An atmometer loses water from an artificial surface, not from the roots of the field
An atmometer’s water level falls during a warm week. The number can help describe evaporative demand, but it does not mean every nearby plant lost the same depth. The instrument has its own wetted surface, supply and exposure. Connecting that loss to a reference crop, then to actual crop use, requires two distinct explanations.
Here we focus on porous-surface agricultural atmometers: instruments that supply water to a ceramic evaporating surface, commonly beneath a specified cover. You will trace the water account, interpret a calibrated scale, test reference-surface compatibility and follow an original crop water balance without treating one gauge as a complete irrigation decision.
The examples use fictional data and supplied conversion relationships. They are not a recommendation for a particular product, field or watering schedule. Begin with the measured change; keep every later model step attached to its definition and independent checks.
Hydrology subject guide · Learning that transfers into life · How X Works hub
Choose your reading route
Route 1: Understand the artificial surface · Chapters 1–3
Route 2: Read exposure and calibration · Chapters 4–5
Route 3: Connect the reference to crops · Chapters 6–8
Route 4: Check performance and uncertainty · Chapters 9–10
Route 5: Transfer and report · Chapters 11–12
Full chapter index · Independent worked case · Questions and practical checks
Chapter index
Understand the artificial surface · Chapters 1–3
Identify the kind of atmometer being discussed
Trace the water supply and its limits
Read the scale according to its calibration
Read exposure and calibration · Chapters 4–5
Keep exposure and surface condition in the record
Calibrate the reference with matched observations
Connect the reference to crops · Chapters 6–8
Distinguish grass and alfalfa reference surfaces
Connect a reference estimate to actual crop water use
Place the estimate inside a root-zone balance
Check performance and uncertainty · Chapters 9–10
Validate bias, agreement and timing separately
Allocate uncertainty to the weak link
Transfer and report · Chapters 11–12
Test a new cover, a new climate and a new time step
Report observation, reference and decision as different stages
CHAPTER 1 OF 12
1. Identify the kind of atmometer being discussed
Atmometer is a broad instrument term, not the name of one universal measurement geometry. The American Meteorological Society includes several evaporimeter forms within it. This guide narrows the subject to water-fed porous-surface devices used as agricultural reference-evapotranspiration aids.
That scope matters. A paper-wick instrument, an open pan and a covered ceramic device need not respond identically. Their values should not be merged merely because all are described as evaporation gauges. State the instrument type, surface and intended reporting basis before comparing records.
A common agricultural arrangement has a reservoir feeding a wet porous ceramic element through a tube. A cover and, in some designs, a rain-excluding membrane influence the surface exposed to the atmosphere. Colorado State and Montana State describe this general arrangement. The instrument’s documentation determines its exact construction and calibration.
Draw the chain as stored water, delivery path, wetted surface and atmospheric loss. Then add a separate arrow from instrument loss to the reference estimate. The first chain describes physical water movement; the second describes an empirical or designed equivalence. Separating them prevents the reference label from hiding how the observation was produced.
CHAPTER 2 OF 12
2. Trace the water supply and its limits
For evaporation to continue from the intended wet surface, water must be supplied through the instrument. A dry reservoir or interrupted delivery changes what the surface can evaporate. A low reading could then indicate a supply problem rather than low atmospheric demand.
Use a fictional ideal instrument with adequate water supply. If its surface loses 0.30 litre over a period, the reservoir must supply that water unless the surface’s stored water decreases. With constant internal storage and no other loss, reservoir depletion is also 0.30 litre. This equality follows from the specified balance, not from assuming every real instrument has zero internal storage change.
Now suppose the delivery path becomes restricted and only 0.15 litre reaches the surface. The instrument may become supply-limited. Calling its smaller loss evidence of half the reference demand would confuse the apparatus condition with the atmosphere. An independent reference or condition check is needed to distinguish the explanations.
The practical lesson is to check the observation process before applying a new multiplier. A fitted coefficient might improve agreement during one restricted period while becoming wrong after the supply is restored. Repairing or identifying the measurement fault is a different task from recalibrating a normally operating instrument.
CHAPTER 3 OF 12
3. Read the scale according to its calibration
A physical reservoir level change can be converted to volume if the geometry is known. A product’s graduated scale may instead be calibrated directly in reference-equivalent depth. Those are not necessarily the same conversion, and a reader should not replace a documented calibration with a guessed area.
In a deliberately simple raw-volume example, a constant reservoir area is 0.003 m² and its water level falls 0.08 m. The volume decrease is 0.00024 m³, or 0.24 litre. That is an instrument water quantity. Without a specified reporting area or calibration, it is not yet 80 mm of reference crop evapotranspiration.
Suppose an exercise explicitly supplies a calibration of 20 mm reference-equivalent depth per litre of reservoir depletion. The 0.24 litre corresponds to 4.8 mm on that basis. The coefficient belongs to this fictional instrument and calibration; it is not a recommendation to divide every atmometer’s volume by the ceramic area.
Before computing rates, record whether scale numbers increase with accumulated loss or decrease with remaining water. A sight tube can be read in either convention. Write ending accumulated loss minus starting accumulated loss, or starting remaining water minus ending remaining water, as appropriate. Correct signs begin with the scale definition.
CHAPTER 4 OF 12
4. Keep exposure and surface condition in the record
An artificial surface exchanges radiation, heat and vapour with its surroundings. Shade, obstacles and local airflow can make an instrument’s exposure different from the field it is intended to represent. A convenient mounting location is not automatically a representative one.
Montana State’s explanation emphasises representative placement away from obstructions. That is a principle to apply through the instrument’s specific guidance, not a universal installation dimension supplied here. The appropriate height and surroundings depend on the design and reference task.
Imagine two fictional identical devices: one in open exposure and one shaded for half the day beside a wall. If their records differ, the comparison contains a deliberate exposure difference. Averaging the values does not produce a better open-field estimate simply because there are two instruments.
Record changes in the cover, wetting condition and surroundings. Dust, damage or replacement can change the physical measurement chain and may need investigation. Do not automatically diagnose a particular fault from a lower reading; use those observations to define competing explanations and check them against independent evidence.
CHAPTER 5 OF 12
5. Calibrate the reference with matched observations
A calibration states how an instrument reading estimates a defined reference quantity over supported conditions. Reference evapotranspiration may be computed from suitable meteorological observations using an established reference model. Such a comparison still depends on weather data quality and surface definitions; it is not an error-free revelation of every actual crop’s use.
For an original simple model, write Eref = a + bX, where X is the atmometer’s supplied interval reading. Units must match the coefficients: if both quantities are millimetres per day, a is millimetres per day and b is dimensionless. A model fitted to five-day totals has a differently interpreted intercept from a daily model.
Suppose fictional matched daily pairs are X = 2, 4 and 6 mm, with Eref = 2.5, 4.5 and 6.5 mm. They fit a = 0.5 and b = 1 exactly. That is a constructed teaching dataset. It demonstrates the model; it does not establish that a positive intercept makes physical sense at zero evaporation under all conditions.
Inspect residuals across the range and reserve independent periods. If high-wind or rainy intervals behave differently, do not claim validation based only on the remaining calm days. Colorado State notes condition-dependent limitations, and field studies investigate local calibration. The useful conclusion is bounded performance, not a single fitted number detached from its circumstances.
CHAPTER 6 OF 12
6. Distinguish grass and alfalfa reference surfaces
Reference evapotranspiration describes a specified well-watered reference surface. Grass-reference ETo and alfalfa-reference ETr are not labels to exchange casually. A gauge cover, reference model and crop coefficient must use compatible definitions.
Suppose a fictional data source reports 6 mm ETr for an interval. A crop coefficient of 0.80 explicitly defined against that alfalfa reference gives a modelled crop depth of 4.8 mm. Applying a grass-based coefficient to the same 6 mm without checking compatibility could introduce a systematic error even though the units look correct.
There is no universal conversion supplied here between ETo and ETr. Their relationship depends on conditions and reference formulation. If a product or dataset says only ET, recover the full definition before choosing a crop factor. Familiar abbreviations can conceal a more important difference than the last digit of the value.
Keep a small compatibility statement beside the calculation: instrument setting, reference definition, coefficient basis and output quantity. If one is unknown, identify that missing link. The disciplined response is not to assume whichever definition makes a preferred crop coefficient convenient.
CHAPTER 7 OF 12
7. Connect a reference estimate to actual crop water use
A well-watered reference is deliberately not a measurement of a particular field’s root uptake. Actual evapotranspiration also depends on crop development, soil water and management conditions. A reference gauge can continue losing water while a dry field becomes limited by available water.
For a simplified supplied model, ETcrop = Kc × Eref under the specified non-stressed conditions. With Eref = 5 mm/day and Kc = 0.60, the model gives 3 mm/day. The coefficient is an input for the fictional crop stage; the gauge has not independently observed 3 mm leaving the real soil.
If an exercise introduces a uniform stress factor Ks = 0.70 and explicitly uses ETactual = Ks × Kc × Eref, the answer becomes 2.1 mm/day. This simple structure is not every crop-water method. In methods separating soil evaporation and plant transpiration, stress need not multiply the entire combined term identically.
The distinction is practical. A fall in actual field ET can reflect stress while atmospheric demand remains high. A low actual-use estimate is not necessarily evidence that water supply is ample. Conversely, a high reference number alone does not tell us when a particular root zone reaches a chosen depletion limit.
CHAPTER 8 OF 12
8. Place the estimate inside a root-zone balance
To connect demand with stored water, define a root-zone boundary and track its inputs, outputs and storage. Reference ET is only one input to the model that estimates actual loss. Rain, irrigation, runoff, drainage and upward supply can matter separately.
Use a fictional simplified deficit account: Dnext = Dstart + ETcrop − effective rain − net irrigation. Here deficit means water needed to restore the specified reference storage condition. The exercise excludes upward groundwater supply and assumes that effective rain and net irrigation are already the amounts retained within the represented root zone.
Start with a 12 mm deficit, add 4 mm modelled crop ET and subtract 3 mm effective rain, with no irrigation. The next deficit is 13 mm. An atmometer reporting 5 mm reference ET was used with a supplied crop coefficient of 0.80 to produce the 4 mm loss term. The final 13 mm comes from several quantities, not the gauge alone.
If incoming water exceeds the available deficit, the simple equation needs a stated treatment of excess water and storage limits. A negative deficit is not automatically a physical store above any possible capacity. This guide demonstrates accounting and interpretation, not an operational watering threshold for an actual field.
CHAPTER 9 OF 12
9. Validate bias, agreement and timing separately
A high correlation can coexist with systematic error. Two records may rise and fall together while one remains larger. Calibration performance should therefore include differences, totals and relevant condition classes, not just a correlation coefficient.
Consider fictional reference values of 2, 4 and 6 mm and instrument estimates of 3, 6 and 9 mm. Their linear association is perfect, but each instrument estimate is 50% high and the total is 18 rather than 12 mm. Calling the instrument accurate because the lines move together would ignore the quantity the model needs.
Timing can fail even when totals agree. A delayed response of 0, 4 and 8 mm has the same 12 mm total as 2, 4 and 6 mm but distributes the loss differently. Depending on the decision interval, that distinction may matter. Match the assessment to daily tracking, multi-day totals or whichever task was defined.
Compare independent periods across the expected exposure and operating range. A greenhouse calibration is not automatically an open-field calibration; a dry-weather comparison does not settle rainy-weather performance. When evidence is limited, state the supported domain and retain a check for new conditions rather than announcing universal validity.
CHAPTER 10 OF 12
10. Allocate uncertainty to the weak link
An estimate can be uncertain because of reading precision, instrument condition, calibration, reference-weather observations or the later crop factor. These uncertainties enter at different steps. Collapsing them into one statement that ET is approximate does not help the reader decide what to improve.
Suppose fictional X = 5 mm has standard uncertainty 0.2 mm and a multiplier b = 1.10 has standard uncertainty 0.08. With zero intercept and independent small errors, Eref = 5.5 mm has approximate uncertainty √[(1.10 × 0.2)² + (5 × 0.08)²], about 0.46 mm. Calibration contributes more than reading uncertainty in this example.
Now use a supplied independent crop coefficient 0.70 with standard uncertainty 0.10. The modelled crop estimate is 3.85 mm. Its approximate uncertainty includes both 0.70 × 0.46 and 5.5 × 0.10, combining to about 0.64 mm. A precise instrument cannot remove uncertainty in the crop relationship.
These are conditional uncertainty calculations, not confidence intervals for a real farm. Dependencies and systematic errors require different treatment. If the device is supply-limited or the reference definition is unknown, a small numerical uncertainty from the equation would give false reassurance because the model itself is in question.
CHAPTER 11 OF 12
11. Test a new cover, a new climate and a new time step
Change one condition at a time. First replace the cover with a different specified surface while leaving the site unchanged. The water-balance principle survives, but the reference-equivalent calibration may change. Do not preserve the old factor merely because the reservoir and display look familiar.
Next take an unchanged device into a different climate or enclosure. Different radiation, airflow and wetting conditions can alter its relationship to the target. A successful local study supports its tested setting. Transfer needs an independent comparison rather than assuming similarity from one shared air-temperature value.
Then change a multi-day reading interval to hourly estimates. Internal water storage, supply response and measurement resolution can become more influential. Dividing a weekly total into equal hours creates an average; it does not reveal the original hourly pattern or validate the instrument at that resolution.
Finally change from reference-demand monitoring to estimating actual field loss. Add crop and soil evidence rather than relabelling the same number. Each transfer is legitimate when the added assumptions are visible and tested. The method becomes more useful, not less, when it says which parts of the original interpretation remain supported.
CHAPTER 12 OF 12
12. Report observation, reference and decision as different stages
A clear result identifies the device and surface, raw or calibrated reading basis, observation period, exposure, service history and reference definition. It then gives the conversion model and validation scope. Later crop and storage calculations should be listed separately so the reader can locate each assumption.
For example: the specified porous-surface device recorded 10 mm on its calibrated scale over the stated interval; a supplied local correction produced 11 mm alfalfa-reference ET; a compatible crop factor then gave 8.8 mm modelled crop use. This is not the same claim as directly measuring 8.8 mm field evapotranspiration.
When a reading is suspect, preserve its status rather than filling it silently with a confident prediction. A nearby reference calculation may support an estimate during an outage, but the provenance should remain clear. Observed and substituted records can coexist without pretending they have the same measurement chain.
For retrieval, explain why a dry field and a wet instrument can have different losses, why an 80 mm reservoir-level fall is not automatically 80 mm crop ET, and why perfect correlation can conceal a 50% bias. Those answers show that the instrument has become part of an understood system rather than a number to follow without context.
Independent worked case and evidence checks
Independent case: Ben receives three fictional two-day instrument readings on a documented reference-equivalent scale: 8, 10 and 6 mm. A supplied locally validated model for these two-day intervals is Eref = 1 + 0.90X, in millimetres per interval. The resulting estimates are 8.2, 10.0 and 6.4 mm, totalling 24.6 mm over six days.
Why not apply the model once to the combined 24 mm reading? Doing so gives 1 + 0.90 × 24 = 22.6 mm because it includes the interval intercept only once. For the fitted two-day model applied over three intervals, the total intercept contribution is 3 mm. The calibration’s time basis is part of its meaning.
This does not prove that a constant intercept should accumulate indefinitely or that the model is physically valid outside its training scope. It explains how the stated model operates. A new six-day calibration would be a separate model requiring evidence. Arithmetic consistency and model validity are different checks.
The exercise supplies a compatible crop coefficient of 0.80 for the first two intervals and 0.60 for the third. Modelled crop loss is 6.56 + 8.00 + 3.84 = 18.40 mm. Multiplying the reference total by the unweighted mean coefficient, about 0.733, gives about 18.04 mm instead. Reference-depth weighting preserves the interval account.
Start the simplified root zone with a 15 mm deficit. Over the six days, supplied effective rain totals 7 mm and supplied net irrigation is 5 mm. With no other terms and no capacity constraint reached, ending deficit is 15 + 18.4 − 7 − 5 = 21.4 mm. The calculation does not prescribe an irrigation action because no site-specific management threshold was supplied.
Now change the third interval: the reservoir ran dry before the reading ended. Its 6 mm value no longer has demonstrated equivalence to the normally supplied device. Mark the interval suspect, investigate duration and independent reference data, and do not retain the 6.4 mm model output as though the calibration had been validated for a dry instrument.
For a final evidence test, explain what each additional observation improves: a water-supply check addresses instrument operation; a weather reference addresses calibration; crop-stage information addresses the crop factor; soil water addresses the storage account. None converts one gauge into direct proof of every downstream term. A useful monitoring system combines evidence according to the question each observation can actually answer.
A second validation exercise separates wet-weather coverage from a favourable overall average. Suppose eight fictional dry intervals have a mean error of +0.2 mm, while two rain-affected intervals have a mean error of −2.0 mm. The overall mean error is (8 × 0.2 − 2 × 2.0)/10 = −0.24 mm. That modest combined bias conceals a much larger error in the rain-affected group. Whether it matters depends on the intended decision, but the condition-specific results should remain visible.
Do not infer that rain caused the discrepancy solely because the group labels coincide. Reference-weather errors, altered wetting or other correlated conditions could contribute. Review instrument and reference observations, then test a defined explanation in independent intervals. A calibration fitted separately after every disappointing comparison can appear excellent while never demonstrating transfer. Reserve a later dataset, preserve the original predictions and distinguish a repaired instrument from a recalibrated normally operating one.
Consider one more unit check. A fictional calibrated scale records 0.40 inch reference-equivalent loss. Multiplication by 25.4 mm per inch gives 10.16 mm. Reporting 0.40 mm would understate the depth by a factor of 25.4. If a chart later converts that depth to volume over one hectare under an explicitly justified uniform-depth assumption, the corresponding volume is 101.6 m³. The scale calibration and the field-area assumption remain separate even after the unit conversion is correct.
For independent study, write a short explanation without equations first: the atmosphere acts on the specified artificial surface; the supplied calibration estimates a particular reference; a compatible crop relationship estimates use; a root-zone balance connects use with stored water. Then attach the numbers. If the explanation jumps directly from a falling sight tube to an irrigation volume, locate the omitted reference, crop or storage step. Restoring that missing relationship is the useful correction, not merely repeating the final arithmetic more carefully.
Worked comparison
| Fictional two-day interval | Scale X | Reference estimate | Crop estimate |
|---|---|---|---|
| First | 8 mm | 8.2 mm | 6.56 mm at Kc 0.80 |
| Second | 10 mm | 10.0 mm | 8.00 mm at Kc 0.80 |
| Third | 6 mm | 6.4 mm | 3.84 mm at Kc 0.60 |
| Total | 24 mm | 24.6 mm | 18.40 mm |
Questions and practical checks
Does an atmometer measure actual crop evapotranspiration?
Its physical loss occurs from an artificial surface. A supported calibration can estimate a defined reference. Actual crop use requires compatible crop and water-condition evidence; it is not directly observed by the reservoir.
Can I divide reservoir depletion by the ceramic area?
Only if the instrument’s stated measurement and calibration method requires that conversion. A reference-equivalent scale may already incorporate a designed relationship. Inspect its documentation rather than guessing a reporting area.
Are ETo and ETr interchangeable?
No. They refer to different reference surfaces in the usage discussed here. Match the instrument setting, reference model and crop coefficient. No universal conversion between the two is supplied by this guide.
Does high correlation establish accuracy?
No. The fictional 3, 6, 9 mm estimates correlate perfectly with 2, 4, 6 mm references while being 50% high. Compare bias, totals, timing and performance under relevant conditions.
What if the reservoir runs dry?
The intended water-fed surface may no longer represent normal operation. Flag the affected interval and investigate. A normal-operation calibration should not be applied without evidence that its conditions still hold.
Continue with the connected subjects
Follow reference demand into water accounts
Evaporation depth, surface area and water volume
Drought indices and the meaning of evaporative demand
Compare other observations and their support
Eddy-covariance evapotranspiration and the changing flux footprint
Satellite soil moisture and the inference to root-zone storage
Water-budget closure and uncertainty in unmeasured components
Compare the observation with another measurement route
Choose the quantity first: these companion studies distinguish water lost from a maintained surface, precipitation accumulated in a vessel, and water condensed or intercepted on a collector. Their readings are connected by water and energy accounts, not interchangeable labels.
Evaporation pans: exposure, fetch and coefficient transfer
Precipitation storage gauges: accumulated catch and maintenance
Dew collectors: cooling, retained water and source attribution
Sources and scope
Colorado State University Extension: reference surfaces, atmometers and water-balance use
Montana State University: porous-surface instruments, covers and representative exposure
American Meteorological Society: scope of the atmometer instrument family
FAO Chapter 1: reference evapotranspiration and actual-crop distinctions
Chen and Robinson (2009): atmometer comparison and local calibration in a semiarid environment
The numerical cases in this guide are original fictional teaching examples, not measurements from a real site. This is an educational explanation, not a field safety instruction, engineering approval or an official certification. For a real project, use qualified practitioners, current product documentation and site-specific evidence. Sources were checked on 30 October 2026.