ($ET_0$) — the water use of a hypothetical, well-watered grass surface. It bakes in assumptions for a vegetated canopy: an albedo near 0.23 and a fixed canopy/surface resistance.

Open water is not grass. It has a much lower albedo (it absorbs more radiation), no stomatal resistance, and large thermal mass that stores heat and releases it later — including at night. Applied unmodified, reference ET will generally over-predict open-water loss, and it gets the timing wrong because it ignores the heat a deep body stores during the day and sheds after dark.

The fix is to use an open-water adaptation — the Penman open-water formulation with an open-water albedo and a removed canopy-resistance term — or to cross-check against a radiation-driven method like Priestley-Taylor (good for calm lakes) or an aerodynamic mass-transfer method (Harbeck 1962; robust for wind-exposed ponds). We walk through choosing among these in how to calculate evaporation.

How accurate is it, really?

For moderate conditions, the major estimation methods typically agree within about 10–20% of one another, and Penman-Monteith generally lands within roughly ±10–20% of observed open-water loss when properly adapted and fed good data. That is good enough for planning, budgeting, and comparing reduction methods — but it is not a precision instrument. Treat any single-method figure as an estimate with an error bar, not a measurement.

Two habits improve reliability:

1. Run more than one method and look at the spread. A wide spread flags a site where one driver (e.g. strong wind) is being mishandled.
2. Calibrate to local data where you have a pan record or a water-balance history.

Getting numbers without building a model

Implementing FAO-56 correctly — radiation estimation, pressure corrections, the open-water albedo swap — is fiddly, and reservoir managers rarely need to write the code themselves. For live, multi-method estimates that pull real weather, AWTT publishes a free evaporation calculator that runs Penman-Monteith alongside Priestley-Taylor, Hargreaves-Samani, and mass-transfer methods. It is a practical way to get a defensible number and to see how much your estimate moves between methods — which is itself the most honest indicator of confidence.