Dataset on microclimate and drone-based thermal patterns within an oil palm agroforestry system

Microclimate and Land Surface Temperature (LST) are important analytical variables used to understand complex oil palm agroforestry systems and their effects on biodiversity and ecosystem functions. In order to examine experimental effects of tree species richness (0, 1, 2, 3 or 6), plot size (25 m2, 100 m2, 400 m2, 1600 m2) and stand structural complexity on microclimate and Land Surface Temperature, related data were collected following a strict design. The experiment was carried out in the Jambi province, in Sumatra (Indonesia), as part of the collaborative project EFForTS [Ecological and Socioeconomic Functions of Tropical Lowland Rainforest Transformation Systems]. Microclimate data collected using miniaturized data loggers combined with drone-based thermal data were considered within an oil palm plantation enriched with six target tree species. The timeframe considered for data analysis was 20th September 2017 to 26th September 2017. The experiment data can be used for comparison with data from conventional oil palm agroforestry systems in the tropics. They can more specifically be used as reference to assess microclimate and Land Surface Temperature patterns within similar agroforestry systems.


Keywords:
Microclimate Land surface temperature iButton Drone-based thermography Oil palm Agroforestry Vegetation structure Biodiversity experiment was carried out in the Jambi province, in Sumatra (Indonesia), as part of the collaborative project EFForTS [Ecological and Socioeconomic Functions of Tropical Lowland Rainforest Transformation Systems]. Microclimate data collected using miniaturized data loggers combined with dronebased thermal data were considered within an oil palm plantation enriched with six target tree species. The timeframe considered for data analysis was 20th September 2017 to 26th September 2017. The experiment data can be used for comparison with data from conventional oil palm agroforestry systems in the tropics. They can more specifically be used as reference to assess microclimate and Land Surface Temperature patterns within similar agroforestry systems.  Table   Subject Ecology Specific subject area Assessment of microclimate and land surface temperature variability within oil palm plots enriched with tree species. Type of data Table  Image Figure How data were acquired Data were obtained using miniaturized microclimate sensors (hygrochron and thermochron loggers), and an octocopter drone equipped with radiometric thermal and RGB (red-green-blue) cameras. Data format Raw Analyzed Parameters for data collection Ambient air temperature, soil temperature, relative humidity and land surface temperature data acquired over a week (20 th September 2017 to 26 th September 2017) and around noon were considered. Description of data collection Microclimate data were collected from 28 experimental plots and 4 control plots varying in size (25 m 2 , 100 m 2 , 400 m 2 and 1600 m 2 ) and in diversity level (0, 1, 2, 3 and 6

Value of the Data
• This dataset provides valuable spatial and temporal temperature and humidity data obtained from permanent plots established in an agroforest stand. It might be useful to understand the effect of mixed species tree planting on microclimate and land surface temperature variability. • These data will benefit those who are interested in agroforestry management systems, biodiversity and ecosystem functions enhancement and efficient techniques for microclimate data collection and assessment. • The data can be used to be compared with other microclimate and land surface temperature data collected in tropical agroforestry systems or pure oil palm plantations. They can also be used to assess effect of other parameters (e.g., vapour pressure deficit, tree growth, transpiration, etc.) on thermal patterns.

Microclimate data
Ambient air temperature, soil temperature and relative humidity represent the microclimate data of interest recorded. The dataset containing microclimate data is structured with text files categorized with date, Time, values (temperatures/relative humidity values) and Unit . These are raw data extracted from microclimate sensors.

Land surface temperature data
Land Surface Temperatures were measured at the surface of oil palm and tree canopy. Thermal images from 56 plots and a csv file summarizing climatic variables for these plots and their metrics (average, minimum, maximum etc.) are available in the data repository. Thermal images were stored as tif images that can be visualized in GIS software (e.g. QGIS) and their respective properties can also be checked there. Abbreviations used in the csv file named "All-Data_Humusindo_56Plots_stats.csv " are summarized ( Table. 1 );

Other abbreviations used in the dataset
RH: relative humidity; LST: land surface temperature

Appendices
Raw data available in the Mendeley repository include a first folder containing appendices summarized in a docx file. Appendix A contains a detailed enumeration and denomination of microclimate sensors positioned above and under the ground, in the respective plots within which they were found. It shows which sensors successfully collected data and which didn't. Appendix B display metrics (mean, maximum, minimum, median and standard deviation) of microclimate (for 32 plots) and land surface temperature (for 56 plots) calculated considering the timeframe 10 am to 3 pm. Appendix C summarizes the proportion of microclimate sensors with available data for each investigated plot. Appendix D makes a state of all plots and categories of plot size and diversity level within which they are positioned. Appendices E and F show calculated values of mean, median, standard error of the mean, minimum/maximum values of ambient air,  To support these statistics, appendix J presents an illustrative figure of the distance effect on microclimate variables. Appendix H summarizes statistics explaining the effect of distance variation from plot centre on microclimatic values. Appendix I contains two figures showing microclimate patterns over the period of data acquisition, for microclimate variables ambient temperature, relative humidity and soil temperature. Appendix K illustrates the specific design of miniaturized microclimate sensors disposition within varying-sized plots. Appendix L presents an illustrative image of field installations (sensors and hand-made protection shields).

Experimental Design, Materials and Methods
Microclimate data were obtained from 28 plots systematically varying in size and tree species diversity, while land surface temperature data were collected from 52 plots. Microclimate data represent environmental variables recorded above (ambient air temperature and relative humidity) and below (soil temperature) the ground. Four additional control plots differently considered for microclimate and land surface temperatures data were also investigated. 198 miniaturized microclimate sensors (100 thermochron and 98 hygrochron iButtons, Maxim integrated, USA) to measure ambient air temperature, relative humidity, and soil temperature, which were acquired in 10-minutes intervals all over considered days. Due to a limited number of available sensors, we selected from the 56 initial plots, 32 that represented a great variability in vegetation structure and with varying species diversity level and plot size, including the four control plots.
Within each plot, mini microclimate sensors were positioned on each sampling point located at increasing distance on a logarithmic scale (1, 2, 4 and 8 m distant from each other) oriented along three main directions: North, South-East, and South-West. The purpose of this fractal design was to account for spatial variations in temperature and humidity values [1] and to have comparable data across plot sizes. We used two types of mini microclimate sensors: hygrochron temperature/humidity loggers (DS1923-F5#), installed 1.5 m above the ground to measure both the ambient air temperature and relative humidity and thermochron temperature loggers (DS1922L-F5#), placed 10 cm under the ground to measure soil temperature. The mini microclimate sensors were protected from water and direct solar radiations using hand-made multi-plate radiation shields ( Fig. 1 ).
Precision and accuracy as provided by the manufacturer are 0.063 °C and 0.5 °C, respectively for hygrochron and thermochron loggers. When measured values were negative, these were considered as aberrant and therefore excluded from the analysis to prevent any bias. We validated the mini microclimate sensors by applying a linear regression of the measured values with reference values (air temperature and relative humidity measured with Hygro-Thermo Transmitter, Thies Clima, Göttingen, Germany; soil temperature measured with Trime-Pico32, IMKO, Ettingen, Germany) across a range of controlled microclimatic conditions (from 25 to 35 °C), indicating no systematic biases (mean slope = 0.86, mean intercept = 5.68 °C, mean R 2 = 0.91 for temperature, mean slope = 0.96, mean intercept = 3.56%, mean R 2 = 0.91 for relative humidity). Collected data were saved as txt. files.
Thermal images were acquired with an octocopter drone (MK EASY Okto V3; HiSystems, Moormerland, Germany) equipped with radiometric thermal and RGB (red-green-blue) cameras, to capture thermal and RGB images of all the 56 study plots. We used the thermal camera Flir Tau 2 640 with TeAx ThermoCapture module attached; the focal length 13 mm covers spectral bands ranging from 7.5 to 13.5 μm. RGB camera was based on an Omni vision OV12890 CMOS-Sensor 148 (Omni vision, USA) with a 170 °FOV fish-eye lens [2] . For each day considered, flights were operated at noon (12 pm local time) at the average height of 50 m above the starting point, but varying up to 20 m over plots. Thermal images were recorded and only images falling within plot dimensions were cropped and saved as rgb tif files. The resulting plot thermal images were used to compute minimum, maximum and average values of temperature in degree Kelvin per plot, using thermal pixels. Thermal values per plot were saved as csv file.
We derived the stand structural complexity from terrestrial laser scans in October and November 2016 based on a procedure described by Ehbrecht et al. (2021) [3] . A FARO Focus terrestrial laser scanner (Faro Technologies Inc., Lake Mary, USA), placed at the centre of each plot was used to obtain three-dimensional point clouds of each plot for the computation of the Stand Structural Complexity Index (SSCI). SSCI is an integrated measure of the three-dimensional arrangement of the vegetation above the herbaceous layer and quantifies the heterogeneity of biomass distribution in three-dimensional space [4] . Index values increase with increasing efficiency of canopy space occupation and vertical stratification. Further details on SSCI construction and functioning can be found in Ehbrecht et al. (2021). Control plots considered to assess stand structural complexity, Land Surface Temperature (LST) and microclimate were all different, and correlations between the three variable categories were therefore performed considering 28 plots instead of 32.
We used R version 3.6.3 to calculate different metrics (mean, median, maximum/minimum, standard error of the mean and range) of each variable (ambient air temperature, relative humidity, soil temperature and LST). Microclimate metrics calculation took into consideration microclimate sensors at equivalent distance away from central sensor (1 meters) for all plots, so as to cancel the distance effect.

Ethics Statement
None

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships which have or could be perceived to have influenced the work reported in this article.