Microclimate proxy measurements from a logging gradient in Malaysian Borneo (BALI project)

Description: Overview We focused our empirical measurements in Sabah, Borneo. We measured a microclimate thermal proxy sensitive to radiative, convective, and conductive heat fluxes. This thermal proxy is relevant to a range of organismal processes including plant regeneration, animal behavior, and soil nutrient fluxes. Measurements were made across space and time within three 1-ha plots comprising a gradient from old growth to heavy selective logging. We then combined these data with nearby weather station air temperature data, as well as measurements of topography and canopy structure derived from detailed ground surveys and airborne LIDAR. Sampling design Study location We identified three sites along a logging intensity gradient in the Malaysian state of Sabah in Borneo. Sites each contain a 1-ha plot in lowland mixed dipterocarp forest. The plots are chosen to contrast an unlogged old growth forest with a moderately logged forest and a heavily logged forest Microclimate To map the microclimate thermal proxy in each plot, we installed dataloggers on a semi-regular grid pattern varying in minimum distance from 1 to 14 meters. Each datalogger was a Thermochron iButton (DS1921G, Maxim), capable of logging up to 2048 temperature values between -30°C and 70°. Each datalogger was waterproofed by wrapping in plastic paraffin film (Parafilm, Bemis) and then in light yellow duct tape. Dataloggers were attached using plastic zip-ties onto PVC stakes at a height of 1-3 cm immediately above the forest floor. The tape color was chosen to approximate the albedo of vegetation/soil, and the size of the sensor package was chosen to have a similar boundary layer to many small organisms (e.g. tree seedlings, fallen branches, large insects). The sensor packages intentionally did not include a radiation shield, as the intent was not to measure air temperature. Temperatures recorded by the loggers therefore reflect a combination of conductive, convective, and radiative heat fluxes, and can be considered a rough proxy for those experienced by small organisms. Each plot contains 25 20 × 20 m2 subplots, demarcated by 1 m-high PVC stakes embedded in the soil. Each subplot also contains at its center a mesh litter trap suspended on PVC stakes at 1-m height. Exact locations for all subplot corner and center stakes were determined using ground-based Field-Map software (IFER, Jílové u Prahy, Czech Republic). Spatial positions were recorded in three-dimensional space (local x, y, z-coordinates) using an Impulse 200 Standard laser rangefinder, MapStar Module II electronic compass (Laser Technology, Colorado, USA). We installed a datalogger on these stakes at the corner of each subplot and the center of each subplot. We also chose at random three focal subplots in each plot for higher-resolution sampling. Within these subplots we established a cross-type design, with six additional dataloggers deployed at 1 to 5 m distance on additional PVC stakes located near each litter trap. A total of 239 dataloggers were installed. Dataloggers were deployed during the end of the dry season in late 2015. Each recorded 28 days of data at 20-minute intervals. Start times were synchronized among dataloggers within plots. The exact date of deployment was 1 November for the heavily logged plot and 9 November for the moderately logged and old growth plot. Weather conditions during November-December 2015 were consistently dry and hot, so we do not anticipate any biases from the differing start dates. In nearly all cases dataloggers were recovered in their original location, except for a small number that were transported 1-2 meters down slopes. We treated data from these as though they were in their original position. A small number of dataloggers also failed due to being lost or punctured by animal bites. 90% of dataloggers (214/239) were successfully recovered and downloaded. Air temperature data To compare the microclimate thermal proxy to other temperature metrics, we obtained off-plot (open site) and on-plot (below canopy) weather station data. To represent off-plot data for both the moderately and heavily logged plots, a weather station was located in a cleared area at the SAFE base camp (4.724341°N, 117.601449°E), at a distance of 2.0 km from the heavily logged plot and 3.9 km from the moderately logged plot. Data were logged continuously (Datahog, Skye Instruments, UK). Measurements included air temperature (°C) and photosynthetically active radiation (W m-2). Data were available for all of the study period. To represent off-plot data for the old growth plot, another weather station was located in a cleared area at the Maliau Basin Studies Center (4.736263°N, 116.97662°E), at a distance of 1.4 km from the plot. Available data only included photosynthetically active radiation (W m-2). Data were available for approximately 25% of the study period. We predicted air temperature values at this plot for these dates by calibrating a LOESS regression model of air temperature based on time of day (seconds after midnight) and photosynthetically active radiation, calibrated with weather station data from an open clearing at the SAFE base camp (78 km distance, 184 m lower elevation). Because of the small elevation change we did not include a further lapse rate correction for temperature. The fitted model, which had a residual standard error of 0.9°C, was used to predict off-plot air temperature at the old growth plot. To represent on-plot air temperature, we located air temperature sensors (HOBO, U23-002) within radiation shields at 1.5 m height in a subplot within each plot (corresponding to a focal subplot with a higher density of microclimate dataloggers: old growth, subplot 18; moderately logged, subplot 24; heavily logged, subplot 25). Temperature was measured hourly. LiDAR data Discrete airborne LiDAR data were acquired by NERC's Airborne Research Facility (ARF) in November of 2014 using a Leica ALS50-II LiDAR sensor flown on a Dornier 228-20 at 41 points m-2 density, with up to four returns recorded per pulse. Georeferencing of the point cloud was ensured by incorporating data from a Leica base station in the study area. LiDAR point clouds were classified into ground and non-ground points, and used to produce a 1 m resolution canopy height model by averaging the first returns. Gaps in the canopy height model were filled by averaging neighboring cells. Topography The ground-mapped coordinates of the subplot corners, subplot centers, and all stems were used to construct a digital elevation model (DEM) for the plot. Elevation was interpolated onto a 1 m grid using ordinary kriging with a minimum of 4 points and search radius of 30 meters. This grid was then aligned to the LIDAR-determined location and elevation of the plot corners. The DEM was then used to estimate slope (in degrees) and cosine of aspect (with higher values indicating more southerly exposures) for each location. Forest structure Forest structure was determined from field surveys and from airborne laser scanning. For the field survey, all trees ≥10 cm diameter at 1.3 m height were censused in each plot in 2016. Diameter was measured at 1.3 m with a tape measure, height with laser rangefinders, and x-y position of each stem were determined using the same system as the subplot corners. The horizontal crown projection of every tree was mapped by measuring spatial positions (x and y-coordinates) of 5 to 30 points (depending on the size of the crown) at the boundary of a crown projected to the horizontal plane and then smoothed using Field-Map software. Field stem maps were then converted into raster grids of stem basal area density (smoothed with 2-meter Gaussian kernel, and then rasterized to 1 m resolution), canopy density (number of overlying canopies per unit area) (1 m resolution), and plant area index (PAI) (10 m native resolution, interpolated to 1 m resolution). Spatial variation in PAI was mapped from the LiDAR point cloud using the MacArthur-Horn method. The method assumes that the leaves are randomly distributed within the laterally homogeneous canopy layers, so the PAI is proportional to the logarithm of the fraction of LiDAR pulses, β, penetrating through the canopy: PAI = -1/κ ln(β), where κ is a correction factor that accounts for canopy features, such as clumping and the distribution of leaf angles. We assumed a constant value of κ=0.7. Only the first returns, representing the first interaction of each LiDAR pulse with the canopy, are considered. We employed a lower cutoff of 2 m to avoid confusing ground returns with low-lying vegetation. PAI was estimated for point locations along a 1 m regular grid using circular sampling neighborhood of 10 m. This sampling window size is used to capture a sufficient number of LiDAR returns to avoid saturation effects in the more densely vegetated parts of the plots. This approach for calculating canopy closure may be biased, as clumping of vegetation, variation in leaf angle, and canopy edges (i.e. at gaps) should lead to spatial variation in the κ coefficient. It was not possible with our data to constrain κ using hemispherical photos due to saturation effects. Project: This dataset was collected as part of the following SAFE research project: Drivers of microclimate variation in disturbed forests Funding: These data were collected as part of research funded by: NERC independent research fellowship (Standard grant, NE/M019160/1) This dataset is released under the CC-BY 4.0 licence, requiring that you cite the dataset in any outputs, but has the additional condition that you acknowledge the contribution of these funders in any outputs. Permits: These data were collected under permit from the following authorities: Sabah Biodiversity Council (Research licence JKM/MBS.1000/2/2(JLD.3((126) ) XML metadata: GEMINI compliant metadata for this dataset is available here Files: This dataset consists of 2 files: Blonder_Mircoclimate.xlsx, Mircoclimate_proxy.zip Blonder_Mircoclimate.xlsx This file contains dataset metadata and 2 data tables: Mircoclimate datalogger data (described in worksheet Data_Sheet_S1) Description: Raw temperature measurements for all microclimate dataloggers Number of fields: 7 Number of data rows: 444416 Fields: Site: The name of the site, corresponding to logging treatment intensity (Field type: Categorical) Tag: The field code for the data logger (Field type: ID) Time.elapsed..s.: Number of seconds elapsed since data logger began recording (Field type: Numeric) Forest.floor.temperature..degC.: Microclimate proxy reading (Field type: Numeric) X..m.: Location of the data logger in meters east in UTM zone 50N (Field type: Numeric) Y..m.: Location of the data logger in meters north in UTM zone 50N (Field type: Numeric) Z..m.: Location of the data logger in meters above sea level (Field type: Numeric) Weather station data (described in worksheet Data_Sheet_S3) Description: Raw off-plot and on-plot weather station air temperature data Number of fields: 4 Number of data rows: 1397 Fields: Site: The name of the site, corresponding to logging treatment intensity (Field type: Categorical) Time.elapsed..s.: Number of seconds elapsed since data logger began recording (Field type: Numeric) Off.plot.weather.station.air.temperature..degC.: Air temperature measurement of the off-plot weather station (Field type: Numeric) On.plot.weather.station.air.temperature..degC.: Air temperature measurement of the on-plot weather station (Field type: Numeric) Mircoclimate_proxy.zip Description: S2 Spatial datasets for microclimate, topography, and canopy structure. All datasets are spatially interpolated to 1m resolution and projected into UTM coordinates, zone 50N. Date range: 2015-11-01 to 2015-12-30 Latitudinal extent: 4.5000 to 5.0700 Longitudinal extent: 116.7500 to 117.8200