High-resolution tropical rain-forest canopy climate data

Abstract Canopy habitats challenge researchers with their intrinsically difficult access. The current scarcity of climatic data from forest canopies limits our understanding of the conditions and environmental variability of these diverse and dynamic habitats. We present 307 days of climate records collected between 2019 and 2020 in the tropical rainforest canopy of the Yasuní National Park, Ecuador. We monitored climate with a 10-min temporal resolution in the middle crowns of eight canopy trees. The distance between canopy climate stations ranged from 700 m to 10 km. Apart from air temperature, relative humidity, leaf wetness, and photosynthetically active radiation (PAR), measured in each canopy climate station, global radiation, rainfall, and wind speed were measured in different subsets of them. We processed the eight data series to omit erroneous records resulting from sensor failures or lack of the solar-based power supply. In addition to the eight original data series, we present three derived data series, two aggregating canopy climate for valleys or for ridges (from four stations each), and one overall average (from the eight stations). This last derived data series contains 306 days, while the shortest of the original data series covers 22 days and the longest 296 days. In addition to the data, two open-source tools, developed in RStudio, are presented that facilitate data visualization (a dashboard) and data exploration (a filtering app) of the original and aggregated records.


Introduction
Forest canopy climate is important for the performance of trees (Luo, 2007;O'Grady et al., 2011) as well as for sessile and mobile organisms living in the dynamic canopy habitat (Suggitt et al., 2017). Forest canopies host a wealth of biological diversity (Barker and Pinard, 2001) and canopy climate factors such as temperature, relative humidity, and light intensity drive and constrain the physiological performance, ecology, and fitness of sporadic and permanent inhabitants (Suggitt et al., 2017). In addition to hosting sessile organisms, such as plants and lichens (i.e., epiphytes), canopies host mobile organisms ranging in size from microorganisms and arthropods to all groups of vertebrates (Nakamura et al., 2017). This astonishing diversity is a common descriptor for canopies in mature tropical forests, yet canopy climate records describing the conditions met by these diverse organisms are particularly scarce, despite the expansion of canopy research at tropical latitudes in recent decades (Nakamura et al., 2017). Here, we present >300 days of canopy climate records with high temporal resolution (each 10 min) obtained in the lowland tropical forest surrounding the Yasuní scientific station in Amazonian Ecuador, an iconic biodiversity hotspot (Bass et al., 2010).
The intrinsic difficulty to access forest canopies challenges us while studying canopy climate in forests around the world. Yet, the relevance of canopies for biodiversity maintenance calls for persistence, creativity, and cooperation (Barker and Pinard, 2001). In addition, forest canopies contribute importantly to vegetation-climate feedback (Lin et al., 2010;O'Grady et al., 2011). The exchange of relevant gases between canopies and the atmosphere can be highly dynamic, in dependence of climatic fluctuations above and inside the canopy (Luo, 2007;Lin et al., 2010). However, high-temporal-resolution field measurements of canopy climate remain scarce and scattered, limiting our ability to understand such dynamics (Nakamura et al., 2017). This shortage is especially pronounced in tropical forests. Although a modeling approach can supply estimates for the canopy climate of forests in different parts of the world (Maclean and Klinges, 2021), the robustness and accuracy of those estimates are hard to evaluate in forest types with no or insufficient available field data. Hence, monitoring canopy climate is critical to improving and validating such climate models. In turn, understanding climatic fluctuations in forest canopies is essential to understand and model vegetation-atmosphere interactions, as well as the functional ecology of trees and the incredible diversity of canopy-dwelling organisms found, in particular, in tropical rainforests.
Canopy climate parameters can be estimated from macroclimate with mechanistic models (Maclean, 2020), but akin to other models, calibration and validations demand field measurements . Recently, a decided leadership and the contribution of a wealth of field measurements resulted in a body of global estimates of high temporal and spatial resolution microclimate for soils and near the soil surface (Lembrechts et al., 2021). This integrative research highlighted the divergence between soil microclimate and macroclimate, summarized consistent patterns across biomes, and incentivized using microclimate to tackle ecological research in the frame imposed by global change. For the canopy, a recent mechanistic model can estimate climate conditions from a climate input and from a series of canopy descriptors (Maclean and Klinges, 2021), however remote and under-described forests configure challenging applications of such models. While extrapolated and modeled macroclimate estimates have supported a generation of studies assessing the effects of climate on organisms and ecosystems (Suggitt et al., 2017), the divergence between macroclimate and climate of specific habitats  can obscure the efforts to interpret the impact of climate change on organismal responses (Suggitt et al., 2017;. Therefore, high-resolution microclimate records constitute a relevant input to unveil ecological nuances of the habitat where they were gathered. Tropical regions host vast forested areas, like the Amazon, yet detailed canopy climate records for these forests remain scattered (e.g., Löbs et al., 2020). Here, we present a set of climate measurements with high temporal resolution obtained in the canopy of the forest in Yasuní National Park, Ecuador, a global biodiversity hotspot (Bass et al., 2010). Data were collected within the crowns of eight canopy trees, in the Johansson (1974) zone corresponding to the middle canopy. We monitored temperature, relative humidity, photosynthetically active radiation (PAR), and leaf wetness, and we derived the vapor pressure deficit (VPD). In selected crowns, we additionally registered precipitation. Solar radiation, and wind speed and direction. Despite several data gaps due to the harsh conditions for electronics in the forest coupled with logistical challenges, we compiled canopy climate series with more than 300 days of data.
We anticipate that these data series will enrich the view and methodological possibilities of ecologists studying the wealth of organisms dwelling in the canopy of this and other tropical lowland forests. Even if the structure of this canopy remains relatively stable despite recent climate warming trends (Nabe-Nielsen and Valencia, 2020), the biology, ecology, and phenology of its inhabitants may lead to new ecological dynamics and emergent properties (Pincebourde et al., 2016). We encourage using these data series to assess organismal responses of specific organism groups to short-term canopy climate. It is likely that a modeling approach can assess how relevant are fine temporal resolution climate patterns for organismal responses (Lembrechts et al., 2019).
The section below corresponds to the Metadata of our climate data series. Given the potential use of these data in ecological studies, the metadata follows the standard descriptors suggested by Michener et al. (1997), excluding non-applicable fields to avoid redundancies while maintaining the suggested numbering system.

Class I. Data set descriptors
A. Data set identity: High-resolution tropical rain-forest canopy climate data B. Data set identification codes: Suggested Data set Identify Codes. "S1R_Canopy_Climate_2019-20.csv," "S1V_Canopy_Climate_2019-20.csv," "S2R_Canopy_Climate_2019-20.csv," "S2V_Canopy_Climate_2019-20.csv," "S3R_Canopy_Climate_2019-20.csv," "S3V_Canopy_Climate_2019-20.csv," "S4R_Canopy_Climate_2019-20.csv," "S4V_Canopy_Climate_2019-20.csv," "Ridges_Canopy_Climate.csv," "Valleys_Canopy_Climate.csv," and "Yasuni_Canopy_Climate. csv" (DataS1  Figure 1). The YSS is located in the Orellana province (0⁰40 0 27 00 S 76⁰23 0 50 00 W, 230 m a.s.l.,~90 ha) to the South of the Tiputini river. c. Habitat: Medium crown of eight canopy trees (Table 1). d. Geology, landform: Modestly undulating terrain, where local valleys and ridges differ by less than 100 m of elevation, with poor and clayey soils originated from weathering of dominant materials of the intersection between two geological shields, Andes and Brazilian (stratified clays and sediments of the Curaray formation from the tertiary; Tschopp, 1953). Figure 1. Study area and stratified study design used to monitor climate with a 10-min temporal resolution in the tropical rainforest canopy of the Yasuní National Park, Ecuador. Large circles represent study sites located at different distances from the Tiputini river (black indicates the largest distance and the lightest grey indicates the shorter distance). Two canopy climate stations were established in each site, one in a valley (blue dot) and the other in a ridge (orange dot) as highlighted by colour-keyed altitudinal belts.  (Holdridge, 1947(Holdridge, , 1964 with a wet equatorial climate with imperceptible seasonality (Bailey, 2014), a mean annual temperature of~25⁰C and an average annual precipitation of~2,240 mm. 2. Experimental or sampling design a. Design characteristics: To monitor canopy climate, eight canopy climate stations were established in a stratified sampling design targeting ridges (four stations) and valleys (four stations). Nearby valley and ridge trees correspond to a site (Table 1). Air temperature, relative humidity, PAR, and leaf wetness were recorded by all canopy climate stations, while precipitation, wind, and global radiation were recorded by selected canopy climate stations (Table 1). b. Data collection period, frequency, etc.: All canopy climate stations measured climate parameters every 30 s and recorded data every 10 min (mean, minimum, maximum, or sampling values of the 20 measurements) between April 2019 and February 2020. However, the different stations and sensors have gaps within this period ( Figure 2 and Table 2). 3. Research methods a. Field/laboratory: We installed eight climate stations, each in the crown of a tree, using adapted climbing techniques (Perry 1978) in the second half of April 2019. These trees belonged to seven species distributed in six botanical families and averaged 64 AE 25.6 cm (mean AE SD) for DBH, 26 AE 2.7 m of height, and 7.2 AE 1.5 m of crown radius (Table 1); we targeted canopy trees, that is, those immersed in the canopy stratum of the forestexcluding sub-canopy and emergent trees. In each tree, the sensor set ( Figure 3) was established in the medium section and upper side of crown branches (i.e., in the middle canopy sensu Johansson 1974). We verified the performance of the climate stations in the lab before installation, and from the ground via WIFI after installation, using the interface provided by the datalogger manufacturer (Device Configuration Utility, 2.21.16 by Campbell scientific). Verifications and data downloading from the ground were performed 1 week, 4 months, and 10 months after installation. After the last verification, we removed the climate stations.    . Set of sensors used to monitor climate with a 10-min temporal resolution in the tropical rainforest canopy of Yasuní National Park, Ecuador. In each station, a radiation shield (a) protected the sensor for temperature and relative humidity from direct sunshine and rain. The leaf wetness sensor (b) and the sensor for photosynthetic active radiation (c) were placed on the upper side of the medium section of a crown branch. Selected canopy climate stations were instrumented with a generic rain gauge (d), a sensor for global radiation (e), or a digital anemometer (f), as indicated in Table 1. Instrument details are provided in Table 3. August 2020 4. Data verification: Data were checked in by the authors. In each data series, a variable of each climate parameter was plotted to identify gaps and suspicious data in the records, suggesting partial or definitive sensor failure. This data exploration was supported by field notes taken at the time of  1664,1666,1772,1773,1774,1775,1776,1777 Tipping bucket rain gauge (Figure 3d) Record total precipitation during the recording period (10 min) Kalyx 190847, 190848, 190849, 190850, 190851, 190852 Digital thermopile pyranometer (Figure 3e) Record average solar radiation, average dew point, and a second measurement for average air temperature and its standard deviation Campbell Scientific CS320 1835, 1836

ELM
Two-dimensional ultrasonic wind sensor ( Figure 3f) Record a sample of wind direction, as well as average, maximum, and minimum wind speed The Gill WindSonic4 18480149, 18500164, 18500165, 19480151 removing the stations (e.g., "rain gauge clogged with canopy debris," "rain gauge broken; the rain collector is absent," "climate station box colonized by termites," "logger under water within the climate station box"). To ease comparisons among stations and sites, data series were aligned, and suspicious data were labeled as NA (R script Data_ Compilation.Rmd). By the end of the data verification process, the data series contained only trustable data (Figure 4).  There is no restriction for using data from this data paper, as long as the data paper is cited as the source of the information used. C. Costs: None. Data size details are presented in Table 4. 3. Format and storage mode:

Header information:
Data series of stations in the same site (e.g., S1V and S1R) share headers (See Table 4 in Section B, Variable information) 5. Alphanumeric attributes: Mixed (See Table 4 in Section B, Variable information).

B. Variable information
The details of the climatic variables included in the data are summarized in Table 5.
Environmental Data Science e13-11 C. Data anomalies: In the original data series, absent data were filled with NA either by the data logger or during the data verification process, where suspicious data points were replaced by NA. In the derived data series, absent data were filled with NA during the data compilation process.

Class V. Supplemental descriptors
A. Data acquisition Automated data loggers. B. Quality assurance/quality control procedures: The original data series went through a data verification process while the derived data series resulted from a data compilation process. The data verification, including the calculation of derived variables, consisted of four steps: (a) visualization and diagnosis of suspicious data, that is, those exceeding the possible values for the climate parameter in the locality or those resulting from sensor failure (for instance, PAR that was above zero during night hours), resulting from malfunctioning of specific sensors; (b) replacement of suspicious data with NA; (c) calculation of derived climate variables (Table 5), average relative humidity as the mean value of the recorded relative humidity variables (maximum and minimum), VPD derived from average relative humidity and average air temperature following Fenton and Frego (2005), calculated dew point (DewPtC) derived from average air temperature and average relative humidity following Lawrence (2005); and transformation of solar radiation data, recorded in kWm 2 , to W/m 2 , because this is the most commonly used unit; and (d) creation of fields used to filter data, that is, Date, Hour, and DateHour (Table 4), by using functions of R base (trunc; R Core Team, 2020), and   "lubridate"(hour; Grolemund and Wickham, 2011) and "chron" (as.chron; James and Hornik, 2020) R packages. The data compilation consisted of three steps: (a) temporally aligning series for each climate parameter; (b) assigning a value for each climate parameter to each timestep, by either copying the value recorded by a single station or calculating a mean value when two or more stations registered that parameter; absent data were filled with NA; and (c) writing the aggregated data series as an output (in format csv file). C. Related materials: None D. Computer programs and data-processing algorithms: Data were downloaded using the logger software (Campbell Scientific), while data verification and processing were performed using R (see details in literal B, Quality assurance/quality control procedures, above and in the R scripts: Data_Compilation.Rmd, Dashboard_CanopyClimate. Rmd, and Filtering_CanopyClimate.Rmd). E. Archiving 1. Archival procedures: Description of how data are archived for long-term storage and access 2. Redundant archival sites: Locations and procedures followed Figure 5. Appearance of a box station (S2R) used to monitor canopy climate with a 10-min temporal resolution in the tropical rainforest canopy of the Yasuní National Park, Ecuador. Tables 2 and 4 detail equipment and station components.
F. Publications and results: None. G. History of data set usage 1. Data request history: None. 2. Data set update history: First update in August 2019, second update in February 2020. 3. Review history: None. 4. Questions and comments from secondary users: None.

Conclusion
The main practical lesson from this data collection and compilation is that sustaining cooperation with local researchers is critical to strengthen our capabilities toward understanding ecological processes occurring in tropical canopies, as likewise suggested by Haelewaters et al. (2021). This lesson highlights the need for budgeting qualified partners to ensure proper equipment maintenance. Then, relocating a part of the available budget toward qualified colleagues rather than to equipment would likely result in a more complete and clean raw data series. We observed that equipment becomes less reliable as forest growth occurs in or around critical components of sensors and electronics, including the power supply.
Abbreviations PAR photosynthetically active radiation VPD vapor pressure deficit