General comments

The manuscript is of very high quality. I have some minor comments and technical correction that, in my view, would benefit the text.  

Specific comments

First paragraph of section 4.2: This entire paragraph discusses the (lack of) effect of planted tree diversity on EF. I think this discussion is somewhat not relevant and maybe misleading, since the observed tree diversity (the actual tree diversity of the environments in the time of measuring) had a positive effect on EF. The lack of effect of planted tree diversity is just and artifact of the lack of correlation between planted diversity and current diversity (due to mortality and other ecological mechanism not addressed in the manuscript). For example the second sentence: “This contrasts with widespread observations of increasing ecosystem functions with greater tree diversity, as reported for aboveground productivity (Zheng et al., 2024).”, is problematic because the authors are talking about planted tree diversity and not actual observed diversity and, additionally, the authors did found indirect effect of tree diversity on EF via structural complexity.  My suggestion is to re-write this paragraph focusing on why planted tree diversity did not affected observed tree diversity and consequently EF.

The discussion presented in lines 434-437 contradicts the discussion presented in the first paragraph of section 4.2 (commentary above).

Technical corrections

Standardize the supplementary material  naming. You call it supplementary material in the main text, and Appendix at the end. Same for supplementary figure naming (ie.: S1 vs. A1)

Figure S1/A1 is not a “conceptual structural equation model, relationships and assumed mechanisms”, ratter a scatter plot ET/EF vs. Shortwave radiation. The authors should include the graphical representation of the conceptual structural equation model, and scatter plots of individual relationships (ie.: diversity vs. ET).

Table S1 is not shown in the appendix.

Figure 3: the path from “Island Area” to “Structural Complexity” can be confused with a path from “Observed woody plant diversity” to “Structural Complexity”. Maybe change the figure a bit to make this path more clear.

Line 395. Supplementary material S3 not shown.

This manuscript investigates evapotranspiration (ET) and evaporative fraction (EF) across 52 experimental tree islands in an oil-palm landscape using UAV thermography, the DATTUTDUT energy-balance model, random-partitions analysis, and piecewise structural equation modelling. The study addresses a relevant gap by quantifying ET and EF across the full tree-island area and planted-diversity gradients of the EFForTS-BEE experiment.

The manuscript is generally well structured and presents a high-quality study.

I identified some aspects for which additional clarification or sensitivity analyses would strengthen the interpretation. These principally concern the selection of one observation per island, the partial confounding of island area with palm removal, the proposed edge-based explanation, and the reporting of direct and indirect SEM effects. Overall, the positive relationship between island area and EF appears to be an important contribution. Clarifying the robustness of this relationship, its separation from the establishment treatment, and the evidence for the proposed underlying mechanisms would further improve the manuscript.

Major Comments

1. Selection of one observation per island

Selecting the observation with the highest incoming radiation is an understandable approach to standardize measurement conditions, particularly because 98 % of the selected observations fall between 850 and 1000 W m-2. However, comparable radiation does not necessarily imply comparable soil-water availability, vapour-pressure deficit, wind conditions, or antecedent rainfall.

A clearer explanation of the selection rule would help readers understand this analytical choice. This could include reporting within-island EF variability and comparing the reported area effect with an analysis based on the repeated observations or another suitable island-level summary. If the magnitude of the relationship is sensitive to the selection method, the reported 17% increase could be described more specifically as applying to the selected high-radiation observations.

2. Island area is partially confounded with palm removal

The 25 m² islands were established without removing oil palms, whereas 40 % of the palms were removed from the 100, 400, and 1600 m² islands. The smallest area class therefore differs from the larger classes in both island area and establishment treatment. These differences could include residual palm density, light availability, competitive conditions, and potentially soil disturbance and water availability.

Because palm removal is tied to area class, its contribution cannot be fully separated simply by adding it as another predictor. I suggest acknowledging this feature of the experimental design more explicitly. If supported by the available sample size, a useful sensitivity analysis would be to assess the area relationship using only the 100, 400, and 1600 m² classes, which received the same nominal palm-removal treatment. If the positive relationship remains evident among these three classes, this would strengthen its interpretation as an area effect. If it depends mainly on the contrast with the 25 m² class, the interpretation should acknowledge that island area and establishment treatment cannot be clearly separated for this particular comparison.

3. Edge explanation

The manuscript reports higher EF near the edges of the 400 and 1600 m² islands. Under comparable island geometry and edge-effect width, however, smaller islands would be expected to contain a larger proportion of edge-proximate area. A higher EF near edges might therefore raise plot-mean EF more strongly in smaller islands, whereas the observed relationship shows increasing mean EF with island area.

It would be helpful if the authors could explain quantitatively how the reported distance-to-edge gradient relates to the observed increase in plot-mean EF with island area. This could include a clearer definition of the edge zone, its estimated spatial extent, and the proportion of each island represented by different distance-to-edge classes. The principal methods and results of the analyses referenced in Supplementary Tables S2 and S3 should also be summarized sufficiently in the main manuscript to make this interpretation traceable. A concise analysis or conceptual explanation connecting the within-island edge gradient to the among-island means would considerably strengthen the proposed mechanism. Until this relationship is demonstrated more clearly, I suggest moderating the statement in the Abstract that the area effect “can be explained” by the edge gradient. A formulation such as “may be related to the observed edge gradient” would better reflect the current evidence.

4. Direct and indirect effects

The path coefficients reported in the Results differ from those shown in Figure 3. It would be helpful to clarify whether the text and figure present different effect metrics or standardization procedures and to ensure numerical consistency across the manuscript.

Figure 4 already presents bootstrapped direct, indirect, and total effects. Reporting the corresponding numerical estimates and confidence intervals in a table would make the relative contributions of the pathways easier to evaluate. In particular, it would be useful to distinguish:

• the positive serial pathway through observed woody plant diversity and structural complexity,
• the pathway from island area through structural complexity to EF, which opposes the positive serial pathway,
• the combined indirect effect, and
• the direct and total effects.

The conclusion states that island area enhances evaporative cooling “primarily” by increasing woody diversity and structural complexity. This wording does not appear fully aligned with the stronger direct area-to-EF relationship shown in the results. I therefore suggest distinguishing more clearly between the consistently detected overall relationship between island area and EF and the more model-dependent interpretation of the indirect pathways.

Minor Comments

1. Spatial context: Figure 1 shows the distribution of the tree islands but does not include the plantation boundary, major roads, or surrounding land cover. Some additional spatial context would help readers assess whether proximity to plantation edges or contrasting land-cover types could covary with island-area class or influence the thermal measurements. If this information is available in the original experimental-design paper, a brief summary and reference here may be sufficient.
2. Temperature endpoints. Please clarify whether the DATTUTDUT hot and cold endpoints were calculated separately for each island or from the complete mission-level orthomosaic. If they were calculated separately for each island, it would be helpful to assess sensitivity to the strongly unequal pixel numbers among area classes.
3. Sample composition. The random-partitions analysis uses N=48, whereas the SEM uses N=52. Please summarize clearly which plots were included in each analysis and the rationale for the different sample sizes.
4. Language and typographical corrections. Please replace the noun “analyzes” with “analyses” throughout, correct “Priestly–Taylor” to “Priestley–Tylor,” and correct “pievewiseSEM” to “piecewiseSEM.” The sentence “The direct path was associated with at high standardized effect size” should read “The direct path was associated with a high standardized effect size.”