fieldClim: Choosing fieldClim Heat-Flux Methods by Measurement Design
A decision guide for one- and two-height station setups, canopy-layer context, target quantity, and defensible findings
Jörg Bendix, Chris Reudenbach
2026-06-01
Source:vignettes/fieldclim_m2m_en.Rmd
fieldclim_m2m_en.RmdPurpose
This page is a practical decision guide. Its starting point is not the question “Which heat-flux method is best?”, but the question “What measurement architecture do I have, which exchange context do I mean, and what kind of finding can I defend from this dataset?”
The same energy-balance equation can be used in very different ways. A one-height station with radiation and soil heat flux supports a different interpretation than a two-height profile station. A tower above, within, or below a canopy requires an explicit canopy-layer interpretation before any flux statement is meaningful.
The core notation used here is:
\[ A = Q^* - B \]
\[ Q^* - B = H + LE + R_E \]
\[ R_E = Q^* - B - H - LE \]
Here, Q* means net radiation, B means soil heat flux, A means available energy, H means sensible heat flux, LE means latent heat flux, and R_E means the residual or non-closed energy term. Positive B means soil heat flux into the soil as a sign convention. The real soil heat flux can change direction; this is represented by the sign of B.
Operational selection matrix
This table is the practical entry point. It tells you what to run and what kind of statement can be defended from the dataset.
| Measurement situation | First decision | Main method choice | Secondary comparison | Defensible finding | Avoid claiming |
|---|---|---|---|---|---|
| One height plus Q* and B over a defined surface | Treat as available-energy and LE-oriented setup | Penman or Priestley–Taylor | Available-energy diagnostics | LE-oriented estimate, empirical partition, or unresolved remainder | Direct gradient-based H |
| Two heights plus Q* and B over a defined surface | Treat as two-height gradient setup | Bulk–Residual for H and residual LE; Bowen if gradients are reliable | Penman/PT comparison; Monin/Profile diagnostic | Method-sensitive H/LE partition and closure behaviour | Bulk closure as validation |
| One height in canopy/tower context | Define the canopy-layer context of that level | Penman/PT only if Q* and B fit that layer | Available-energy diagnostics | Local ground, under-canopy, within-canopy, canopy-top, or above-canopy energy context | Generic canopy flux |
| Two heights in canopy/tower context | Define whether the target is within-canopy coupling, canopy-top exchange, or above-canopy exchange | Bulk–Residual, Bowen, or Monin/Profile only with explicit height and roughness logic | Penman/PT comparison if Q* and B are meaningful for the same layer | Exchange diagnostics within or above the canopy layer | Simple surface-layer flux without canopy-layer context |
| Missing Q* or B | Treat available energy as incomplete | Profile-only diagnostics if two heights exist | Penman only if required meteorology exists | Incomplete flux comparison | Energy-balance closure |
This table is not a validation hierarchy. It maps measurement design to defensible interpretation. A method is useful only if the dataset contains the information that the method actually uses.
Refinement scheme
After the operational table, the decision is refined in four steps: exchange context, measurement architecture, target finding, and method role. These steps prevent the common error of treating all methods as competing estimates of the same physical quantity.
Step 1: Define the exchange context
The exchange context is the physical layer to which the flux interpretation refers. Over short homogeneous surfaces such as meadow or bare soil, this is often straightforward. In canopy or tower setups it is not. A sensor near the forest floor, a sensor inside the canopy, a sensor near the canopy top, and a sensor above the canopy describe different exchange situations.
| Exchange context | What must be clarified | Consequence for method choice |
|---|---|---|
| Short homogeneous surface, such as meadow or bare soil | Sensor height above the surface, Q*, B, wind exposure | One-height or two-height method choice is usually straightforward |
| Forest floor or under-canopy station | Whether the target is local ground microclimate or exchange with the canopy air space | Available-energy and microclimate interpretation may be valid; canopy-top exchange is not automatic |
| Within-canopy layer | Whether the target is vertical coupling or decoupling inside the canopy | Treat primarily as canopy-layer exchange diagnosis, not as a simple surface-layer flux |
| Canopy-top transition | Whether the target is exchange across the canopy top | Methods require explicit height and roughness interpretation |
| Above-canopy layer | Whether the target is exchange above the canopy | Bulk and Monin/Profile become highly parameter- and reference-height dependent |
| Unknown canopy-layer context | Exchange layer and reference height are not defined | Do not report robust fluxes; report exploratory diagnostics only |
Step 2: Identify the measurement architecture
The available station architecture controls which methods can be used defensibly. One height supports LE-oriented or available-energy interpretations. Two heights support gradient-based estimates and partitions. In the current fieldClim method set, profile-based heat-flux methods are effectively one- or two-height workflows. If more than two measurement heights exist, they must be reduced explicitly to a selected height pair before applying the two-height methods.
| Measurement architecture | Data pattern | Structurally supported methods | Main finding type |
|---|---|---|---|
| One height with Q* and B | T, RH, u at one height; Q*; B | Priestley–Taylor, Penman, available-energy diagnostics | LE-oriented estimate, partition, or unresolved remainder |
| Two heights with Q* and B | T, RH, u at two heights; Q*; B | Bulk–Residual, Bowen, Monin/Profile diagnostic, Priestley–Taylor, Penman | H baseline, H/LE partition, method comparison, diagnostic residual |
| One height in canopy/tower context | One selected level within, below, near, or above canopy | Priestley–Taylor, Penman, available-energy diagnostics if Q* and B fit the same context | Local energy and microclimate interpretation |
| Two heights in canopy/tower context | Two selected levels assigned to a canopy-layer exchange context | Bulk–Residual, Bowen, Monin/Profile diagnostic only with explicit height and roughness logic | Within-canopy coupling, canopy-top exchange, or above-canopy exchange diagnostics |
| Missing Q* or B | Profile information may exist, but available energy is incomplete | Profile-only diagnostics or partial Penman-type comparison | Incomplete flux interpretation |
Step 3: Choose the target finding
The target finding determines the method role. The same dataset can support several methods, but each method answers a different question.
| Target finding | Best method role | Defensible interpretation | What not to claim |
|---|---|---|---|
| Direct sensible heat estimate | Bulk | H is estimated from a vertical temperature gradient and an exchange assumption | The residual closure validates H |
| Latent heat or evaporation estimate | Penman, Priestley–Taylor | LE is estimated from available energy, atmospheric demand, or empirical partitioning | The remaining energy is automatically H |
| H/LE partition of available energy | Priestley–Taylor, Bowen, Bulk–Residual | A is distributed according to a method assumption | Partition closure is independent validation |
| Profile and stability diagnosis | Monin/Profile diagnostic | H and LE are diagnostic profile outputs and R_E remains visible | This is automatically a complete MOST truth |
| Formal closure demonstration | Bulk–Residual, Priestley–Taylor, Bowen | H + LE equals A because of definition or partition assumption | Formal closure proves physical accuracy |
| Open residual diagnosis | Penman remainder, Monin/Profile residual | The unresolved part remains visible | The residual is automatically a known physical flux |
Step 4: Translate method output into a finding
The final step is to translate the selected method into a defensible statement. This is where most interpretation errors occur. The same numerical output can mean a direct estimate, an empirical partition, a residual definition, or a diagnostic remainder.
| Method | What the method determines | What the result means | What the result does not prove |
|---|---|---|---|
| Priestley–Taylor | LE_PT from available energy and empirical partitioning | LE-oriented partition; H_PT is the complement | Independent H or validation of alpha_PT |
| Bulk–Residual | H_bulk from a two-height temperature gradient and exchange assumption | Direct H estimate plus residual LE definition | Physical validation of H_bulk or independent LE |
| Bowen Ratio | H/LE partition from gradient ratio beta | Partition based on T/RH gradients | Robustness under weak or noisy gradients |
| Penman | LE_Penman | LE-oriented estimate and unresolved energy remainder | Paired H estimate |
| Monin/Profile diagnostic | H_MO and LE_MO from profile/stability logic | Diagnostic profile result with visible residual | Energy-balance closure or automatic MOST truth |
| Closure diagnostics | R_E, closure ratio, open remainders | Semantics of closure, residual, or open remainder | New flux estimate |
Evidence base and reference frame
This page separates implementation evidence from physical reference
evidence. Implementation evidence describes what the current
fieldClim code, tests, and generated documentation actually
do. The current implementation audit reports that the Bulk path
estimates sensible heat with a neutral bulk aerodynamic-resistance
structure and then defines latent heat as an algebraic residual in the
Bulk–Residual workflow. It also reports that the optional
ri_guard is a diagnostic filter rather than a full
stability correction, and that Bulk must not be described as MOST.
Physical reference evidence defines the method families: available-energy partitioning, Bowen-ratio energy balance, aerodynamic resistance or bulk transfer, Penman-type latent heat estimation, and Monin–Obukhov or profile-gradient diagnostics. These references provide the comparison frame, but they do not prove that every package implementation is a complete canonical implementation.
Priestley–Taylor is treated here as an available-energy partitioning approach for latent heat. In the original formulation, evaporation is parameterized from available energy and a coefficient rather than derived from a measured humidity-gradient flux (Priestley and Taylor 1972). In fieldClim, this role is mirrored by estimating LE_PT from available energy and a surface-dependent coefficient, while H_PT is the complement.
Bowen-ratio methods are treated as gradient-based energy partitioning methods. Their key assumption is that the ratio of sensible to latent heat can be inferred from temperature and humidity or vapour-pressure gradients under suitable measurement conditions (bowen1926?). Because this ratio becomes fragile when gradients are small or noisy, Bowen-ratio calculations require rejection or filtering logic in practical applications (ohmura1982?).
Penman-type methods are treated as latent-heat or evaporation-oriented methods. They combine available energy with an aerodynamic drying-power term and therefore produce an LE estimate rather than a complete paired H/LE solution (Penman 1948). This is why the Penman remainder is described here as an unresolved remainder, not as sensible heat.
Monin–Obukhov Similarity Theory is the canonical reference frame for surface-layer profile-gradient and stability-based flux interpretation (monin1954?). In this page, however, the package method is deliberately called Monin/Profile diagnostic. This avoids overstating the implementation as a fully validated canonical MOST solution. Foken’s review of MOST provides the theoretical reference frame, but it is not evidence that the package implementation satisfies all canonical requirements (Foken 2006).
Case A: One-height station with radiation and soil heat flux
A one-height station can define available energy but cannot estimate H from a vertical air-temperature gradient. It can support Priestley–Taylor, Penman, and basic energy-balance interpretation.
\[ A = Q^* - B \]
For Penman, the useful interpretation is latent-heat-oriented:
\[ LE_\mathrm{Penman} = f(A, VPD, r_a, r_s) \]
The remainder is:
\[ U_\mathrm{Penman} = A - LE_\mathrm{Penman} \]
U_Penman is an unresolved energy remainder. It can contain sensible heat, closure error, input error, and model mismatch. It should not be labelled H unless an explicit additional closure assumption is introduced.
| Dataset | Primary methods | Reportable finding | Avoid |
|---|---|---|---|
| One height: T, RH, u, Q*, B | Penman, Priestley–Taylor | Available energy and LE-oriented estimate | Direct gradient-based H |
| One height: T, RH, u, no reliable Q* or B | Limited Penman-type comparison if inputs exist | Atmospheric demand or incomplete LE comparison | Energy-balance closure |
| One height with soil data only | Energy-state description | Ground and radiation context | Turbulent flux partitioning |
A defensible wording is: “Available energy was quantified as A = Q* − B. Latent heat was estimated using Penman or Priestley–Taylor. A direct gradient-based sensible heat estimate was not possible from this dataset.”
Case B: Two-height station over a defined reference surface
A two-height station is the first configuration that supports gradient-based interpretation. If T, RH, and u are measured at two known heights, the station can support Bulk, Bowen, and Monin/Profile diagnostics, in addition to Priestley–Taylor and Penman.
Bulk–Residual estimates sensible heat from the two-height temperature difference and an exchange assumption:
\[ H_\mathrm{bulk} = \rho c_p \frac{t_1 - t_2}{r_a} \]
The Bulk–Residual latent heat is then defined as:
\[ LE_\mathrm{res} = Q^* - B - H_\mathrm{bulk} \]
The closure follows by definition:
\[ R_E = Q^* - B - H_\mathrm{bulk} - LE_\mathrm{res} = 0 \]
This does not validate H_bulk. It only states how LE_res was defined.
| Data available | Method role | Reportable finding | Diagnostic risk |
|---|---|---|---|
| T at two heights and wind | Bulk | Direct H baseline | Exchange assumption and stability |
| T and RH at two heights | Bowen | Gradient-based partition | Weak or noisy humidity gradients |
| T, RH, and wind at two heights | Monin/Profile diagnostic | Profile/stability-sensitive diagnostic outputs | Not force-closed; parameter-sensitive |
| Q* and B also reliable | Method comparison | How assumptions distribute A into H and LE | Closure is not validation |
A defensible wording is: “Bulk–Residual estimated sensible heat from the vertical temperature gradient and defined latent heat as the remaining available energy. The resulting closure is algebraic and follows by definition.”
Case C: Canopy-layer and tower setup
A canopy or tower setup is not just a question of sensor height above ground. The central question is which exchange layer is being interpreted: the under-canopy air space, the within-canopy layer, the canopy-top transition, or the above-canopy layer. A sensor pair inside the canopy, across the canopy top, or above the canopy represents different physical processes.
If one relevant height is used, the result mainly describes the local microclimate and available-energy state at that level. If two relevant heights are used, Bulk–Residual, Bowen, or Monin/Profile diagnostics can be used only after the height pair has been assigned to a canopy-layer interpretation. The result should then be reported as exchange or coupling diagnostics, not as a generic surface-layer flux.
| Canopy / tower situation | First decision | Method role | Reportable finding |
|---|---|---|---|
| One height near forest floor or below canopy | Define local ground or under-canopy context | Penman, Priestley–Taylor, available-energy diagnostics if Q* and B fit that context | Local under-canopy energy context |
| One height within canopy | Define local within-canopy context | Penman, Priestley–Taylor, available-energy diagnostics if Q* and B fit that context | Local within-canopy energy context |
| Two heights within canopy | Define whether the target is vertical coupling inside the canopy | Diagnostic only; Bulk/Bowen only with caution | Within-canopy coupling or decoupling signal |
| One height within canopy and one height above canopy | Define whether the target is canopy-top exchange | Bulk–Residual, Bowen, or Monin/Profile only with explicit height and roughness logic | Canopy-top exchange diagnostic |
| Two heights above canopy | Define the above-canopy reference layer | Bulk–Residual, Bowen, or Monin/Profile diagnostic | Above-canopy exchange comparison |
| Unknown canopy-layer context | Do not force flux interpretation | Exploratory diagnostics only | Flux interpretation remains underdefined |
A defensible wording is: “The interpretation depends on the canopy-layer exchange context. Within-canopy gradients describe coupling or decoupling inside the canopy. Gradients across the canopy top may support canopy-top exchange diagnostics if height and roughness assumptions are explicit. Above-canopy gradients may support above-canopy exchange comparison, but should not be treated as generic surface-layer fluxes without context.”
Method roles in the implemented canon
The methods are not competitors in a single accuracy ranking. They determine different things.
| Method | What it determines directly | What is assumed or residual | Best used for | Main limitation |
|---|---|---|---|---|
| Priestley–Taylor | LE_PT from available energy and empirical partitioning | H_PT is the complement | LE-oriented partition for moist vegetation | Depends on alpha_PT and surface assumptions |
| Bulk–Residual | H_bulk from temperature gradient and exchange assumption | LE_res is the remaining available energy | H estimate from two-height station data; residual closure demonstration | Neutral transfer assumption; LE_res absorbs errors in Q*, B, and H_bulk |
| Bowen Ratio | H/LE partition through beta | Closure only for valid finite beta | H/LE partition when T/RH gradients are reliable | Fragile for weak or noisy humidity gradients |
| Penman | LE_Penman | U_Penman remains open | Latent heat or evaporation comparison | No direct H |
| Monin/Profile diagnostic | H_MO and LE_MO from profile/stability logic | R_E,MO remains visible | Profile and stability diagnosis | Input- and parameter-intensive; not force-closed |
| Closure diagnostics | R_E, closure ratio, open remainders | No flux is computed | Interpreting method semantics | Diagnostic only |
Method-specific decision notes
Bulk–Residual
Bulk–Residual estimates sensible heat from a two-height temperature gradient and an exchange assumption. The corresponding latent heat is defined as the remaining available energy.
\[ H_\mathrm{bulk} = \rho c_p \frac{t_1 - t_2}{r_a} \]
\[ LE_\mathrm{res} = Q^* - B - H_\mathrm{bulk} \]
Bulk–Residual is appropriate when the dataset contains two measurement heights and the target is a direct sensible-heat estimate plus an explicitly residual latent-heat term. Its limitation is the neutral transfer assumption. The residual term absorbs errors in Q*, B, and H_bulk.
Priestley–Taylor
Priestley–Taylor is an available-energy partitioning method. It is useful when the target is LE-oriented partitioning, especially for moist vegetated surfaces.
\[ LE_\mathrm{PT} = \alpha_\mathrm{PT} \frac{sc}{sc + \gamma} (Q^* - B) \]
\[ H_\mathrm{PT} = (Q^* - B) - LE_\mathrm{PT} \]
Its limitation is that H_PT is a complement, not an independently estimated sensible heat flux. The result depends on alpha_PT and surface assumptions.
Bowen Ratio
Bowen is useful when temperature and humidity gradients are reliable. It partitions available energy through a gradient ratio.
\[ H_\mathrm{BR} = \frac{\beta}{1 + \beta} (Q^* - B) \]
\[ LE_\mathrm{BR} = \frac{1}{1 + \beta} (Q^* - B) \]
Its limitation is denominator fragility. Weak or noisy humidity gradients can dominate the result.
Penman
Penman is an LE-oriented method.
\[ LE_\mathrm{Penman} = f(A, VPD, r_a, r_s) \]
\[ U_\mathrm{Penman} = A - LE_\mathrm{Penman} \]
U_Penman is an unresolved remainder. It is not H. Penman is useful for evaporation or latent-heat comparison, not for a full paired H/LE solution.
Monin/Profile diagnostic
Monin/Profile is a profile and stability diagnostic path. It should not be used as automatic truth or forced closure.
\[ R_{E,\mathrm{MO}} = Q^* - B - H_\mathrm{MO} - LE_\mathrm{MO} \]
Its value is that R_E,MO remains visible. Its limitation is that the result is input- and parameter-sensitive and depends strongly on the quality and meaning of the selected two-height profile.
Final rule
A heat-flux method should be selected by measurement design and target quantity. Priestley–Taylor partitions available energy through an empirical latent-heat formulation. Bulk–Residual estimates sensible heat from a two-height gradient and defines latent heat as the residual. Bowen partitions available energy through temperature and humidity gradients. Penman estimates latent heat and leaves the remaining energy unresolved. Monin/Profile provides a process-oriented diagnostic and keeps the residual visible.