Highly sampled measurements in a controlled atmosphere at the Biosphere 2 Landscape Evolution Observatory

Site and instruments description

LEO (https://biosphere2.org/research/projects/landscape-evolution-observatory) is located at Biosphere 2, Oracle, Arizona, USA (https://biosphere2.org) and operated by the University of Arizona. It is composed of three near-identical greenhouses (Fig. 1a) covered but not sealed, by an 11 mm thick glass with an interior mylar sheet. The glass has a solar heat gain coefficient of 0.7, transmitting between 50% and 60% of total solar radiation but less than 1% of UV solar radiation28,29. The three greenhouses (Fig. 1a) are named East, Center and West bays, each of them containing an air volume of approximately 12,550 m3, 12,950 m3 and 12,550 m3, respectively; they are all facing to the south-southwest. Although the enclosed atmosphere could be highly controlled, it has been most of the time naturally driven except for precipitation during the experiments and temperature with the purpose of keeping the bays at temperatures allowing the work of the scientists.

Fig. 1

The Landscape Evolution Observatory. Schematic figure of LEO (a) and a picture of the above ground instrumentation in the West bay (b).

Inside each bay there is one artificial bare soil hillslope of 11 × 30 m2 of surface with an average slope of 10° and 1 m average soil depth. The soil corresponds to ground basaltic tephra with a loamy sand texture and a dry bulk density of 1.5 gcm−3; more detailed information on soil physical and chemical properties can be found in the main article describing LEO15.

Buried in the soil of each hillslope are more than 1,200 sensors measuring soil water content, soil water potential, soil temperature, soil carbon dioxide concentration, heat flux, electrical resistivity, and hydrostatic water pressure. There are also more than 630 sampling points, allowing physicochemical analyses of water and gases within the soil. Outside the hillslopes, water storage in the soil is accounted for through 10 large load cells for each hillslope whereas discharge is monitored by a combination of tipping buckets and electromagnetic flowmeters. Above ground, there are more than 50 sensors in each bay to monitor the enclosed atmosphere by measuring temperature, relative humidity, wind speed and direction, and radiation fluxes at different heights (Fig. 1b). Precipitation is not measured directly but precisely controlled by the irrigation system. The most recent data acquired in LEO can be visualized in http://biosphere2.org/research/leo-data.

Although all the monitored variables in LEO are valuable for the scientific community, the scope here is to provide data that helps to improve our understanding of the soil-atmosphere interactions. Careful processing, quality control, and then sharing of other data, including that from more than 3,000 sensors buried deeper in the soil, are left to future efforts. Hence, this dataset30 compiles part of the LEO measurements in the three individual hillslopes of (1) meteorological variables above the hillslopes’ surface, (2) soil moisture, heat flux and temperature of the soil near the surface, (3) precipitation, discharge and water storage aggregated for the entire hillslope, and (4) meteorological variables outside LEO from an automatic weather station. A basic description of each instrument used for the measurements included in this dataset is available in Tables 1, 2, and 3, while their locations within each bay are shown in Fig. 2.

Table 1 Summary of the above ground instrument’s specifications.
Table 2 Summary of the in-soil instrument’s specifications.
Fig. 2

Instrument locations on each LEO artificial hillslope. Soil sensors are at 5 cm depth. T is temperature, RH is relative humidity, HF is soil heat flux, and VWC is Volumetric Water Content. Net radiation includes the four radiation fluxes.

The atmospheric sensors are located on five retractable masts at 0.25, 1, 3, 6, and 9–10 m above ground, with the exception of the mast at 4 m from the bottom of the hillslope where only the four lowest levels are present. At each level along the masts (Fig. 2), air temperature, relative humidity, and wind speed and direction are measured. On the two masts located at 10 m from the bottom of the hillslope, there are 4-channel radiometers measuring downward and upward longwave and shortwave radiation fluxes. In addition, a 3D sonic anemometer is located on the mast at 17 m from the bottom of the hillslope. Specifications for the above ground instruments are listed in Table 1. All masts are lifted during each rain event to avoid interference with the rain and nonuniform erosion of the soil through dripping. When the masts are lifted, most of the above ground measurements are not usable, as the location and orientation of the sensors are changed. Only the topmost measurements of temperature and relative humidity show some temporal consistency during such rain periods as they do not have much dependence on the orientation of sensors and their location is only slightly modified.

Among the sensors buried in the soil, there are measurements of near surface (at 5 cm depth) soil heat flux, soil temperature, and soil volumetric water content which are also included in the dataset (Fig. 2, Table 2). Heat flux is measured with two independent instruments at each location, HFP-1 and HFP-1SC.

Precipitation, discharge and water storage are also monitored in LEO, but they required additional processing, which was performed and explained in the following paragraphs. Details of the instruments and the aforementioned hillslope-scale quantities are available in Table 3.

Table 3 Summary of hillslopes instrument’s specifications.

Precipitation is not directly measured but highly controlled through a complex irrigation system. The total volume of water flowing through the irrigation main line is recorded and then converted to rain rates by dividing the water volume by the area of the hillslopes (330 m2) and by the time aggregation period in hours, to be included in this dataset. Droplet size distributions, terminal velocities of the droplets, and spatial homogeneity of the precipitation have been studied in the past15, showing that the irrigation system is able to produce rain droplets that achieve velocities close to that of natural rain. The spatial distribution has coefficients of variation between 0.2 and 0.7 with more homogeneous distributions occurring at higher rain rates.

LEO’s discharge is routed through a porous plate at the seepage face (11 m2) to six dividers and flow through a SeaMetrics PE102 flow meter (for high flow discharge) and then through a NovaLynx 26-2501-A tipping bucket gauge (for low flow discharge), both of which are included in this dataset. The Center hillslope had developed a leak at the bottom of the hillslope and additional flow meter and tipping bucket gauge were added to account for this in the total discharge. Flow meters have a low accuracy for flows below the equivalent to 0.025 mmh−1, while tipping buckets tend to underestimate the high flows. Our dataset also includes the computed total discharge as a reference. This quantity was calculated as the sum of the flows of the best available measurement, in each divide, based on the discharge rate. The flow meter (tipping bucket) value was selected for flows higher (lower) than 0.025 mmh−1 if both measures passed the quality control or whichever is available if only one passed the quality control. For this computation, a zero value was assigned to a divider when there was no discharge flow or measurement were missing or below the range.

Our dataset also includes measurements from 10 Honeywell Model 3130 load cells on each hillslope to monitor the total water storage. Relative water storage was computed by adding the weights from the 10 load cells (only when all of the measures passed the quality control) and then subtract the total mass weight from the lowest water storage content during the time period of this dataset, occurred on 6 November 2016. Assuming a water density of 1,000 kgm−3, the results were divided by the slope area (330 m2) to obtain the relative water storage content in mm which is also included in the dataset.

Although the greenhouses in LEO allow the control of many variables, some of them are still impacted by the external conditions. For instance, solar radiation inside the bays is directly dependent on the amount of solar radiation outside LEO and indoor temperature is highly impacted by the outside temperature. Hence, data from a weather station WeatherHawk 710 located outside the building were also included. Pressure, solar radiation (300–1100 nm), temperature and relative humidity of the air, are the prevalent variables impacting the internal environment of LEO but precipitation, wind speed and wind direction were also included for completeness.

Experiments and precipitation control in LEO

Several specific experiments have been conducted in LEO which has led to particular rain patterns. In the provided dataset, there were two extensive tracer experiments carried out by the end of 2016 and in mid-2019.

The first extensive experiment was a 28-day tracer experiment conducted from 1 to 28 December 2016. The experiment was designed to observe the transit time distributions (TTDs) and the StorAge Selection (SAS) functions31, which are system-scale hydrologic transport signatures. Those functions were directly observed using the experimental protocol PERiodic Tracer Hierarchy (PERTH) method32. The method required driving the hillslopes to a periodic steady state. Therefore, the hillslopes were irrigated with two 3-hour pulses of 12 mmh−1 at 7-hour intervals every 3.5 days.

Before this experiment, no irrigation was performed within the period covered by this dataset until 6 November 2016, leading to the driest period recorded in those hillslopes. After this experiment was concluded, no irrigation was performed for about 4 months. In order to support the biogeochemical dynamic of the soil, a quasi-regular irrigation sequence, with precipitation almost every two weeks but with a few longer dry periods in between was performed until late June 2019 when the next major experiment started.

The second major experiment was conducted from 24 June to 16 August 2019, with only 7 days within this dataset. Its goal was to test a new TTD estimation method, which is not limited to a periodic steady state. The irrigation sequence was generated stochastically using a rainfall generator33. During this experiment, the total irrigation amount was about 750 mm with a mean irrigation rate of about 5.3 mmh−1. The mean duration of the irrigation pulses was 2.5 hours, and the mean inter-irrigation time was about 17 hours. Due to the complicated irrigation sequence, the mast operation was different from its regular operation and the corresponding times when masts were lifted were flagged in the quality control companion files.

Instruments calibration and uncertainty

All the instruments were calibrated in the factory by the manufacturers, some of which have unique calibration coefficients that are applied before storing the data. The uncertainty reported in Tables 1, 2, and 3 correspond to that reported by the manufacturers. Additional uncertainties arise from different sources such as the data-acquisition devices, numerical roundoff, and dependence on other environmental conditions, among others.

The DVI7911 anemometers are known to have a very low accuracy for wind speeds below 0.5 ms−1 which occurs within LEO more than 99% of the time. Their corresponding wind vanes have even lower accuracy for such wind speeds and hence they were not included in this dataset. The CNR4 net radiation sensor is composed of 2 pyranometers and 2 pyrgeometers, each of them having unique calibration coefficients that are applied before storing the data. They also have internal temperature sensors that automatically compensates for thermal drifts.

For the volumetric water content, a calibration curve was developed specifically for the LEO soil through lab experiments based on four 5TM sensors buried in a sample of the same basaltic material with the same bulk porosity. This fitting curve allows to derive the volumetric water content from the directly measured dielectric permittivity, with a 95% confidence. The custom calibration was applied by the factory for Biosphere 2.

The EX81 flowmeter sensors use a calibration coefficient dependent on the material and size of the irrigation piping and provides the total volume of water sprayed by the irrigation system. This does not account for the small losses of water that fall outside the hillslopes’ surface nor the evaporation from the rain drops that occurs before they reach the surface. Hence, the precipitation is slightly overestimated by this measure. To reduce the uncertainty in the total precipitation rates, the changes in water storage content derived from the load cells and mass conservation can also be used.

The PE102 flowmeters have a low flow cutoff of 0.025 mmh−1, hence the flow is then routed to the tipping buckets which in contrast are known to underestimate the high flows. The Novalynx conversion rate between tip pulses and volume of water was derived empirically through manual calibration. The total discharge computed for this dataset has a larger uncertainty due to the assumption of no flow when data were missing or below the range. Therefore, a more detailed filling of the missing data is recommended for case studies.

Data aggregation

The data were typically logged every 15 minutes but during specific experiments where high temporal resolution was required, the data were logged every minute for many sensors. Hence, in order to homogenize the series, all the data were aggregated/averaged to 15-minute intervals and labeled with the time for the end of the aggregation period.

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *