Crop Water Stress Index: Canopy Temperature and Baselines

A thermal image of a field may show one patch of crop canopy several degrees warmer than another. That contrast can be an important clue about plant water stress, but a warmer pixel is not automatically a measure of soil moisture, transpiration or irrigation demand. To interpret the image, we need to know how the thermal sensor was calibrated, whether it saw leaves or exposed soil, what the air and sky were doing, and which reference temperature counts as well watered.

This guide explains canopy temperature measurements and the crop water stress index, CWSI, from first principles. We distinguish thermal radiance from physical leaf temperature, derive the baseline-based index, carry through checked numerical examples and discuss vapour-pressure deficit, canopy cover, instrument drift, weather, uncertainty and irrigation decision thresholds. The objective is to help a reader translate a coloured heat map into a claim that has the correct units, denominator, assumptions and limitations.

If one field has CWSI of 0.6 and another has 0.3, the first has a higher position between its specified lower and upper thermal baselines. That does not by itself prove twice the water deficit, twice the expected yield loss or a fixed number of millimetres of irrigation required. The practical skill is to preserve what the index was designed to describe and then decide what independent measurements would make a management conclusion defensible.

Why the canopy warms or cools

Leaves exchange energy with the environment through absorbed sunlight, emitted longwave radiation, sensible heat and latent heat associated with transpiration. When water evaporates from leaf surfaces and stomata, it consumes energy and can cool the canopy relative to the surrounding air. When transpiration is restricted, less latent heat removal may leave the foliage warmer, assuming other energy terms are comparable. This is the physiological intuition behind many thermal water-stress methods.

The mechanism is conditional. A leaf in shade absorbs less solar energy and may be cool even if its stomata are relatively closed. Wind changes the rate of sensible heat exchange. Humidity alters evaporative demand. A crop with sparse foliage may expose hot soil to the camera. A warm canopy can be caused by water limitation, but the same temperature can arise through different combinations of sunlight, air temperature, wind, leaf structure and measurement geometry.

A useful interpretation begins with energy balance rather than a colour legend. Net radiation is partitioned among sensible heat, latent heat and storage or other smaller terms. Thermal temperature is one observable in that coupled system. The thermal camera does not directly measure the individual energy fluxes. Its reading becomes biologically meaningful only when measurement conditions and comparison references constrain alternative explanations.

What a thermal sensor measures

An infrared radiometer detects thermal radiation arriving from the scene across a particular spectral response. It converts radiance into a brightness temperature using calibration and assumptions. A target’s actual surface temperature also depends on emissivity, reflected surrounding radiation and the sensor’s spectral band. A camera does not insert a thermometer into each leaf. It estimates temperature from detected photons, often applying an emissivity setting chosen by the operator or software.

Vegetation usually has high thermal infrared emissivity, but using a reasonable nominal value is not a substitute for checking the camera setup. Exposed wet soil, dry soil, sky reflections, plastic mulch and vegetation can differ. Errors are particularly noticeable when surface and surrounding radiative temperatures differ. Atmospheric transmission and water vapour matter more with longer paths or particular wavelengths. The correct measurement statement is therefore radiometrically estimated canopy-surface temperature under a specified method.

When translating a research plot, record whether temperature is a single-point infrared thermometer measurement, a handheld thermal image, a stationary sensor time series or an airborne mosaic. Each has different spatial and temporal support. A point sensor may miss field heterogeneity. A mosaic may contain residual calibration shifts across flight lines. The word temperature alone hides the method and what area was actually sampled.

Canopy temperature versus air temperature

A common diagnostic quantity is ΔT = T_canopy − T_air. Both temperatures may be reported in degrees Celsius; their difference is in Celsius degrees, numerically the same as a kelvin temperature difference. If canopy temperature is 31 °C and air temperature is 29 °C, ΔT is +2 °C. If canopy temperature is 26 °C while air is 29 °C, ΔT is −3 °C. The sign matters because a negative difference indicates a canopy cooler than the reference air.

A field instrument’s air temperature should be measured with adequate shielding, ventilation and placement. A thermometer warmed by direct sunshine can read too high and produce an artificially negative ΔT. Air temperature at a remote meteorological station may not represent the microclimate of a field at the same moment. The spatial and temporal alignment of canopy and air readings is therefore part of the data, not an afterthought.

ΔT is not yet a water stress index. The same +2 °C canopy-air difference could occur under strong radiation and dry air or under milder conditions with different wind and canopy architecture. The baseline-based CWSI attempts to put the observed difference into the context of the thermal behaviour expected for well-watered and minimally transpiring reference conditions.

Vapour-pressure deficit changes the reference

Vapour-pressure deficit, or VPD, is the difference between saturated water-vapour pressure at a specified temperature and the actual water-vapour pressure of the air. It is commonly expressed in kilopascals. A higher VPD generally means a stronger potential atmospheric demand for evaporation, though actual transpiration depends on plant and boundary-layer conductance and available energy. Temperature alone cannot specify VPD; humidity information is needed.

For a well-watered crop under suitable conditions, stronger evaporative demand can produce a more negative canopy-air temperature difference. Empirical non-water-stressed baselines often express the lower reference as ΔT_lower = a + b × VPD, with crop- and setting-specific coefficients. A negative b means the expected well-watered canopy becomes cooler relative to air as VPD rises. This linear description is empirical and cannot automatically be transferred to every plant species or climate.

When a paper displays VPD on one axis and canopy-air temperature on the other, first check whether the fitted baseline is for the same crop growth stage, radiation conditions and instrument configuration as the observed points. If the baseline was derived under one irrigation or canopy condition and applied elsewhere without validation, the numerical CWSI can be biased even when all temperatures have been measured correctly.

The two reference limits of CWSI

The lower baseline represents the expected temperature difference for a non-water-stressed, well-transpiring canopy under the stated conditions. The upper baseline represents a reference condition with severely restricted transpiration or an equivalent upper thermal limit, depending on the particular method. These are operational reference states, not universal fixed temperatures found in a handbook. Their construction is a central part of the measurement protocol.

A common empirical form is CWSI = (ΔT_observed − ΔT_lower)/(ΔT_upper − ΔT_lower). The numerator tells us how far the measured canopy-air temperature lies above the well-watered reference. The denominator measures the thermal interval between reference limits. Because both are temperature differences, their units cancel and CWSI is dimensionless. Under ideal reference conditions values near zero indicate the lower reference and values near one approach the upper.

Some publications use equivalent formulas written in terms of canopy temperature rather than canopy-air temperature. If the same air temperature is subtracted from observation and both references, the algebra is equivalent. However, different published CWSI methods do not always use identical upper and lower baselines. Before comparing results, verify the formula, reference generation and whether out-of-range values have been clipped.

Worked example: a value of one half

Consider a hypothetical fully developed crop whose lower baseline is ΔT_lower = 1.0 − 2.0 × VPD, with ΔT in degrees Celsius and VPD in kilopascals. The intercept has degrees Celsius as its unit, and the slope has degrees Celsius per kilopascal. On a day when VPD = 2.0 kPa, the lower baseline is 1.0 − 2.0 × 2.0 = −3.0 °C. Suppose the appropriate upper reference is +5.0 °C.

If measured canopy temperature is 30 °C and properly shielded air temperature is 29 °C, the observed difference is +1.0 °C. Substituting into the formula gives CWSI = [1 − (−3)]/[5 − (−3)] = 4/8 = 0.50. The arithmetic says this observation lies halfway between the two specified thermal reference limits. It does not say that soil has lost half its water-holding capacity.

Check the endpoints. If the observed ΔT equalled −3 °C, CWSI would be zero; if it equalled +5 °C, CWSI would be one. A measured difference of −4 °C would yield −1/8 = −0.125, and +6 °C would give 9/8 = 1.125. Values outside zero and one can arise from noise, mismatched baselines or conditions beyond the calibration domain. Automatically truncating them without reporting the rule can conceal diagnostic information.

What does one half mean biologically?

The 0.50 estimate is a dimensionless position between reference thermal states. It does not mean a leaf has precisely half of its stomata closed, half the gas exchange of an irrigated plant or half the available soil water. Stomatal response is nonlinear and depends on species, light, atmospheric demand and the plant’s hydraulic state. CWSI is a proxy constructed from temperature behaviour, not a direct volumetric moisture measurement.

Even when CWSI correlates with midday leaf water potential or yield, the relationship may be site- and crop-specific. It may differ by phenological stage or by the weather at observation. A robust study calibrates the index against appropriate independent measures, such as soil water status, plant water potential or transpiration, and tests its performance outside the calibration period. One fitted relationship is evidence, not a universal conversion table.

A cautious report would say the canopy was thermally intermediate between the locally defined well-watered and restricted-transpiration limits. If an irrigation threshold is proposed, it should come from crop performance and economic or agronomic validation, not from treating every CWSI value greater than a convenient round number as proof of emergency water deficiency.

How to build the non-water-stressed baseline

The lower baseline is commonly established from well-watered reference measurements covering a meaningful range of VPD under suitable radiation and canopy conditions. A plot of ΔT against VPD can be fitted to derive an intercept and slope. The goal is not merely to draw a visually attractive line. The reference observations should represent a physiologically appropriate, adequately supplied crop, and the regression should be checked for scatter, outliers and changes of behaviour.

A baseline obtained from a single clear afternoon has little ability to demonstrate how thermal response varies across atmospheric demand. If a study covers only VPD between 1.5 and 2.5 kPa, using its line at 5.0 kPa is extrapolation. The uncertainty may increase greatly outside the observed range. Crop varieties, plant height, canopy closure and field management can also shift the response even when VPD matches.

Before applying a published baseline, ask when and where it was measured, whether leaf temperatures came from comparable geometry, whether air readings were shielded, and whether reference plots remained well watered. A scientifically responsible comparison may need a new local baseline rather than borrowing coefficients from a different crop or climate.

How to define the upper baseline

The upper CWSI limit is often interpreted as the condition of little or no transpiration, but methods differ in how this is estimated. Some studies model the upper reference from surface energy balance, while others use wet and dry reference targets or empirical extremes. Each approach has assumptions about radiation, aerodynamic resistance and canopy emissivity. An upper limit from a dry artificial target may not reproduce the thermal behaviour of real non-transpiring foliage.

If an experimental protocol sets an upper limit of +5 °C for one condition, that limit should not be assumed fixed for all wind speeds and radiation levels. A hot canopy exchanges sensible heat with air; the resulting temperature depends on energy supply and air movement. It is possible for thermal contrasts to narrow as weather changes, making the same measured ΔT represent a different position between the relevant references.

A reader should inspect whether the upper baseline was measured, modelled or imposed as a convention, and whether uncertainty in that upper limit was included. If a paper reports precise CWSI values but gives no account of the upper reference, the apparent numerical precision is not supported by the method description.

A second worked example under different VPD

Retain the illustrative lower-baseline equation ΔT_lower = 1.0 − 2.0 × VPD. At VPD = 1.0 kPa the predicted lower reference is −1.0 °C. If the upper reference stays +5.0 °C for this deliberately simple exercise and measured ΔT is +2.0 °C, then CWSI = [2 − (−1)]/[5 − (−1)] = 3/6 = 0.50.

The observed ΔT of +2.0 °C in this second scenario differs from +1.0 °C in the first scenario, yet both index values equal 0.50 because the atmospheric reference changed. This illustrates why comparing raw canopy-air temperature differences without VPD can be misleading. It also illustrates why one must verify that a constant upper reference is appropriate before using the example to guide real irrigation.

Now suppose a reader mistakenly uses the first scenario’s lower baseline of −3.0 °C for the second scenario. The incorrect index becomes [2 − (−3)]/[5 − (−3)] = 5/8 = 0.625. The change from 0.50 to 0.625 is solely a reference mismatch. No leaf warmed, no soil dried and no irrigation occurred. A large part of translating CWSI is therefore protecting the denominator and its environmental context.

Thermal mosaics and the soil-mixing problem

A drone thermal pixel may contain green leaves, shadow, dry soil and wet soil. Its apparent brightness temperature emerges from a radiance mixture rather than from an arithmetic average of the components’ temperatures. Even if vegetation is accurately segmented in an accompanying visible image, imperfect alignment between thermal and visible cameras can leave soil contributions near crop edges. Mixed pixels are especially common before canopy closure.

Bare soil can be much hotter than transpiring leaves under sunshine and may raise a pixel’s apparent temperature. A gap between plants may then be falsely classified as severe crop stress. The correct repair is to examine spatial resolution, canopy fraction, vegetation masks, registration, emissivity assumptions and whether the pixel support is appropriate. Treating every hot pixel as a thirsty plant produces attractive maps but unreliable agronomic diagnoses.

For publication, specify whether the canopy statistic is a median of foliage-only pixels, a mean of an entire plot, a high percentile or a single sampled spot. The choice changes sensitivity to hot soil, cooler shade and outliers. A comparison across treatments should use a consistent statistic and explain what happens to plot pixels excluded by quality or vegetation masks.

Time of day and rapidly changing weather

Thermal stress contrasts often depend on solar radiation and evaporative demand. Near midday, the combination of high radiation and atmospheric demand may make differences in transpiration visible, but that is not a universal best hour. High clouds, wind gusts, irrigation events and transient shading can change canopy temperatures quickly. A flight that takes an hour to cover a field can include real temporal variation disguised as spatial variation.

If the first field block was imaged in strong sun and the last under cloud, differences in temperature may reflect illumination rather than irrigation treatment. One approach is to sample reference plots repeatedly, record simultaneous meteorology and incorporate flight time in the analysis. Another is to compare matched environmental conditions rather than assuming all pixels were captured in a single thermal instant.

When reading a time series, distinguish single-pass photographs from fixed-sensor records averaged over minutes. A 1-minute anomaly and a daily average answer different questions. A chart with no time-of-day information may show a biological effect, but it does not allow independent readers to assess whether weather and measurement scheduling provide alternative explanations.

The role of wind and boundary layers

A leaf loses sensible heat more readily when wind increases air exchange around it. Boundary-layer conductance therefore changes the relation between leaf temperature, transpiration and surrounding air. Two fields with equal stomatal closure might show different canopy-air differences under different wind regimes. Tall sparse crops, dense uniform crops and sheltered greenhouse plants have different aerodynamic environments.

Baselines derived under one wind distribution may not apply exactly to another. A local weather station that reports only air temperature and humidity leaves part of this aerodynamic uncertainty invisible. Some physically based energy-balance approaches incorporate aerodynamic resistance explicitly, but those estimates require their own inputs and assumptions. CWSI should not be described as independent of weather simply because it is dimensionless.

A field worker can reduce confusion by documenting wind speed, sensor angle, crop height and canopy exposure with every measurement campaign. An analysis should consider whether treatment plots were equally exposed to shelter or edge effects. Otherwise a warmer boundary row may indicate different air circulation rather than lower plant water availability.

Emissivity and reflected sky radiation

Thermal radiation is related to absolute temperature, not to Celsius temperature inserted into a physical radiation law. The Stefan–Boltzmann relation uses kelvin, and spectral infrared sensors use wavelength-dependent radiative models. For a surface at 30 °C, absolute temperature is 303.15 K. Applying a fourth-power radiation law to the number 30 instead of 303.15 would be physically invalid. Although a simple CWSI temperature difference can use Celsius, radiometric corrections cannot casually switch units.

A thermal camera sees emitted energy from the target and may also collect environmental radiation reflected by the surface. Emissivity is part of the conversion from radiance to temperature. When the sky is much colder in radiative terms than the canopy, an incorrect emissivity setting can bias estimated temperatures. Wet foliage, angled leaves and bright soil patches can complicate assumptions about what the detector sees.

Good practice includes instrument calibration, stable sensor warm-up, emissivity documentation, appropriate reference targets and checks on atmospheric or reflected-temperature settings. If a paper states a thermal accuracy figure, determine whether it describes instrument specifications under controlled conditions or actual accuracy of the entire aerial field workflow. Those are different claims.

Uncertainty in a normalised index

CWSI combines measured canopy temperature, air temperature and two reference baselines. Each has uncertainty, and the final index can be especially sensitive when upper and lower references are close together. A simple error analysis should not quote only the thermal camera’s nominal precision. The air sensor, VPD calculation, baseline fitting and spatial masking may contribute errors of comparable or greater size.

Take the illustrative baseline interval from −3 to +5 °C, so its width is 8 °C. If the observed ΔT increases by 0.4 °C while both baselines are fixed, CWSI increases by 0.4/8 = 0.05. But if the lower baseline is wrong by 1 °C, both the numerator and denominator change; its effect is not identical to a simple temperature offset. Correct propagation must reflect the actual functional dependence and correlations among inputs.

An uncertainty interval should also identify whether observations fall outside the baseline development range. A low reported standard error from many pixels does not remove systematic calibration error shared by the pixels. A scientifically useful summary states the dominant uncertainty sources and whether the treatment difference is large compared with them.

Example: propagating an uncertain temperature

Suppose a camera’s canopy temperature estimate is 30.0 ± 0.4 °C and the coincident air temperature is 29.0 ± 0.3 °C, where the uncertainties are independent standard uncertainties for this example. The observed difference is 1.0 °C. Its uncertainty is sqrt(0.4² + 0.3²) = 0.5 °C. The result is therefore ΔT = 1.0 ± 0.5 °C under these assumptions.

If the lower and upper references are treated as exact for teaching purposes, with an 8 °C interval, CWSI = 0.50 and the component of its uncertainty from ΔT is 0.5/8 = 0.0625. Reporting 0.50 ± 0.06 from this calculation would still be incomplete in a real study because the baselines are fitted or otherwise estimated quantities, not perfect constants.

If all canopy pixels share a calibration offset, averaging hundreds of them does not eliminate that offset. If the air sensor is poorly shielded, its error may correlate with intense sunlight. The assumptions behind the square-root-of-squares formula then need scrutiny. The point of the worked calculation is to make uncertainty visible, not to prescribe a universal ±0.06 accuracy for CWSI.

How to compare treatments fairly

Consider irrigated and water-limited plots measured on the same afternoon. First verify they have comparable canopy stage, leaf area and exposure. Compare plot-level estimates while treating plants or replicate plots—not millions of correlated pixels—as the independent experimental units when that matches the design. A thermal image with one million pixels is not a million biological replicates. Spatial autocorrelation and shared weather sharply limit independent information.

Suppose treatment A has mean CWSI 0.32 across four independent plots and treatment B has 0.53 across four plots. The difference is 0.21 index units. To know whether it is reliable, inspect plot-to-plot variability, uncertainty and the experimental design. If all plots were placed in separate soil zones rather than randomly assigned treatments, soil texture and irrigation system geometry may be confounded with treatment.

A robust result reports experimental replicates, measurement time, baseline derivation, canopy segmentation and uncertainty. It avoids claiming that a difference of 0.21 is directly a 21 percent increase in drought damage. CWSI is an index with particular anchors, and a treatment comparison is only as causal as the study design that produced it.

From index to irrigation decisions

An index may help identify where a crop’s thermal behaviour departs from a well-watered reference, but deciding whether to irrigate requires more information. Soil water availability, rooting depth, upcoming weather, irrigation efficiency, crop growth stage and yield sensitivity all matter. A hot canopy can indicate atmospheric demand even if soil water is not yet critically depleted. Conversely, deep soil water deficits may be developing before a thermal index crosses an arbitrary threshold.

Many irrigation decisions benefit from combining thermal observations with volumetric soil moisture, soil matric potential, crop coefficients and water-balance estimates. Existing eduKate guides on wetting-front detectors, root-zone soil water and evapotranspiration explain those other measurements. The useful role of CWSI is to contribute a canopy response signal, not to erase all the independent evidence needed for scheduling and amounts.

If a paper recommends irrigation at CWSI = 0.4, ask how that threshold was validated: crop, soil, phenological stage, climate, yield response, baseline and economic trade-offs. A threshold validated for one setting is an operational rule, not a universal plant physiological constant.

Related thermal indices are not identical

Degrees Above Non-Stressed, or DANS, compares the temperature of an observed plant or canopy with an adequately watered reference measured under comparable conditions. Degrees Above Critical Temperature, or DACT, compares temperature with a specified critical benchmark. Both can be simpler to compute than baseline-derived CWSI, but their meanings and requirements differ. In USDA research these simpler indices have been compared with CWSI for particular crops and conditions.

A DANS value of +3 °C is a temperature difference, while CWSI of 0.5 is dimensionless. The numbers should not be plotted on an axis with a shared unnamed unit. DACT is also sensitive to how the critical temperature was defined and whether it applies to the crop and environment. The convenience of an index does not excuse a weak reference or an untested assumption.

When comparing methods, ask what practical decision each was designed to improve. A simple index with directly measured controls may outperform a complicated model that uses poorly matched coefficients. Conversely, a robust physically informed baseline may generalise better when measured accurately. Good translation compares validation against the real task rather than ranking formulas by mathematical sophistication alone.

A practical acquisition and quality-control workflow

Before imaging, document the crop, growth stage, irrigation history, field geometry, expected canopy cover and intended inference. Check camera calibration and emissivity settings, air and humidity sensor placement, and time synchronisation across systems. Identify reference plots or conditions if they are part of the chosen CWSI approach. If using UAV imagery, determine flight altitude, overlap, field of view and the effective ground footprint of thermal pixels.

During acquisition, record radiation conditions, wind, clouds, air temperature, humidity, measurement times and any interruptions. Place reference targets where appropriate and capture repeated observations that can reveal drift. Inspect whether soil and foliage can be separated reliably and whether shade is systematic between plots. A consistent protocol matters more than collecting an impressive number of coloured frames with uncertain reference conditions.

After acquisition, validate registration and segmentation, calculate canopy temperature from accepted pixels, align weather observations, apply the declared baseline and inspect out-of-range CWSI values. Document excluded pixels and dates. Recompute sample calculations by hand. Before making a map, decide whether each pixel represents comparable vegetation, not merely whether software can assign it a number.

Independent problem one: calculate the index

A hypothetical crop has lower baseline ΔT_lower = 2 − 1.5 × VPD, where the slope is in °C/kPa. The upper baseline for the exercise is +6 °C. At VPD = 2.0 kPa, a radiometer reads canopy temperature 32 °C and the air sensor reads 30 °C. Calculate the observed difference, lower baseline and CWSI. State what extra evidence would be needed to convert the result into irrigation depth.

The observed difference is 32 − 30 = +2 °C. The lower baseline is 2 − 1.5 × 2 = −1 °C. Therefore CWSI = [2 − (−1)]/[6 − (−1)] = 3/7, approximately 0.429. A sensible rounded answer is 0.43 under the given references. Checking endpoints shows a numerator of 3 °C within a 7 °C thermal interval, so the dimensionless result is reasonable.

The value does not determine millimetres of irrigation. That requires root-zone water storage and depletion, soil hydraulic properties, expected crop water use, irrigation efficiency and management objectives. The thermal index can be one indicator of a water-limitation response, but using 0.43 as if it meant 43 millimetres would be a dimensional and conceptual error.

Independent problem two: misaligned observations

A drone captures a crop at 14:05 when canopy temperature is 33 °C. The analysis pairs it with air temperature 30 °C measured at 13:00 and VPD 1.2 kPa from a different weather station. A nearby field logger shows air temperature 32 °C and VPD 2.0 kPa at 14:05. Explain why the original comparison is unreliable before calculating an index.

With the stale temperature, ΔT would be 33 − 30 = +3 °C; with the coincident temperature it is 33 − 32 = +1 °C. The VPD used to choose a non-water-stressed baseline also changes. Both numerator and denominator can therefore be wrong in the original calculation. The correction is not a cosmetic timestamp edit. It can materially alter the interpretation of how warm the canopy was relative to its proper atmosphere.

A valid answer identifies both the air-reference error and the baseline error, then recommends time-synchronised, properly shielded local weather measurements. If representative local meteorology is unavailable, report the CWSI as uncertain or withhold it. Computing a precise index from mismatched inputs is not better than admitting the data are insufficient.

Independent problem three: a hot mixed pixel

A thermal mosaic shows 38 °C in a row gap, 31 °C in the adjoining dense leaf canopy and 30 °C air temperature. A vegetation mask leaves the gap in the analysed plot because the visible and thermal images were misregistered. If the lower and upper reference differences are −2 and +6 °C, calculate the index for the leaf-only observation and for the erroneously selected 38 °C pixel.

For the leaf-only canopy, ΔT = 31 − 30 = +1 °C and CWSI = [1 − (−2)]/[6 − (−2)] = 3/8 = 0.375. For the hot gap, ΔT = 38 − 30 = +8 °C and the same formula gives 10/8 = 1.25. The latter is above the expected range, a clue that the pixel or reference assumptions deserve investigation rather than automatic clipping to one.

The repair is to check registration, identify vegetation pixels and describe the selected canopy statistic. Applying a plant-water-stress formula to exposed soil temperature creates a category error. A precise software output cannot correct an incorrect target definition.

Independent problem four: compare a treatment

Two plots of the same crop are imaged with the same calibrated camera and meteorology. The mean foliage temperature is 29 °C for plot A and 31 °C for plot B; air temperature is 30 °C, the lower ΔT baseline is −3 °C and the upper is +5 °C. Calculate both index values and interpret the difference without changing the measured quantity.

Plot A has ΔT = −1 °C, so CWSI_A = [−1 − (−3)]/8 = 2/8 = 0.25. Plot B has ΔT = +1 °C, so CWSI_B = [1 − (−3)]/8 = 4/8 = 0.50. Plot B is 0.25 index units higher under the specified references. That is an absolute index difference, not a demonstrated 25 percent difference in soil water or crop production.

To argue that irrigation caused the difference, verify randomisation or matching, replicate plots, and similar exposure and canopy structure. To estimate water required, add independent soil and crop measurements. The interpretation is stronger when the researcher does not overstate what a thermal comparison proves.

FAQ about thermal crop stress maps

Can CWSI be negative? Yes, especially when observations are cooler than the fitted lower baseline or measurement noise and reference mismatch occur. Can CWSI exceed one? Yes, if observations lie above the upper reference or input conditions violate assumptions. Neither number should be silently forced into range without reporting the processing rule. Is one universally the point of permanent plant injury? No. It is a reference-index endpoint under a method, not a guaranteed biological mortality threshold.

Is a warmer canopy always under-watered? No. Radiation, wind, leaf area, disease, soil background and viewing geometry can all affect its apparent temperature. Does a single aerial image show plant water status throughout the root zone? No. It samples a radiometric surface state at one time and scale. Does a well-watered baseline stay valid forever? Not necessarily; crop growth, varieties, climate and measurement conditions can change the expected relationship.

Can the thermal index guide irrigation? Yes, when calibrated and combined with crop context and water-balance evidence. Is it interchangeable with soil moisture measured by TDR or cosmic-ray neutrons? No; those methods observe or infer different physical variables and volumes. Is it interchangeable with the stomatal conductance curve in another study? No; the processes are related, but an optical temperature proxy does not directly measure stomatal conductance.

Write the conclusion the measurements support

A defensible interpretation states that thermal canopy temperature was estimated for a defined foliage support, compared with simultaneous air temperature, adjusted against lower and upper reference differences and translated into a dimensionless CWSI. It gives a numerical value with its reference equation and relevant uncertainty. It does not call the result a direct measurement of irrigation depth, root water content or photosynthetic productivity.

If there are independent plant and soil measurements, explain how they support or constrain the inference. For example, elevated CWSI accompanied by declining predawn water potential and soil moisture strengthens the case for water limitation; a thermal anomaly coincident with a cloud transition weakens a simple causal reading. The point is not to avoid drawing conclusions, but to make their evidential basis visible.

A scientific heat map is useful when its colours remain attached to measurements, units, quality controls and tested reference conditions. The best translation describes what the sensor saw, how the index was computed and what question the result can answer. That method carries beyond crop stress to environmental imaging generally: first define the observable, then inspect the model, then state only the inference the evidence allows.