Evapotranspiration (ET) – the combined water flux from soil and vegetation to the atmosphere – is a key component of water cycling and climate regulation, and strongly affected by land-use changes. The evaporative fraction (EF), representing the proportion of available energy allocated to ET, is often preferred over ET as a target variable in studies involving repeated measurements under varying weather conditions. In Sumatra's lowlands in Indonesia, (evapo)transpiration of dominant land-use types including oil palm monocultures is well studied; however, there is a lack of studies assessing ET (or EF) across diverse mosaic landscapes and types of land-use such as oil palm agroforests. Across 52 experimental plots – forest restoration patches known as “tree islands” – in an oil palm landscape (EFForTS-BEE), we tested whether the experimental treatments “planted tree diversity” and `tree island area' influence ET and EF as derived from UAV (uncrewed aerial vehicle)-based thermography and subsequent energy balance modeling. A random partition linear model showed that planted tree diversity (1, 2, 3, or 6 species) did not affect plot-level ET or EF, whereas tree island area (25, 100, 400, or 1600 m2) had a positive effect, with EF increasing by 17 % from the smallest to the largest tree islands. A structural equation model revealed that the effect of tree island area on EF was mediated by both direct and indirect pathways. Specifically, a strong direct effect of island area on EF (Std.Beta = 0.44, p<0.001) was complemented by an indirect pathway through increased observed woody plant diversity and stand structural complexity. Stand structural complexity had a positive effect on EF (Std.Beta = 0.20, p<0.05), while neither the vegetation index GNDVI nor tree height variability had significant effects. The decrease in EF with increasing distance from the edge within larger tree islands suggests that edge-related spatial variation may be related to, but does not fully explain, the positive area effect. Our findings suggest that larger tree islands enhance ET and EF through structural and biodiversity-related mechanisms. This underscores the importance of tree islands in human-modified landscapes, not only as biodiversity refugia but also as functional elements that support climate regulation.
Deutsche Forschungsgemeinschaft192626868532868192532776526Introduction
The conversion of tropical rainforests is ongoing and impairs biodiversity, ecosystem functioning and services (Edwards et al., 2013; Grass et al., 2020; Potapov et al., 2024). Commodity-driven deforestation is a major driver and often leads to landscapes dominated by large-scale monocultural plantations. In recent decades, oil palm cultivation has expanded, particularly in the humid tropics, with Indonesia being the world's largest producer of palm oil (FAO, 2023). Oil palm cultivation contributes to food security, improves the livelihoods of local people, and enables macro-economic development (Kubitza et al., 2018; Qaim et al., 2020). However, it has also contributed to a massive decline in biodiversity and has compromised the carbon and hydrological cycles (Kotowska et al., 2015; Merten et al., 2016; Guillaume et al., 2018; Grass et al., 2020), as well as the microclimatic regulation of air and land-surface temperatures (Meijide et al., 2018; Pallavi et al., 2024). Consequently, there is an intensive search for more sustainable modes of oil palm cultivation, which may include reduced application of agrochemicals (Iddris et al., 2023) or the integration of trees into the cultivation system (Teuscher et al., 2016; Zemp et al., 2023; Wenzel et al., 2024).
Tree islands, i.e., local tree plantings of native species within the oil palm-dominated landscape, are being studied within the Biodiversity Enrichment Experiment (EFForTS-BEE) in lowland Sumatra, Indonesia. The experiment consists of 52 experimental tree islands of varying tree diversity and area, distributed across an oil palm-dominated landscape. The experimental enrichment of tree diversity was motivated by the generally positive relationship between biodiversity and ecosystem functioning (Grossman et al., 2018). Tree diversity usually enhances productivity, ecosystem carbon storage, transpiration and decomposition rates (Paquette and Messier, 2011; Kunert et al., 2012; Zemp et al., 2023; Zheng et al., 2024). The treatment island area was motivated by the theory of island biogeography, which states that larger islands harbor more species than smaller islands, driven by species colonization and extinction (MacArthur and Wilson, 1963; Gooriah et al., 2020). This theory also plays a major role in nature conservation (Givnish, 2024), for example regarding the size of protected areas or fragments in patchy landscapes (Arroyo-Rodríguez et al., 2020).
Establishing the tree islands in the experimental EFForTS-BEE landscape had pronounced positive effects on biodiversity, for example by increasing tree diversity by +4.7 compared to oil palm monoculture, and on key ecosystem functions such as water infiltration, which was achieved without reducing landscape-level oil palm yields (Zemp et al., 2023). Both treatments – planted tree diversity and in particular island area – enhanced the taxonomic, phylogenetic and functional diversity of recruiting native woody species (Paterno et al., 2024). While EFForTS-BEE revealed considerable potential to alleviate biodiversity decline and impaired ecosystem functions by integrating tree islands into oil palm landscapes, a comprehensive assessment regarding evapotranspiration (ET) is still missing.
ET is an ecosystem function that contributes to climate regulation and water cycling. The prominent role of the tropical rainforest in global evaporative cooling was highlighted by Lawrence et al. (2022), but this function is threatened by forest conversion to other land-use types. However, ET from oil palm, particularly from mature and intensively managed plantations, has been reported to match or even exceed that of tropical forests (Sabajo et al., 2017; Manoli et al., 2018; Röll et al., 2019; Gómez et al., 2023) – a pattern similarly observed in tree plantations and early successional tree stands, which can also exhibit high ET rates (Dierick et al., 2010). Landscape fragmentation and increasing edge density can influence the species composition of forest remnants by favoring certain functional traits while disfavoring others (Pinho et al., 2024). These trait shifts can influence ecosystem functioning, potentially affecting processes like water use and energy partitioning. Islands intrinsically comprise edges, and edges can have enhanced transpiration rates compared to the interior (Giambelluca et al., 2003; Kunert et al., 2015). This effect might play a role when comparing the average ET of tree islands which differ in edge length.
Assessments of ET from forests or other land-use types in human-modified landscapes are often based on eddy covariance or sap flux measurements (Niu et al., 2015; Meijide et al., 2017). Their strengths include the provisioning of high-resolution time series, potentially also over long periods. Weaknesses include the restriction to a relatively low number of localities and large measurement footprints mixing multiple landscape types or entities such as trees. It is therefore difficult to assess patchy landscapes such as EFForTS-BEE, with its 52 tree islands and more than 10 000 trees, using such methods. At such large scales, remote sensing approaches may be more feasible (Senf, 2022), for example using uncrewed aerial vehicles (UAV) as a monitoring platform. ET derived from UAV-thermography and subsequent energy balance modeling has previously been applied in horticulture (Park et al., 2021; Riveros-Burgos et al., 2021) and forests (Bulusu et al., 2023; Cortés-Molino et al., 2024). A well-established approach uses maps of land surface temperatures and solar radiation data to predict ET with the DATTUTDUT (Deriving Atmosphere Turbulent Transport Useful To Dummies Using Temperature) energy balance model (Timmermans et al., 2015). This approach has been implemented in a freely available QGIS Plugin, QWaterModel (Ellsäßer et al., 2020b). In a mature oil palm plantation of lowland Sumatra, it was tested against eddy covariance measurements (Ellsäßer et al., 2021). The derived estimates of latent heat flux agreed very well between the two methods across variable weather conditions and times of day, and further analyses even suggested statistical interchangeability (Ellsäßer et al., 2021). The same UAV-based approach was also successful in predicting stomatal conductance and sap-flux-derived transpiration rates by trees and palms in a subset of tree islands in EFForTS-BEE (Ellsäßer et al., 2020a). However, one central problem with using UAV-based observations to analyse spatial differences in ET is the variability of weather conditions, such as solar irradiance and air temperature, during the observation period. This can either be addressed by only analysing ET data recorded under specific, comparable climatic conditions (e.g., high evaporative demand), or by using the evaporative fraction (EF), i.e., the ratio of latent heat flux to available energy, or the ratio of actual ET to potential ET (ETa/ETpot), as a target variable. Both ETa/ETpot and EF are less susceptible to differences in available energy, and as such, within typical ranges, fluctuate less with varying weather conditions than ET.
The effects of planted tree diversity and tree island area on ET and EF remain largely unexplored. While both variables have been linked to ET in previous studies (Baldocchi, 2005; Zhang et al., 2022), they are not recognized as direct drivers of ET or EF in oil palm landscapes, including the EFForTS-BEE experiment. In urban settings as well as in Neotropical natural forests and plantations, tree diversity was reported to enhance transpiration (González-Espinosa et al., 2004; Kunert et al., 2012; Wang et al., 2021). For island area, it was found that vapor pressure deficit (VPD) and water stress varied between small urban tree islands and larger urban forest areas (Cregg and Dix, 2001). The effects may indirectly be driven by other factors, such as the structure or tree height variability of a given stand, which may affect turbulent energy exchange. Ehbrecht et al. (2017) reported a significant relation between stand structure and microclimatic variables such as daily VPD amplitude. Tree height is known to be related to forest thermal balance as well (Barbeta et al., 2023) and may be used as an indicator of surface roughness. Vegetation health, expressed in indices such as the NDVI, was also shown to affect ET, and its integration into ET modeling improved the quality of predictions (Yang and Wang, 2011; Er-Raki et al., 2013). Thus, we expect planted tree diversity and island area to have indirect effects on ET or EF via mediator variables such as observed woody plant diversity, vegetation stand structural complexity, tree height variability, and vegetation health.
In our study, we tested the effects of tree diversity and island area on ET and EF across 52 tree islands, ten years after their establishment, using UAV-based thermography and subsequent energy balance modeling.
We hypothesized that EF increases with increasing
H1.
Planted tree species diversity,
H2.
Area of tree islands,
H3.
Observed diversity of woody species,
H4.
Stand structural complexity,
H5.
Variability in tree height, and
H6.
Vegetation health.
For a conceptual structural equation model, relationships, and assumed mechanisms please see the Supplement (Fig. S1, Table S2).
MethodsStudy site
The study was conducted in the lowlands of Sumatra, Indonesia in Jambi province, near the village Bungku. The experiment EFForTS-BEE (103.25° E, 1.95° S) is situated in an oil palm monoculture owned by PT. Humusindo Makmur Sejati. Rainfall in the region is 2235 mm yr−1 and average annual temperature is 26.7 °C; dominant soil types are Acrisols (Drescher et al., 2016). The natural vegetation would be tropical rainforest as described in Rembold et al. (2017). After rainforest conversion oil palms were cultivated. EFForTS-BEE was established in 2013 with 52 tree islands within an area of oil palm monocultures of 140 ha, then between 6 and 12 years old (Teuscher et al., 2016).
Experimental design
The treatments in EFForTS-BEE consist of planted tree species diversity (0, 1, 2, 3 6 different tree species) and tree island area (25, 100, 400, 1600 m2) (Fig. 1). Six native multi-purpose tree species (Parkia speciosa Hassk., Fabaceae; Archidendron jiringa (Jack) I.G. Nielsen, Fabaceae; Durio zibethinus L., Malvaceae; Peronema canescens Jack, Lamiaceae; Rubroshorea leprosula (Miq.) P.S.Ashton & J.Heck., Dipterocarpaceae; Dyera polyphylla (Miq.) Steenis, Apocynaceae) were planted. For each tree island area class, there is one control plot with conventional oil palm management, representing the original monoculture palm pattern with no tree planting. Before tree planting 40 % of oil palms were removed inside of the 100, 400 and 1600 m2 plots to increase light availability (Teuscher et al., 2016). In contrast, the 25 m2 plots were established between oil – palm rows, and no oil palms were removed. This design reflects a compromise between increasing light availability for the planted trees and retaining the prevailing oil – palm management system. Consequently, the 25 m2 class differs from the larger area classes in both island area and establishment treatment. In addition to tree planting, the experiment also allowed for natural plant recruitment. The 25 m2 plots were placed between palm rows and no oil palms were removed. In addition, there are four control plots where oil palm is managed as usual (Teuscher et al., 2016). The experimental design follows the random partitions approach which allows using stepwise linear regression models to analyze diversity gradients (Bell et al., 2009).
UAV data acquisition
UAV flight missions were carried out on 18 d in the early dry season between 9 June 2023 and 11 July 2023. Flights were conducted between 10:00 and 14:00 local time to collect data under peak radiation conditions and also to minimize shadows. As the battery capacity of the UAV limited flight times to approx. 20 min per charge, the area of EFForTS-BEE was divided into 5 routes of roughly equal size. Each route covered between 10 and 12 plots, wherein the sequence of the flight routes was alternated among days to ensure that plots were visited at different times of the day throughout the campaign. Flight routes were planned using the UgCS flight planning software (SPH Engineering, Lativa). Flight speed was set to 5 m s−1 with an 80 % overlap of images. Flight altitude was between 90–100 m above the terrain, resulting in a pixel size of ca. 2 cm for RGB and ca. 10 cm for thermal images.
We used a DJI M300 RTK system (SZ DJI Technology Co., Ltd., China) to carry multiple payloads simultaneously during the flight campaigns. To record land surface temperature, the UAV was equipped with a Zenmuse H20T camera that contains a radiometric thermal sensor with a spatial resolution of 640×512 pixels, a spectral coverage of one long-wave infrared band (7.5–13.5 µm), and a sensitivity of ≤50 mK at f/1.0 (DJI, 2021) (Fig. 1). For measuring shortwave radiation, the UAV carried an Apogee SP-710 albedometer (Apogee Instruments, Inc., USA), consisting of an upward facing (SP-510) and a downward facing (SP-610) thermopile pyranometer. The recorded irradiance values were stored on an Apogee AT-100 microCache Bluetooth logger. For recording air temperature and relative humidity, the UAV was equipped with an iButton DS1923-F5 Hygrochron sensor (iButtonLink LLC, USA). To acquire multispectral images, the Zenmuse H20T was swapped with a MicaSense red-edge MX (MicaSense, Inc., USA), which captures five spectral bands (blue [459–491 nm], green [546.5–573.5 nm], red [660–676 nm], red edge [711–723 nm], near infrared [813.5–870.5 nm]) at a spatial resolution of 1280×960 pixels and a spectral resolution of 12 bit. It is equipped with an upwards oriented downwelling light sensor (MicaSense DLS 2), which records global lighting changes to correct reflectance values of each band (MicaSense, 2020). Both cameras were triggered by the overlap mode, which uses the internal GPS to ensure enough overlap of pictures at flight altitude.
The experimental design of EFForTS-BEE with 52 tree islands varying in planted tree diversity and island area (island size not drawn to scale). The islands are dispersed in a monocultural oil palm plantation of 140 ha. Red-Green-Blue (RGB), land surface temperature (LST, °C) and evapotranspiration (ET, mm h−1) of one large tree island (1600 m2) with three planted tree species.
UAV-data processing
Raw 3-band JPG image files from the thermal camera were converted to 1-band TIFF format using the dji_h20t_rpeg_to_tif R script by Kattenborn (2023) translating them to land surface temperatures. Subsequently, all images of a given mission were processed to one thermal orthomosaic map displaying land-surface temperatures using the photogrammetry software Agisoft Metashape Professional (Agisoft, Russia). The orthomosaics were then re-sampled to a ground resolution of 11 cm and the individual plots were clipped from the orthomosaics using QGIS version 3.34 (QGIS Development Team 2025). Orthomosaics of the single images from the multispectral camera were processed using Pix4Dmapper (Pix4D S.A., Switzerland), therein applying automated image correction from the DLS 2 radiation sensor and respective calibration images that were captured on the ground before each flight (MicaSense, 2020).
Energy-balance modeling
For the energy-balance modeling, we used the QGIS plugin QWaterModel version 1.5 (Ellsäßer et al., 2020b), which implements the DATTUTDUT model by Timmermans et al. (2015). This model was successfully tested against eddy covariance data in an oil palm plantation in the same region (Ellsäßer et al., 2021) and was subsequently also used for other land-use types in the region (Bulusu et al., 2023; Cortés-Molino et al., 2024). DATTUTDUT is a one source energy balance model that defines the maximum and minimum land surface temperature (LST) in an image based on the 100th and 0.5th percentiles of pixel values, respectively. For each observation, the LST orthomosaic was first clipped to the polygon of each tree island, and QWaterModel was then run separately for each clipped tree island raster. Thus, the hot and cold temperature endpoints were calculated from the pixels within each individual tree island. This percentile-based approach avoids relying on individual pixel values, thereby reducing the influence of outliers or noise. The assumption is that cold pixels indicate high evapotranspiration while hot pixels indicate very low to no evapotranspiration (Timmermans et al., 2015). Incoming shortwave radiation as measured by the on-board sensors is a further key input for QWaterModel (Brenner et al., 2018; Ellsäßer et al., 2020b; Cortés-Molino et al., 2024). Additionally, mean air temperature during flight was given to QWaterModel, while all other settings remained at default values. As output, QWaterModel generates a 6-band raster file: net radiation Rn (W m−2), latent heat flux LE (W m−2), sensible heat flux H (W m−2), ground heat flux G (W m−2), evaporative fraction EF and evapotranspiration ET (mm h−1) (Ellsäßer et al., 2020b). Therein, EF is the proportion of available land surface energy that is used for LE. It is calculated as EF =LE/ (LE+H). Out of all conducted flights, 43 met the quality criteria, which were clear skies, high incoming radiation and complete, well aligned orthomosaics. The number of observations per tree island ranged from 5 to 10. To standardize measurement conditions, we selected, for each tree island, the observation recorded under the highest incoming shortwave radiation. This selection was made before analysing the factors controlling EF and was intended to minimize variation in radiation among observations. The selected observation was therefore not necessarily representative of the mean EF of each tree island during the measurement campaign. 98 % of these selected observations (n=52) were obtained under high incoming shortwave radiation (850–1000 Wm-2). The selected flight days showed broadly comparable soil-moisture, antecedent-precipitation, and wind conditions to the other usable flight days, as measured by a climate station in the oil palm plantation (Supplement, Table S1). In contrast to ET, EF is only marginally affected by varying levels of radiation (Fig. A1). For both ET and EF, the mean of all pixels within a given tree island were used for the analyses. By design, EF and ET and their mean values over the area of a given tree island, are independent of scale. To confirm this assumption, tree islands were sub-sampled and the effect of island area within artificially reduced islands was tested (see Table S4).
Vegetation health
Multispectral images were used to create orthomosaics of vegetation indices relating to vegetation health (Hernández-Clemente et al., 2019). The following analyses use the Green Normalized Differential Vegetation Index (GNDVI) since it is sensitive to chlorophyll contents at medium to high LAI (Gitelson and Kaufman, 1998). As described by Gitelson et al. (1996), GNDVI was calculated as:
GNDVI=NIR-Green(NIR+Green).
Observed woody plant diversity
Data on observed woody plant diversity, including planted as well as naturally regenerated plants, were recorded in 2022. The inventory included all free-standing woody plants with a height greater than 130 cm from the soil surface. Specimens were identified as described in Paterno et al. (2024). A total of 77 species (excluding planted trees) were recorded, with a plot mean of 8.35 (SD 6.37). Hill number q1 was used as a measure of observed woody plant diversity; H1 corresponds to the exponential Shannon entropy also referred to as “typical” diversity (Chao et al., 2014). The R package “hillR” version 0.5.2 (Li, 2018) was used to calculate H1.
Tree height
Maximum tree height and oil palm meristem height were measured for all woody plants in 2022, i.e., oil palm, planted trees and natural regeneration. For heights lower than 2 m, we used a standard tape measure, between 2 and 8 m a measuring rod and for trees taller than 8 m a Vertex (Haglöf, Sweden) device (Zemp et al., 2019b). Average tree height was 5.07 m (SD 3.93) while the oil palms had an average meristem height of 7.8 m high (SD 1.13). The Gini index of tree height was calculated using the R package “ineq” version 0.2-13 (Zeileis, 2014); it is a measure of the unevenness of tree height and serves as a proxy for surface roughness of the stand in our study.
Stand structural complexity
Terrestrial laser scanning (TLS) was carried out in 2022 using a FARO Focus M70 (Faro Technologies Inc., Lake Mary, USA) laser scanner. A stand structural complexity index (SSCI) was calculated as described by Ehbrecht et al. (2017). These data were previously published by Kikuchi et al. (2024). SSCI captures the structural complexity of forest stands by combining the mean fractal dimensions (MeanFrac) with the effective number of layers (ENL) (Ehbrecht et al., 2017). A higher SSCI may reflect a greater surface area available for ET, which could promote higher energy partitioning towards latent heat, therefore SSCI could be linked to EF.
Statistical analyses
The effects of island area, tree species richness (linear and nonlinear effects) and tree species identity on EF were tested using a random partitions analysis (Bell et al., 2009; Teuscher et al., 2016). Assumptions for linear models (normal distribution of residuals) were checked for the response variable EF via the quantile–quantile plots using the package “stats” version 4.3.2 (Team, 2023). The assumption of heteroscedasticity was checked using the “performance” package version 0.13.-0 (Lüdecke et al., 2021). EF was logit transformed following recommended practices (Warton and Hui, 2011; Li et al., 2023). The overall random partitions model is stated as:
y=β0+βLRXLR+∑i6βixi+βNLRxNLR+βpxp+e,
where y is the response variable, in our case EF; β0 is the model intercept; XLR is planted enrichment tree-species-richness treated as a continuous variable (“linear richness”); xi is an indicator of the presence or absence of a certain species (i) from the pool of six planted tree species; xNLR is the “non-linear richness”, which is the effect of species richness levels as factors; xp is island area as a factor; and e is a normally-distributed error term (Li et al., 2023). In EFForTS-BEE, the random partitions model focused on 48 plots, comprising 12 plots per island-area class. The four plots without planted trees and the four “business-as-usual” control plots were not part of the analysis, since they do not contribute to the species richness gradient. All analyses were performed in R, using code developed by Li et al. (2023).
Structural equation modeling (SEM) was used to investigate the direct and indirect effects of the treatments (planted tree diversity and island area) on EF. Hypotheses to be tested are the direct influences of planted tree diversity and island area on EF. Further hypotheses include potential indirect effects via endogenous variables. In our study, the endogenous variables observed woody plant diversity, vegetation stand structural complexity, tree height variability and vegetation health are considered. Endogenous variables may act as mediators, indirectly linking planted tree diversity, island area, and observed woody plant diversity to EF. Piecewise SEM are made up of several linear models, which are:
Model 1 = lm(Observed woody plant diversity ∼ Planted tree diversity + Island area)
Model 2 = lm(Structural complexity ∼ Planted tree diversity + Island area + Observed woody plant diversity)
Model 3 = lm(Tree height variability ∼ Planted tree diversity + Island area + Observed woody plant diversity)
Model 4 = lm(Vegetation health ∼ Planted tree diversity + Island area)
Model 5 = lm(EF ∼ Planted tree diversity + Island area + Observed woody plant diversity + Structural complexity + Tree height variability + Vegetation health)
The initial models 1, 2, 3 and 4 were built based on prior knowledge and assumptions, following the “weight of evidence” approach (Grace, 2020). We assumed that structural complexity, observed woody plant diversity, vegetation health and tree height variability are directly influenced by the treatments, planted diversity and island area. For model 5, we assumed that differences in evaporative fraction are driven by the treatments as well as by all endogenous variables. The endogenous variables were not correlated among each other (spearman correlation <0.5, see Fig. A2). The assumption of normality of residuals was tested using the Shapiro–Wilk test. Normality could not be assumed, so EF was transformed (standardized Box Cox transformation) using the “bestNormalize” R package version 1.9.1 (Peterson, 2021). For the calculation of standardized effect sizes in the structural equation model, all variables except – planted tree diversity and tree island area – were standardized using the “vegan” package version 2.6-4 (Dixon, 2003). The SEM was built using the “piecewiseSEM” package version 2.3.0 (Lefcheck, 2016). Non-parametric bootstrapping with 1000 bootstrap replicates was used to calculate direct, indirect, and total effects, including confidence intervals for each predictor, using the “semEff” package version 0.6.1 (Murphy, 2022). Standardized path coefficients describe the individual structural relationships represented by the arrows in the SEM (Fig. 3), whereas the bootstrapped direct, indirect, and total effects summarize contributions across the corresponding direct and mediated pathways (Fig. 4). The complete decomposition of the indirect effects of island area on EF and their bias-corrected and accelerated (BCa) confidence intervals are provided in Table S3. The “business-as-usual” plots were not included in the SEM because their continued management might introduce confounding effects; the SEM therefore included the 52 plots belonging to the factorial experimental design, including the four plots without planted trees.
Results
The mean ET across the 52 islands was 0.74 mm h−1 (SD: 0.09 mm h−1) at the time the thermal images were recorded. The minimum ET observed was 0.49 mmh-1, and the maximum was 0.95 mmh-1. The mean EF was 0.7 (SD: 0.06), with values ranging from 0.56 to 0.83. To assess whether the unequal numbers of pixels among island-area classes affected the island-specific LST endpoints, we compared mean LST, the mean of the 1 % hottest and the mean of the 0.5 % coldest pixels among area classes. None differed significantly (Kruskal-Wallis tests: p=0.50, p=0.48, p=0.48, respectively, Figs. S2–S4). GNDVI ranged between 0.57 and 0.86. Significant linear relationships were found between EF and tree island area (p<0.001), observed woody plant diversity (p<0.001), and tree height variability (p<0.01) (Fig. A3).
Random partitions model: effects of planted tree diversity and tree island area
No significant effects of continuous richness, species identity, or non-linear richness on EF were detected using the random partitions model. However, tree island area had a significant positive effect (p<0.05) on EF (Fig. 2, Table 1).
Controls of evaporative fraction (EF) as indicated by a random partitions model. EF was Logit transformed and unit variance scaled.
Specifically, significant effects (p<0.05) were observed for island areas of 25, 100 and 1600 m2. With each increase in tree island area, the standardized effect on EF rose by approximately 0.5. From the smallest island area (25 m2) to the largest (1600 m2), EF increased by 17 %, from an average of 0.65 to 0.76. To test whether the observed island-area effect depended on selecting the observation with the highest incoming shortwave radiation, we conducted a sensitivity analysis using all EF observations available for the 48 tree islands. In each of 5000 iterations, one observation was randomly selected from each island, and the original random-partition model was refitted. Island area remained significant (p<0.05) in 99.4 % of the iterations, with a median p value of 0.00015. The mean raw EF increase between the 25 and 1600 m2 islands was 18.5 % across the random selections (95 % range: 11.22 %–26.41 %), compared with 17 % based on the selected high-radiation observations. Thus, both the direction and magnitude of the area relationship were similar when observations were selected independently of incoming radiation. Because palm removal differed between the 25 m2 islands and the three larger area classes, we conducted a sensitivity analysis using only the 36 islands in the 100, 400, and 1600 m2 classes, which all received the same nominal 40 % palm-removal treatment. In each of 5000 iterations, one repeated EF observation was randomly selected for each island, and the original random-partition model was refitted. Mean EF increased monotonically from 0.68 in the 100 m2 islands to 0.71 in the 400 m2 islands and 0.74 in the 1600 m2 islands. The ordering of the raw means (100<400<1600 m2) occurred in 92.1 % of iterations. The mean increase in raw EF between the 100 and 1600 m2 islands was 8.8 % (95 % sensitivity range: 3.0 %–14.7 %). Statistical support for the overall island-area effect was weaker than in the full analysis: island area was significant (p<0.05) in 57.2 % of iterations, with a median p value of 0.0347. Therefore, although excluding the 25 m2 class reduced statistical support for the overall area effect, the direction of the relationship and the size of the effect remained consistent among the three area classes that shared the same palm-removal treatment.
Standardized effect sizes of planted tree species diversity and tree island area on evaporative fraction as derived from the random partitions model. Effect sizes are shown with 95 % confidence intervals, N=48. Asterisks indicate p value levels: *** p<0.001, ** p<0.01, * p<0.05. Coefficients and p values are based on a sequential linear model (Bell et al., 2009).
Structural equation model: direct and indirect effects
Planted tree diversity had no significant direct or indirect effects on EF, challenging hypothesis H1. In contrast, hypothesis H2 was supported, as tree island area had a strong positive direct effect on EF (Std.Beta = 0.44, CI[0.28–0.64], p<0.001). The corresponding combined indirect effect was -0.044 (95 % BCa CI: -0.213–0.103), resulting in a total effect of 0.395 (95 % BCa CI: 0.314–0.532; Fig. 4). Among the individual indirect pathways, the serial pathway from island area through observed woody plant diversity and structural complexity to EF was positive (0.045, 95 % BCa CI: 0.010–0.130). In contrast, the pathway from island area through structural complexity alone was negative (-0.066, 95 % BCa CI: -0.185 to -0.014). The remaining indirect pathways were comparatively small, with confidence intervals overlapping zero. Estimates for all decomposed pathways are provided in Table S3. Hypothesis H3 and H4 – which predicted positive effects of observed diversity of woody species and stand structural complexity on EF – were both supported by significant SEM paths. Observed woody plant diversity positively influenced stand structural complexity, which in turn positively affected EF. Contrary to hypothesis H5, tree height variability had no significant effect on EF. Similarly, hypothesis H6 was not supported, as vegetation health showed no significant relationship with EF. Additional results not predicted by any hypothesis included a significant direct path from planted tree diversity to tree height variability (Std.Beta = 0.36, CI[0.14–0.52], p<0.001) and a significant direct path from observed woody plant diversity to tree height variability (Std.Beta = 0.25, CI[0.06–0.5], p<0.05).
Drivers of evaporative fraction as indicated by a piecewise structural equation model (N=52, Fisher's C=2.01, d.f. = 8, P=0.98). Solid lines indicate significant path coefficients (p<0.05). Symbols indicate p value levels: * p<0.05, ** p<0.01, *** p<0.001.
Standardized direct, indirect and total effect sizes of predictors on evaporative fraction in SEM. Error bars show 95 % CI, derived through bootstrapping with 1000 iterations. Symbols indicate p value level: * p<0.05.
Discussion
Our UAV-based assessment of 52 tree islands in an oil palm landscape revealed that tree island area had a direct positive effect on EF. Additionally, tree island area had an indirect positive effect on EF via observed woody plant diversity and structural complexity. Whilst planted tree diversity did not have any direct or indirect effects on EF, observed woody plant diversity had an indirect positive effect on EF via structural complexity.
ET and EF in context
The mean ET during high incoming radiation conditions from the tree islands in EFForTS-BEE was 0.74 mmh-1, with an evaporative fraction of 0.70. Data were filtered for high incoming radiation conditions (>850.00Wm-2) to ensure comparability among islands. These ET values are similar to those reported in other UAV-based studies in the same region. For example, in a mature and intensively managed oil palm plantation an ET of 0.78 mmh-1 was observed around noon, and the UAV-based method performed well compared to the eddy covariance method (Ellsäßer et al., 2021). Additional UAV- derived data from rain forest sites in the region showed ET rates between 0.5 and 0.8 mm h−1 under favorable atmospheric conditions, with plot-level variation attributed to site conditions (Bulusu et al., 2023). Overall, the observed ET rates and EF from the tree islands fall within the regional range and may be at the higher end of the spectrum compared to other land-use types. This is consistent with other studies indicating high evaporation or transpiration rates from early secondary forests and plantations in the tropics (Hölscher et al., 1997; van Dijk and Keenan, 2007; Dierick et al., 2010).
Direct and indirect effects of planted tree diversity and island area (H1 & H2)
In EFForTS-BEE, we did not find significant effects of planted tree diversity on ET or EF. It is likely that any influence of planted tree diversity on EF was superimposed by effects of observed woody plant diversity. Paterno et al. (2024) reported a positive effect (P=0.049) of planted tree diversity on taxonomic diversity of recruiting species six years after establishment; however tree island area showed a much stronger positive effect. The effects of observed woody plant diversity are discussed in the following section (Sect. 4.3). Notably, planted tree diversity did show a significant positive effect on tree height variability, consistent with the tree Height Variation Hypothesis (VHV) (Torresani et al., 2023), which uses tree height heterogeneity as a remote sensing-based biodiversity indicator. Marks et al. (2016) also reported a strong positive correlation between tree species richness and maximum tree heights in forests of eastern as well as western North America.
Island area was a strong predictor of EF, as indicated by both the random partitions model (Fig. 2) and SEM. EF increased significantly with increasing island area from 25 to 1600 m2 by 17 %. Interpretation of this area effect requires consideration of the experimental design. 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. Thus, the 17 % increase between the 25 and 1600 m2 islands cannot be attributed to island area independently of the establishment treatment. However, a sensitivity analysis restricted to the three larger area classes, which received the same nominal palm-removal treatment, showed a consistent and monotonical increase in EF with island area: mean EF increased by 8.8 % between the 100 and 1600 m2 islands. Although statistical support was weaker than in the full analysis, the direction of the relationship was highly consistent, indicating that the positive area relationship was not solely driven by the contrast with the uniquely treated 25 m2 islands. In the SEM (Fig. 3), tree island area had both direct and indirect effects on EF. The direct path was associated with a high standardized effect size (Std.Beta = 0.44, p<0.001) (Fig. 4), indicating that the positive island area – EF relationship was predominantly represented by the direct island area – EF effect in this model. This direct area effect may partly reflect differences in EF between the edges of the tree islands and the surrounding oil-palm plantation. Our spatial analysis detected significant edge-associated declines in EF with increasing distance from the edge in the 400 and 1600 m2 islands (Table S5), with the transition occurring at approximately 30 % of the island side length. However, because smaller square islands contain a greater proportion of edge-proximate area, this within-island gradient alone cannot explain the observed increase in mean EF with island area. We therefore interpret the edge gradient as evidence of spatial heterogeneity within the islands rather than as a sufficient explanation for the among-island area effect. This pattern aligns with findings by Giambelluca et al. (2003), showing an increased transpiration at the edge of a secondary forest in Vietnam. Using similar methods in the Brazilian Amazon, Kunert et al. (2015) likewise found enhanced transpiration at edges. However, other studies have reported the opposite. In a model-based study in the southwestern Amazon, Numata et al. (2021) found no ET difference between forest edges and interiors under high water availability. As dry seasons progressed, ET declined more near edges (up to 100 m) than in interior forests, especially during extreme droughts. Another modeling study in the Brazilian Atlantic Forest region reported that ET was reduced by 43 % at the edges of forest fragments when compared to the core (Dantas de Paula et al., 2015). We conclude that edges can induce changes in ET or EF, although the direction and magnitude of these effects vary among studies. This variability may reflect the origin of the forest patch, such as whether it is a remnant of old-growth forest with biomass loss at the edge (Ordway and Asner, 2020) or a recovering secondary forest, as well as the adjacent vegetation and energy advection, for example by wind. In our experiment, the observed edge-associated gradient indicates spatial variation in EF within the tree islands. However, because smaller islands contain a greater proportion of edge-proximate area, this gradient alone does not explain the increase in island-level mean EF with island area. The positive relationship between island area and EF may instead reflect a combination of edge-related and structural effects. On the indirect path, tree island area showed a strong positive path coefficient (R2=0.8, p<0.001) for observed woody plant diversity. This result aligns with a previous study in the same experiment (Paterno et al., 2024), which showed that increasing tree island area leads to higher taxonomic diversity of recruiting woody species. This result reflects the species-area relationship (SAR), were larger areas support greater diversity (Drakare et al., 2006). However, when accounting for sampling effort, woody plant diversity increases with tree island area, probably due to greater environmental heterogeneity in larger islands that supports a broader range of species and ecological strategies (Paterno et al., 2024). The SEM pathway continues from observed woody plant diversity to structural complexity. Earlier, 3 years after planting, Zemp et al. (2019a) found a non-linear increase of SSCI with planted tree diversity. Later, 9 years after planting, Kikuchi et al. (2024) found that diversity and size variation of recruiting trees became more important, as planted tree diversity had no effect on SSCI. Applying a structural equation model, Ma et al. (2024) found a significant positive indirect path linking tree species richness to coarse wood productivity in forests across a latitudinal gradient in China. Mean height, used as an indicator of vertical structure in their study, served as the mediator in this indirect path. In young temperate forest plots, a significant positive indirect effect of tree species richness on annual wood productivity (AWP) was identified (Ray et al., 2023). As in this study, SSCI acted as a mediator in their SEM. In a study in the Sardinilla tropical tree diversity experiment, Ray et al. (2024) found a positive direct significant path from tree species richness to canopy space filling (CSFI). In their SEM, CSFI was linked by a significant positive path to wood productivity, showing an indirect positive effect of tree species richness on wood productivity. Their result is in line with the outcome of our study, as both report an indirect path from diversity via structure towards an indicator of productivity. The positive serial pathway from island area through observed woody plant diversity and structural complexity was significant but comparatively small. Moreover, it was counteracted by other indirect pathways, including a negative pathway from island area through structural complexity to EF (Table S3). These results suggest that the positive island area – EF relationship in the SEM was dominated by the direct island area – EF, whereas the indirect pathway through observed woody plant diversity and structural complexity provides evidence for one possible, but limited mechanism, contributing to this relationship. Further underlaying patterns and mechanisms deserve additional scientific attention.
Direct effects of observed diversity, vegetation structure and vegetation health (H3–H6)
Direct and indirect effects of the treatments planted tree diversity and island area on EF were tested by structural equation modeling. Direct effects of observed woody plant diversity, stand structural complexity derived from laser scanning, tree height variability and plant health (as indicated by GNDVI) on EF were also tested.
Ten years after establishment of EFForTS-BEE, we did not find a direct effect of observed woody plant diversity (H3) (Fig. 3) on energy partitioning. Across EFForTS-BEE, a total of 84 woody species were observed, consisting of the six planted tree species complemented by 78 woody species originating from natural recruitment. Despite one plot showing no species (Shannon Index = 0), the interquartile range (0.99–1.87) and a median of 1.45 suggest a relatively uniform distribution of moderate to high woody plant diversity across tree islands, reflecting the presence of a wide range of species throughout EFForTS-BEE. Our finding that observed woody plant diversity enhances EF indirectly through stand structural complexity aligns with diversity – ecosystem function (BEF) theory (Morin et al., 2011), supporting the idea that diversity-driven complementarity – through more efficient 3D – space filling (Jucker et al., 2015; Juchheim et al., 2017) – can strengthen ecosystem functioning even at moderate to high diversity levels.
A significant positive effect of structural complexity on EF was detected ten years after tree island establishment (Fig. 3). When modeling ET in boreal ecosystems, Leonard et al. (2022) integrated forest stand complexity variables into their model and thereby reduced unexplained variance by 10 %. Running their Boreal Ecohydrological Tree Algorithm (BETA+) model (Leonard et al., 2021) with variables describing the stand structure, their ET estimate increased by 26 %. This is consistent with the positive effect size of SSCI on EF in our SEM. Enhanced stand structure may change the microclimate and increase energy exchange.
However, in EFForTS-BEE we did not find a significant effect of the Gini index of tree height on energy partitioning toward latent heat flux. Canopy height is an axis of surface roughness in forests (Maurer et al., 2015). In a study on the biogeophysical effects of deforestation, a decrease of surface roughness led to decreased evapotranspiration in the tropics (Winckler et al., 2019). In eucalypt forests of south-eastern Australia, Meili et al. (2024) attributed fast ET recovery after fire to aerodynamic warming from shorter forest height, highlighting the importance of height in ecohydrology. In urban landscapes, vertical canopy structure was a key predictor of LST (Chen et al., 2020). Tree height variability may not serve as a substitute for surface roughness in the tree islands of EFForTS-BEE, which could explain the lack of a significant effect on EF. As some of the planted trees, as well as the natural regeneration, is below oil palm canopies, tree height variability might not be the dominating factor of surface roughness.
Despite our assumption, vegetation health (GNDVI) did not predict EF. Vegetation indices are often used as calibration factors in ET modeling. For example, the space borne level-3 evapotranspiration (ET_PT-JPL) product uses normalized difference vegetation index (NDVI) and soil adjusted vegetation index (SAVI) to convert potential ET to actual ET (Fisher). Yang and Wang (2011) successfully modeled EF based on spaceborne LST maps using a combined approach of the Priestley-Tylor equation, NDVI and day-night LST differences. Both studies however are addressing larger spatial scales with space borne images, on such larger scales stronger differences in vegetation indices between different land-use types are to be expected. In EFForTS-BEE, the tree islands comprise of dense vegetation at a relatively good health status across all tree islands.
Conclusion
Tree-island area enhances evaporative cooling through both direct and indirect pathways, primarily through a direct area – EF effect. An indirect serial pathway through observed woody plant diversity and structural complexity provides evidence for a secondary mechanism contributing to this relationship. A decrease in EF along the edge gradient found in larger tree islands may be related to the direct tree-island-area effect. Other factors – planted tree diversity, tree height variability, and vegetation health – did not show significant effects on EF, possibly because natural regeneration has become more influential than the initially planted trees at this stage of the experiment. These findings indicate that tree-island area, observed woody plant diversity, and structural complexity are important factors enhancing energy partitioning toward latent heat flux, thereby cooling the environment. Further research is needed to better understand the mechanisms behind the tree-island-area effect on evaporative fraction. Our results suggest that larger and structurally more complex tree islands play a role in regulating microclimate in human-modified landscapes. These findings provide a basis for integrating tree islands into agricultural landscapes to improve evaporative cooling and climate regulation.
For a conceptual structural equation model, relationships and assumed mechanisms please see the Supplement (Fig. S1, Table S2).
Evapotranspiration (ET) and evaporative fraction (EF) for the selected observations (one record per island with the highest radiation during the flight; n=52) against shortwave incoming radiation during the flight. Planted tree diversity is indicated by different symbols and island is given in different colours. 98 % of the observations were obtained at incoming shortwave radiation >850Wm-2. ET increased with incoming shortwave radiation (R2=0.31, p<0.01) whereas EF did not change (R2= 0.01, p=0.48).
Correlation plot for variables used in the structural equation model.
Relationship between evaporative fraction (EF) and predictor variables. Panels show scatter plots and linear regression for (a) planted tree diversity, (b) island area, (c) observed woody plant diversity, (d) structural complexity, (e) tree height variability, and (f) vegetation health (n=52). Regression lines indicate significant relationships (p<0.05).
Code and data availability
The final data set including ET and EF values used for statistical tests were uploaded to Göttingen Research Online database (10.25625/Q7B9TR; Wintz, 2025). Raw thermal and multispectral orthomosaics, QWater raster files and R code used for modeling are available upon request to the corresponding author.
The supplement related to this article is available online at https://doi.org/10.5194/bg-23-7643-2026-supplement.
The contact author has declared that none of the authors has any competing interests.
Disclaimer
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. The authors bear the ultimate responsibility for providing appropriate place names. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.
Acknowledgements
We thank PT. Humusindo for allowing us to work on their land. We further thank our research assistant Tiarma Rezeki Manalu for her support during data collection. This study was conducted using research permits: no. 212/SIP/IV/FR/10/2022 (Thorge Wintz) and no. 100/SIP/IV/FR/2/2023 (Gustavo Brant Paterno). Tree island and oil palm plantation icons used in the graphical abstract were provided by Friederike Arndt, Formenorm Leipzig. ChatGPT was used to generate parts of the R code used for data analysis.
Financial support
This research has been supported by the Deutsche Forschungsgemeinschaft (grant nos. 192626868, 532868192, and 532776526).This open-access publication was funded by the University of Göttingen.
Review statement
This paper was edited by Cornelius Senf and reviewed by Danie lAugusto da Silva and one anonymous referee.
ReferencesArroyo-Rodríguez, V., Fahrig, L., Tabarelli, M., Watling, J. I., Tischendorf, L., Benchimol, M., Cazetta, E., Faria, D., Leal, I. R., Melo, F. P. L., Morante-Filho, J. C., Santos, B. A., Arasa-Gisbert, R., Arce-Peña, N., Cervantes-López, M. J., Cudney-Valenzuela, S., Galán-Acedo, C., San-José, M., Vieira, I. C. G., Slik, J. W. F., Nowakowski, A. J., and Tscharntke, T.: Designing optimal human-modified landscapes for forest biodiversity conservation, Ecol. Lett., 23, 1404–1420, 10.1111/ele.13535, 2020.Baldocchi, D. D.: The role of biodiversity on the evaporation of forests, in: Forest diversity and function: Temperate and boreal systems, edited by: Scherer-Lorenzen, M., Körner, C., and Schulze, E.-D., Springer, Berlin, New York, 131–148, 10.1007/3-540-26599-6_7, 2005.Barbeta, A., Miralles, D. G., Mendiola, L., Gimeno, T. E., Sabaté, S., and Carnicer, J.: Disentangling the Role of Forest Structure and Functional Traits in the Thermal Balance of the Mediterranean–Temperate Ecotone, J. Geophys. Res.-Biogeo., 128, 10.1029/2022JG007264, 2023.Bell, T., Lilley, A. K., Hector, A., Schmid, B., King, L., and Newman, J. A.: A linear model method for biodiversity-ecosystem functioning experiments, Am. Nat., 174, 836–849, 10.1086/647931, 2009.Brenner, C., Zeeman, M., Bernhardt, M., and Schulz, K.: Estimation of evapotranspiration of temperate grassland based on high-resolution thermal and visible range imagery from unmanned aerial systems, Int. J. Remote Sens., 39, 5141–5174, 10.1080/01431161.2018.1471550, 2018.Bulusu, M., Ellsäßer, F., Stiegler, C., Ahongshangbam, J., Marques, I., Hendrayanto, H., Röll, A., and Hölscher, D.: UAV-based thermography reveals spatial and temporal variability of evapotranspiration from a tropical rainforest, Front. For. Glob. Change, 6, 10.3389/ffgc.2023.1232410, 2023.Chao, A., Gotelli, N. J., Hsieh, T. C., Sander, E. L., Ma, K. H., Colwell, R. K., and Ellison, A. M.: Rarefaction and extrapolation with Hill numbers: a framework for sampling and estimation in species diversity studies, Ecol. Monogr., 84, 45–67, 10.1890/13-0133.1, 2014.Chen, J., Jin, S., and Du, P.: Roles of horizontal and vertical tree canopy structure in mitigating daytime and nighttime urban heat island effects, Int. J. Appl. Earth Obs., 89, 102060, 10.1016/j.jag.2020.102060, 2020.Cortés-Molino, Á., Valdés-Uribe, A., Ellsäßer, F., Bulusu, M., Ahongshangbam, J., Hendrayanto, Hölscher, D., and Röll, A.: Combining UAV thermography, point cloud analysis and machine learning for assessing small-scale evapotranspiration patterns in a tropical rainforest, Ecohydrology, 17, 10.1002/eco.2604, 2024.Cregg, B. and Dix, M.: Tree Moisure Stress and Insect Damage in Urban Areas in Relation to Heat Island Effects, AUF, 27, 8–17, 10.48044/jauf.2001.002, 2001.Dantas de Paula, M., Groeneveld, J., and Huth, A.: Tropical forest degradation and recovery in fragmented landscapes – Simulating changes in tree community, forest hydrology and carbon balance, Global Ecology and Conservation, 3, 664–677, 10.1016/j.gecco.2015.03.004, 2015.Dierick, D., Kunert, N., Köhler, M., Schwendenmann, L., and Hölscher, D.: Comparison of tree water use characteristics in reforestation and agroforestry stands across the tropics, in: Tropical Rainforests and Agroforests under Global Change, edited by: Tscharntke, T., Leuschner, C., Veldkamp, E., Faust, H., Guhardja, E., and Bidin, A., Springer Berlin Heidelberg, Berlin, Heidelberg, 293–308, 10.1007/978-3-642-00493-3_13, 2010.Dixon, P.: VEGAN, a package of R functions for community ecology, J. Veg. Sci., 14, 927–930, 10.1111/j.1654-1103.2003.tb02228.x, 2003.DJI: ZENMUSE H20 SERIES: User manual v1.2, https://www.dji.com/de/downloads/products/zenmuse-h20-series#doc (last access: 19 March 2024), 2021.Drakare, S., Lennon, J. J., and Hillebrand, H.: The imprint of the geographical, evolutionary and ecological context on species-area relationships, Ecol. Lett., 9, 215–227, 10.1111/j.1461-0248.2005.00848.x, 2006.Drescher, J., Rembold, K., Allen, K., Beckschäfer, P., Buchori, D., Clough, Y., Faust, H., Fauzi, A. M., Gunawan, D., Hertel, D., Irawan, B., Jaya, I. N. S., Klarner, B., Kleinn, C., Knohl, A., Kotowska, M. M., Krashevska, V., Krishna, V., Leuschner, C., Lorenz, W., Meijide, A., Melati, D., Nomura, M., Pérez-Cruzado, C., Qaim, M., Siregar, I. Z., Steinebach, S., Tjoa, A., Tscharntke, T., Wick, B., Wiegand, K., Kreft, H., and Scheu, S.: Ecological and socio-economic functions across tropical land use systems after rainforest conversion, Philos. T. R. Soc. Lond. B, 371, 10.1098/rstb.2015.0275, 2016.Edwards, F. A., Edwards, D. P., Hamer, K. C., and Davies, R. G.: Impacts of logging and conversion of rainforest to oil palm on the functional diversity of birds in S undaland, Ibis, 155, 313–326, 10.1111/ibi.12027, 2013.Ehbrecht, M., Schall, P., Ammer, C., and Seidel, D.: Quantifying stand structural complexity and its relationship with forest management, tree species diversity and microclimate, Agr. Forest Meteorol., 242, 1–9, 10.1016/j.agrformet.2017.04.012, 2017.Ellsäßer, F., Röll, A., Ahongshangbam, J., Waite, P.-A., Hendrayanto, Schuldt, B., and Hölscher, D.: Predicting Tree Sap Flux and Stomatal Conductance from Drone-Recorded Surface Temperatures in a Mixed Agroforestry System – A Machine Learning Approach, Remote Sens., 12, 4070, 10.3390/rs12244070, 2020a.Ellsäßer, F., Röll, A., Stiegler, C., Hendrayanto, and Hölscher, D.: Introducing QWaterModel, a QGIS plugin for predicting evapotranspiration from land surface temperatures, Environ. Model. Softw., 130, 104739, 10.1016/j.envsoft.2020.104739, 2020b.Ellsäßer, F., Stiegler, C., Röll, A., June, T., Hendrayanto, Knohl, A., and Hölscher, D.: Predicting evapotranspiration from drone-based thermography – a method comparison in a tropical oil palm plantation, Biogeosciences, 18, 861–872, 10.5194/bg-18-861-2021, 2021.Er-Raki, S., Rodriguez, J. C., Garatuza-Payan, J., Watts, C. J., and Chehbouni, A.: Determination of crop evapotranspiration of table grapes in a semi-arid region of Northwest Mexico using multi-spectral vegetation index, Agr. Water Manage., 122, 12–19, 10.1016/j.agwat.2013.02.007, 2013.FAO: World Food and Agriculture – Statistical Yearbook 2023, FAO, Rome, 10.4060/cc8166en, 2023.Fisher, J. B.: Level-3 Evapotranspiration L3 (ET_PT-JPL) Algorithm Theoretical Basis Document, https://ecostress.jpl.nasa.gov/downloads/atbd/ECOSTRESS_L3_ET_PT-JPL_ATBD_20180509.pdf, last access: 12 October 2026.Giambelluca, T. W., Ziegler, A. D., Nullet, M. A., Truong, D. M., and Tran, L. T.: Transpiration in a small tropical forest patch, Agr. Forest Meteorol., 117, 1–22, 10.1016/S0168-1923(03)00041-8, 2003.Gitelson, A. A. and Kaufman, Y. J.: MODIS NDVI Optimization To Fit the AVHRR Data Series – Spectral Considerations, Remote Sens. Environ., 66, 343–350, 10.1016/S0034-4257(98)00065-0, 1998.Gitelson, A. A., Kaufman, Y. J., and Merzlyak, M. N.: Use of a green channel in remote sensing of global vegetation from EOS-MODIS, Remote Sens. Environ., 58, 289–298, 10.1016/S0034-4257(96)00072-7, 1996.Givnish, T. J.: Global conservation priorities for island plant diversity, Nature, 634, 790–792, 10.1038/d41586-024-02921-5, 2024.Gómez, A. M., Parra, A., Pavelsky, T. M., Wise, E., Villegas, J. C., and Meijide, A.: Ecohydrological impacts of oil palm expansion: a systematic review, Environ. Res. Lett., 18, 33005, 10.1088/1748-9326/acbc38, 2023.González-Espinosa, M., María Rey-Benayas, J., Ramírez-Marcial, N., Huston, M. A., and Golicher, D.: Tree diversity in the northern Neotropics: regional patterns in highly diverse Chiapas, Mexico, Ecography, 27, 741–756, 10.1111/j.0906-7590.2004.04103.x, 2004.Gooriah, L. D., Davidar, P., and Chase, J. M.: Species-area relationships in the Andaman and Nicobar Islands emerge because rarer species are disproportionately favored on larger islands, Ecol. Evol., 10, 7551–7559, 10.1002/ece3.6480, 2020.Grace, J.: A `Weight of Evidence' approach to evaluating structural equation models, OE, 5, 10.3897/oneeco.5.e50452, 2020.Grass, I., Kubitza, C., Krishna, V. V., Corre, M. D., Mußhoff, O., Pütz, P., Drescher, J., Rembold, K., Ariyanti, E. S., Barnes, A. D., Brinkmann, N., Brose, U., Brümmer, B., Buchori, D., Daniel, R., Darras, K. F. A., Faust, H., Fehrmann, L., Hein, J., Hennings, N., Hidayat, P., Hölscher, D., Jochum, M., Knohl, A., Kotowska, M. M., Krashevska, V., Kreft, H., Leuschner, C., Lobite, N. J. S., Panjaitan, R., Polle, A., Potapov, A. M., Purnama, E., Qaim, M., Röll, A., Scheu, S., Schneider, D., Tjoa, A., Tscharntke, T., Veldkamp, E., and Wollni, M.: Trade-offs between multifunctionality and profit in tropical smallholder landscapes, Nat. Commun., 11, 1186, 10.1038/s41467-020-15013-5, 2020.Grossman, J. J., Vanhellemont, M., Barsoum, N., Bauhus, J., Bruelheide, H., Castagneyrol, B., Cavender-Bares, J., Eisenhauer, N., Ferlian, O., Gravel, D., Hector, A., Jactel, H., Kreft, H., Mereu, S., Messier, C., Muys, B., Nock, C., Paquette, A., Parker, J., Perring, M. P., Ponette, Q., Reich, P. B., Schuldt, A., Staab, M., Weih, M., Zemp, D. C., Scherer-Lorenzen, M., and Verheyen, K.: Synthesis and future research directions linking tree diversity to growth, survival, and damage in a global network of tree diversity experiments, Environ. Exp. Bot., 152, 68–89, 10.1016/j.envexpbot.2017.12.015, 2018.Guillaume, T., Kotowska, M. M., Hertel, D., Knohl, A., Krashevska, V., Murtilaksono, K., Scheu, S., and Kuzyakov, Y.: Carbon costs and benefits of Indonesian rainforest conversion to plantations, Nat. Commun., 9, 2388, 10.1038/s41467-018-04755-y, 2018.Hernández-Clemente, R., Hornero, A., Mottus, M., Penuelas, J., González-Dugo, V., Jiménez, J. C., Suárez, L., Alonso, L., and Zarco-Tejada, P. J.: Early Diagnosis of Vegetation Health From High-Resolution Hyperspectral and Thermal Imagery: Lessons Learned From Empirical Relationships and Radiative Transfer Modelling, Curr. Forestry Rep., 5, 169–183, 10.1007/s40725-019-00096-1, 2019.Hölscher, D., de Sá, A. T., Bastos, T., Denich, M., and Fölster, H.: Evaporation from young secondary vegetation in eastern Amazonia, J. Hydrol., 193, 293–305, 10.1016/S0022-1694(96)03145-9, 1997.Iddris, N. A.-A., Formaglio, G., Paul, C., Groß, V. von, Chen, G., Angulo-Rubiano, A., Berkelmann, D., Brambach, F., Darras, K. F. A., Krashevska, V., Potapov, A., Wenzel, A., Irawan, B., Damris, M., Daniel, R., Grass, I., Kreft, H., Scheu, S., Tscharntke, T., Tjoa, A., Veldkamp, E., and Corre, M. D.: Mechanical weeding enhances ecosystem multifunctionality and profit in industrial oil palm, Nat. Sustain., 6, 683–695, 10.1038/s41893-023-01076-x, 2023.Juchheim, J., Ammer, C., Schall, P., and Seidel, D.: Canopy space filling rather than conventional measures of structural diversity explains productivity of beech stands, Forest Ecol. Manag., 395, 19–26, 10.1016/j.foreco.2017.03.036, 2017.Jucker, T., Bouriaud, O., and Coomes, D. A.: Crown plasticity enables trees to optimize canopy packing in mixed-species forests, Funct. Ecol., 29, 1078–1086, 10.1111/1365-2435.12428, 2015.Kattenborn, T.: tejakattenborn/dji_h20t_rpeg_to_tif: dji thermal rpeg to tif (v. 0.5), Zenodo [code], 10.5281/zenodo.8315941, 2023.Kikuchi, T., Seidel, D., Ehbrecht, M., Zemp, D. C., Brambach, F., Irawan, B., Sundawati, L., Hölscher, D., Kreft, H., and Paterno, G. B.: Combining planting trees and natural regeneration promotes long-term structural complexity in oil palm landscapes, Forest Ecol. Manag., 569, 122182, 10.1016/j.foreco.2024.122182, 2024.Kotowska, M. M., Leuschner, C., Triadiati, T., Meriem, S., and Hertel, D.: Quantifying above- and belowground biomass carbon loss with forest conversion in tropical lowlands of Sumatra (Indonesia), Glob. Change Biol., 21, 3620–3634, 10.1111/gcb.12979, 2015.Kubitza, C., Krishna, V. V., Alamsyah, Z., and Qaim, M.: The Economics Behind an Ecological Crisis: Livelihood Effects of Oil Palm Expansion in Sumatra, Indonesia, Hum. Ecol., 46, 107–116, 10.1007/s10745-017-9965-7, 2018.Kunert, N., Schwendenmann, L., Potvin, C., and Hölscher, D.: Tree diversity enhances tree transpiration in a Panamanian forest plantation, J. Appl. Ecol., 49, 135–144, 10.1111/j.1365-2664.2011.02065.x, 2012.Kunert, N., Aparecido, L. M. T., Higuchi, N., Santos, J. D., and Trumbore, S.: Higher tree transpiration due to road-associated edge effects in a tropical moist lowland forest, Agr. Forest Meteorol., 213, 183–192, 10.1016/j.agrformet.2015.06.009, 2015.Lawrence, D., Coe, M., Walker, W., Verchot, L., and Vandecar, K.: The Unseen Effects of Deforestation: Biophysical Effects on Climate, Front. For. Glob. Change, 5, 10.3389/ffgc.2022.756115, 2022.Lefcheck, J. S.: piecewiseSEM Piecewise structural equation modelling in r for ecology, evolution, and systematics, Methods Ecol. Evol., 7, 573–579, 10.1111/2041-210X.12512, 2016.Leonard, R., Moore, P., Krause, S., Devito, K. J., Petrone, G. R., Mendoza, C., Waddington, J. M., and Kettridge, N.: The influence of system heterogeneity on peat-surface temperature dynamics, Environ. Res. Lett., 16, 24002, 10.1088/1748-9326/abd4ff, 2021.Leonard, R., Moore, P., Krause, S., Chasmer, L., Devito, K. J., Petrone, R. M., Mendoza, C., Waddington, J. M., and Kettridge, N.: Forest stand complexity controls ecosystem-scale evapotranspiration dynamics: Implications for landscape flux simulations, Hydrol. Process., 36, 10.1002/hyp.14761, 2022.Li, D.: hillR: taxonomic, functional, and phylogenetic diversity and similarity through Hill Numbers, JOSS, 3, 1041, 10.21105/joss.01041, 2018.Li, K., Grass, I., Zemp, D. C., Lorenz, H., Sachsenmaier, L., Nurdiansyah, F., Hölscher, D., Kreft, H., and Tscharntke, T.: Tree identity and canopy openness mediate oil palm biodiversity enrichment effects on insect herbivory and pollination, Ecol. Appl., 33, e2862, 10.1002/eap.2862, 2023.Lüdecke, D., Ben-Shachar, M., Patil, I., Waggoner, P., and Makowski, D.: performance: An R Package for Assessment, Comparison and Testing of Statistical Models, JOSS, 6, 3139, 10.21105/joss.03139, 2021.Ma, Q., Su, Y., Hu, T., Jiang, L., Mi, X., Lin, L., Cao, M., Wang, X., Lin, F., Wang, B., Sun, Z., Wu, J., Ma, K., and Guo, Q.: The coordinated impact of forest internal structural complexity and tree species diversity on forest productivity across forest biomes, Fundamental Research, 4, 1185–1195, 10.1016/j.fmre.2022.10.005, 2024.MacArthur, R. H. and Wilson, E. O.: An Equilibrium Theory of Insular Zoogeography, Evolution, 17, 373, 10.2307/2407089, 1963.Manoli, G., Meijide, A., Huth, N., Knohl, A., Kosugi, Y., Burlando, P., Ghazoul, J., and Fatichi, S.: Ecohydrological changes after tropical forest conversion to oil palm, Environ. Res. Lett., 13, 64035, 10.1088/1748-9326/aac54e, 2018.Marks, C. O., Muller-Landau, H. C., and Tilman, D.: Tree diversity, tree height and environmental harshness in eastern and western North America, Ecol. Lett., 19, 743–751, 10.1111/ele.12608, 2016.Maurer, K. D., Bohrer, G., Kenny, W. T., and Ivanov, V. Y.: Large-eddy simulations of surface roughness parameter sensitivity to canopy-structure characteristics, Biogeosciences, 12, 2533–2548, 10.5194/bg-12-2533-2015, 2015.Meijide, A., Röll, A., Fan, Y., Herbst, M., Niu, F., Tiedemann, F., June, T., Rauf, A., Hölscher, D., and Knohl, A.: Controls of water and energy fluxes in oil palm plantations: Environmental variables and oil palm age, Agr. Forest Meteorol., 239, 71–85, 10.1016/j.agrformet.2017.02.034, 2017.Meijide, A., Badu, C. S., Moyano, F., Tiralla, N., Gunawan, D., and Knohl, A.: Impact of forest conversion to oil palm and rubber plantations on microclimate and the role of the 2015 ENSO event, Agr. Forest Meteorol., 252, 208–219, 10.1016/j.agrformet.2018.01.013, 2018.Meili, N., Beringer, J., Zhao, J., and Fatichi, S.: Aerodynamic effects cause higher forest evapotranspiration and water yield reductions after wildfires in tall forests, Glob. Change Biol., 30, e16995, 10.1111/gcb.16995, 2024.Merten, J., Röll, A., Guillaume, T., Meijide, A., Tarigan, S., Agusta, H., Dislich, C., Dittrich, C., Faust, H., Gunawan, D., Hein, J., Hendrayanto, Knohl, A., Kuzyakov, Y., Wiegand, K., and Hölscher, D.: Water scarcity and oil palm expansion: social views and environmental processes, E&S, 21, 10.5751/ES-08214-210205, 2016.MicaSense: MicaSense RedEdge-MX and DLS 2 Integration Guide, https://support.micasense.com/hc/en-us/articles/360011389334-RedEdge-MX-Integration-Guide (last access: 12 October 2026), 2020.Morin, X., Fahse, L., Scherer-Lorenzen, M., and Bugmann, H.: Tree species richness promotes productivity in temperate forests through strong complementarity between species, Ecol. Lett., 14, 1211–1219, 10.1111/j.1461-0248.2011.01691.x, 2011.Murphy, M. V.: semEff: Automatic Calculation of Effects for Piecewise Structural Equation Models, https://CRAN.R-project.org/package=semEff (last access: 28 October 2026), 2022.Niu, F., Röll, A., Hardanto, A., Meijide, A., Köhler, M., Hendrayanto, and Hölscher, D.: Oil palm water use: calibration of a sap flux method and a field measurement scheme, Tree Physiol., 35, 563–573, 10.1093/treephys/tpv013, 2015.Numata, I., Khand, K., Kjaersgaard, J., Cochrane, M. A., and Silva, S. S.: Forest evapotranspiration dynamics over a fragmented forest landscape under drought in southwestern Amazonia, Agr. Forest Meteorol., 306, 108446, 10.1016/j.agrformet.2021.108446, 2021.Ordway, E. M. and Asner, G. P.: Carbon declines along tropical forest edges correspond to heterogeneous effects on canopy structure and function, P. Natl. Acad. Sci. USA, 117, 7863–7870, 10.1073/pnas.1914420117, 2020.Pallavi, Röll, A., Marques, I., Ramadhani, D. N., Valdes-Uribe, A., Hendrayanto, H., and Hölscher, D.: Changes in leaf area index by tropical forest transformation to plantations increase below-canopy surface temperatures, Global Ecology and Conservation, 53, e03001, 10.1016/j.gecco.2024.e03001, 2024.Paquette, A. and Messier, C.: The effect of biodiversity on tree productivity: from temperate to boreal forests, Global Ecol. Biogeogr., 20, 170–180, 10.1111/j.1466-8238.2010.00592.x, 2011.Park, S., Ryu, D., Fuentes, S., Chung, H., O'Connell, M., and Kim, J.: Mapping Very-High-Resolution Evapotranspiration from Unmanned Aerial Vehicle (UAV) Imagery, IJGI, 10, 211, 10.3390/ijgi10040211, 2021.Paterno, G. B., Brambach, F., Guerrero-Ramírez, N., Zemp, D. C., Cantillo, A. F., Camarretta, N., Moura, C. C. M., Gailing, O., Ballauff, J., Polle, A., Schlund, M., Erasmi, S., Iddris, N. A., Khokthong, W., Sundawati, L., Irawan, B., Hölscher, D., and Kreft, H.: Diverse and larger tree islands promote native tree diversity in oil palm landscapes, Science, 386, 795–802, 10.1126/science.ado1629, 2024.Peterson, R.: Finding Optimal Normalizing Transformations via bestNormalize, The R Journal, 13, 310, 10.32614/RJ-2021-041, 2021.Pinho, B. X., Melo, F. P. L., Braak, C. J. F. T., Bauman, D., Maréchaux, I., Tabarelli, M., Benchimol, M., Arroyo-Rodriguez, V., Santos, B. A., Hawes, J. E., Berenguer, E., Ferreira, J., Silveira, J. M., Peres, C. A., Rocha-Santos, L., Souza, F. C., Gonçalves-Souza, T., Mariano-Neto, E., Faria, D., and Barlow, J.: Winner-loser plant trait replacements in human-modified tropical forests, Nat. Ecol. Evol., 9, 282–295, 10.1038/s41559-024-02592-5, 2024.Potapov, A. M., Drescher, J., Darras, K., Wenzel, A., Janotta, N., Nazarreta, R., Kasmiatun, Laurent, V., Mawan, A., Utari, E. H., Pollierer, M. M., Rembold, K., Widyastuti, R., Buchori, D., Hidayat, P., Turner, E., Grass, I., Westphal, C., Tscharntke, T., and Scheu, S.: Rainforest transformation reallocates energy from green to brown food webs, Nature, 627, 116–122, 10.1038/s41586-024-07083-y, 2024.Qaim, M., Sibhatu, K. T., Siregar, H., and Grass, I.: Environmental, Economic, and Social Consequences of the Oil Palm Boom, Annu. Rev. Resour. Econ., 12, 321–344, 10.1146/annurev-resource-110119-024922, 2020.Ray, T., Delory, B. M., Beugnon, R., Bruelheide, H., Cesarz, S., Eisenhauer, N., Ferlian, O., Quosh, J., Oheimb, G. von, and Fichtner, A.: Tree diversity increases productivity through enhancing structural complexity across mycorrhizal types, Sci. Adv., 9, eadi2362, 10.1126/sciadv.adi2362, 2023.Ray, T., Fichtner, A., Kunz, M., Proß, T., Bradler, P. M., Bruelheide, H., Georgi, L., Haider, S., Hildebrand, M., Potvin, C., Schnabel, F., Trogisch, S., and Oheimb, G. von: Diversity-enhanced canopy space occupation and leaf functional diversity jointly promote overyielding in tropical tree communities, Sci. Total Environ., 951, 175438, 10.1016/j.scitotenv.2024.175438, 2024.Rembold, K., Mangopo, H., Tjitrosoedirdjo, S. S., and Kreft, H.: Plant diversity, forest dependency, and alien plant invasions in tropical agricultural landscapes, Biol. Conserv., 213, 234–242, 10.1016/j.biocon.2017.07.020, 2017.Riveros-Burgos, C., Ortega-Farías, S., Morales-Salinas, L., Fuentes-Peñailillo, F., and Tian, F.: Assessment of the clumped model to estimate olive orchard evapotranspiration using meteorological data and UAV-based thermal infrared imagery, Irrig. Sci., 39, 63–80, 10.1007/s00271-020-00716-w, 2021.Röll, A., Niu, F., Meijide, A., Ahongshangbam, J., Ehbrecht, M., Guillaume, T., Gunawan, D., Hardanto, A., Hendrayanto, Hertel, D., Kotowska, M. M., Kreft, H., Kuzyakov, Y., Leuschner, C., Nomura, M., Polle, A., Rembold, K., Sahner, J., Seidel, D., Zemp, D. C., Knohl, A., and Hölscher, D.: Transpiration on the rebound in lowland Sumatra, Agr. Forest Meteorol., 274, 160–171, 10.1016/j.agrformet.2019.04.017, 2019.Sabajo, C. R., le Maire, G., June, T., Meijide, A., Roupsard, O., and Knohl, A.: Expansion of oil palm and other cash crops causes an increase of the land surface temperature in the Jambi province in Indonesia, Biogeosciences, 14, 4619–4635, 10.5194/bg-14-4619-2017, 2017.Senf, C.: Seeing the System from Above: The Use and Potential of Remote Sensing for Studying Ecosystem Dynamics, Ecosystems, 25, 1719–1737, 10.1007/s10021-022-00777-2, 2022.Team, R. C.: R: A Language and Environment for Statistical Computing, Vienna, Austria, https://www.R-project.org/ (last access: 28 October 2026), 2023.Teuscher, M., Gérard, A., Brose, U., Buchori, D., Clough, Y., Ehbrecht, M., Hölscher, D., Irawan, B., Sundawati, L., Wollni, M., and Kreft, H.: Experimental Biodiversity Enrichment in Oil-Palm-Dominated Landscapes in Indonesia, Front. Plant Sci., 7, 1538, 10.3389/fpls.2016.01538, 2016.Timmermans, W. J., Kustas, W. P., and Andreu, A.: Utility of an Automated Thermal-Based Approach for Monitoring Evapotranspiration, Acta Geophys., 63, 1571–1608, 10.1515/acgeo-2015-0016, 2015.Torresani, M., Rocchini, D., Alberti, A., Moudrý, V., Heym, M., Thouverai, E., Kacic, P., and Tomelleri, E.: LiDAR GEDI derived tree canopy height heterogeneity reveals patterns of biodiversity in forest ecosystems, Ecol. Inform., 76, 102082, 10.1016/j.ecoinf.2023.102082, 2023.van Dijk, A. I. and Keenan, R. J.: Planted forests and water in perspective, Forest Ecol. Manag., 251, 1–9, 10.1016/j.foreco.2007.06.010, 2007.Wang, X., Dallimer, M., Scott, C. E., Shi, W., and Gao, J.: Tree species richness and diversity predicts the magnitude of urban heat island mitigation effects of greenspaces, Sci. Total Environ., 770, 145211, 10.1016/j.scitotenv.2021.145211, 2021.Warton, D. I. and Hui, F. K. C.: The arcsine is asinine: the analysis of proportions in ecology, Ecology, 92, 3–10, 10.1890/10-0340.1, 2011.Wenzel, A., Westphal, C., Ballauff, J., Berkelmann, D., Brambach, F., Buchori, D., Camarretta, N., Corre, M. D., Daniel, R., Darras, K., Erasmi, S., Formaglio, G., Hölscher, D., Iddris, N. A.-A., Irawan, B., Knohl, A., Kotowska, M. M., Krashevska, V., Kreft, H., Mulyani, Y., Mußhoff, O., Paterno, G. B., Polle, A., Potapov, A., Röll, A., Scheu, S., Schlund, M., Schneider, D., Sibhatu, K. T., Stiegler, C., Sundawati, L., Tjoa, A., Tscharntke, T., Veldkamp, E., Waite, P.-A., Wollni, M., Zemp, D. C., and Grass, I.: Balancing economic and ecological functions in smallholder and industrial oil palm plantations, P. Natl. Acad. Sci. USA, 121, e2307220121, 10.1073/pnas.2307220121, 2024.Winckler, J., Reick, C. H., Bright, R. M., and Pongratz, J.: Importance of Surface Roughness for the Local Biogeophysical Effects of Deforestation, J. Geophys. Res.-Atmos., 124, 8605–8618, 10.1029/2018JD030127, 2019.Wintz, T.: Evaporative fraction of tree islands in an oil palm landscape increases with island area, GRO.data, V1 [data set], 10.25625/Q7B9TR, 2025.Yang, J. and Wang, Y.: Estimating evapotranspiration fraction by modeling two-dimensional space of NDVI/albedo and day–night land surface temperature difference: A comparative study, Adv. Water Resour., 34, 512–518, 10.1016/j.advwatres.2011.01.006, 2011.Zeileis, A.: ineq: Measuring Inequality, Concentration, and Poverty, https://CRAN.R-project.org/package=ineq (last access: 28 October 2026), 2014.Zemp, D. C., Ehbrecht, M., Seidel, D., Ammer, C., Craven, D., Erkelenz, J., Irawan, B., Sundawati, L., Hölscher, D., and Kreft, H.: Mixed-species tree plantings enhance structural complexity in oil palm plantations, Agriculture, Ecosystems & Environment, 283, 106564, 10.1016/j.agee.2019.06.003, 2019a.Zemp, D. C., Gérard, A., Hölscher, D., Ammer, C., Irawan, B., Sundawati, L., Teuscher, M., and Kreft, H.: Tree performance in a biodiversity enrichment experiment in an oil palm landscape, J. Appl. Ecol., 56, 2340–2352, 10.1111/1365-2664.13460, 2019b.Zemp, D. C., Guerrero-Ramirez, N., Brambach, F., Darras, K., Grass, I., Potapov, A., Röll, A., Arimond, I., Ballauff, J., Behling, H., Berkelmann, D., Biagioni, S., Buchori, D., Craven, D., Daniel, R., Gailing, O., Ellsäßer, F., Fardiansah, R., Hennings, N., Irawan, B., Khokthong, W., Krashevska, V., Krause, A., Kückes, J., Li, K., Lorenz, H., Maraun, M., Merk, M. S., Moura, C. C. M., Mulyani, Y. A., Paterno, G. B., Pebrianti, H. D., Polle, A., Di Prameswari, A., Sachsenmaier, L., Scheu, S., Schneider, D., Setiajiati, F., Setyaningsih, C. A., Sundawati, L., Tscharntke, T., Wollni, M., Hölscher, D., and Kreft, H.: Tree islands enhance biodiversity and functioning in oil palm landscapes, Nature, 618, 316–321, 10.1038/s41586-023-06086-5, 2023. Zhang, L., Liu, Y., Yin, C., Xue, D., Gui, D., and Qi, Z.: Effects of Farmland Landscape Fragmentation on Agricultural Irrigation in Hotan Oasis, Agriculture, 12, 1503, 10.3390/agriculture12091503, 2022.Zheng, L., Barry, K. E., Guerrero-Ramírez, N. R., Craven, D., Reich, P. B., Verheyen, K., Scherer-Lorenzen, M., Eisenhauer, N., Barsoum, N., Bauhus, J., Bruelheide, H., Cavender-Bares, J., Dolezal, J., Auge, H., Fagundes, M. V., Ferlian, O., Fiedler, S., Forrester, D. I., Ganade, G., Gebauer, T., Haase, J., Hajek, P., Hector, A., Hérault, B., Hölscher, D., Hulvey, K. B., Irawan, B., Jactel, H., Koricheva, J., Kreft, H., Lanta, V., Leps, J., Mereu, S., Messier, C., Montagnini, F., Mörsdorf, M., Müller, S., Muys, B., Nock, C. A., Paquette, A., Parker, W. C., Parker, J. D., Parrotta, J. A., Paterno, G. B., Perring, M. P., Piotto, D., Wayne Polley, H., Ponette, Q., Potvin, C., Quosh, J., Rewald, B., Godbold, D. L., van Ruijven, J., Standish, R. J., Stefanski, A., Sundawati, L., Urgoiti, J., Williams, L. J., Wilsey, B. J., Yang, B., Zhang, L., Zhao, Z., Yang, Y., Sandén, H., Ebeling, A., Schmid, B., Fischer, M., Kotowska, M. M., Palmborg, C., Tilman, D., Yan, E., and Hautier, Y.: Effects of plant diversity on productivity strengthen over time due to trait-dependent shifts in species overyielding, Nat. Commun., 15, 2078, 10.1038/s41467-024-46355-z, 2024.