Towards estimating root‐zone soil moisture using surface multispectral and thermal sensing: A spectral and hydrometeorological dataset from the Dookie experiment site, Victoria, Australia

This paper describes surface hydrometeorological and spectral datasets collected from two tower sites located in the University of Melbourne's Dookie experimental farm, Victoria, Australia. The datasets were collected from different vegetation types including wheat, canola, and grazed pasture during the 2012, 2013, and 2014 cropping seasons. The dataset includes 30‐min frequency latent and sensible heat flux measurements and layer‐average soil moisture data at profile depths of 0–5, 0–30, 30–60, 60–90, and 90–120 cm. Air temperature, wind speed, wind direction, relative humidity, precipitation, and incoming and outgoing longwave and shortwave radiation data were also collected from two locations in the study area. The dataset described in this paper is available online.

All sensors were connected to CR3000 and CR5000 data loggers (Campbell Scientific, Inc.) for data storage. Field visits were planned approximately twice a month in order to download data. Although all sensors collected data automatically, periodic level checks, levelling adjustments, and cleaning were undertaken to maintain accuracy of data collection. Field visit days and activity log file have been added to the online data repository.

| METEROLOGICAL DATASET
Meteorological data consisting of air temperature, wind speed and direction, relative humidity, atmospheric pressure, net radiation, rainfall, latent heat flux, sensible heat flux, and soil heat fluxes were collected at both study sites. Wind speed and wind direction were measured using a wind sentry set (03101 R.M. Young) that consisted of a three-cup anemometer and a potentiometer. Air temperature and relative humidity were measured using an HMP45C probe (Campbell Scientific, Inc.). A CS105 barometric pressure sensor (Campbell Scientific, Inc.) equipped with Vaisala's capacitive pressure sensor was used to measure barometric pressure. The output of the sensor in the form of current ranges from 0 to 2.5 V that corresponds to pressure from 600 to 1,060 mb. All meteorological measurements were collected available at 30-min intervals.

| Net radiation
A CNR1 net radiometer (Kipp & Zonen, Inc.) was used to measure the upwards and downwards shortwave and longwave radiation components. Sensor output (V) was converted to radiative flux (W/m 2 ) using manufacturer-supplied calibration coefficients. Net radiation was calculated as follows: The subscripts S and L represent shortwave and longwave radiation, respectively, and upward and downward arrows represent incoming and outgoing radiation. Daily net radiation for 2013 is shown in Figure 2b.

| Rainfall
The TR-525 rainfall sensor (Texas Electronics, Inc.), a tipping bucket rain gauge (0.2-mm tip size), was installed at a height of 1 m at both study sites in order to measure rainfall. Cumulative rainfall was recorded over a 30-min interval. A collecting rain gauge was installed at study sites to measure the total rainfall between bimonthly

| Flux measurements
Latent and sensible heat fluxes were measured using the eddy covariance method. The eddy covariance system consisted of an LI-7500 open path gas analyser (LI-COR, Inc.) and CSAT3 three-dimensional sonic anemometer (Campbell Scientific, Inc.) connected to a CR5000 data logger ( Figure 3). Turbulent fluctuation measurements of threedimensional velocity, humidity, and sonic temperature were recorded at a frequency of 20 Hz for post-processing. The Eddy-Pro software (LI-COR, Inc.) was used to correct high-frequency data and for quality control. A metadata file was configured with instrument height, direction, sensor separation, and dynamic canopy height from field observations. Flux data were corrected to reduce the effects of density fluctuations due to humidity and temperature fluctuations, and default spectral corrections applied (Moncrieff et al., 1997;Webb, Pearman, & Leuning, 1980). The high-frequency 20-Hz eddy covariance data can be shared upon request; however, the high-frequency dataset was converted to a 30-min flux dataset and added to the online data repository.
Soil heat flux (G) measurements were obtained using two sets of soil heat flux plates, a TCAV-averaging soil thermocouple and soil moisture probe (Campbell Scientific, Inc.) installed below the soil surface. Soil heat flux was measured by averaging measurements collected using HFP01 (Hukseflux, Inc.) and CN3 (Middleton Instruments) soil heat flux plates. Output voltage from heat flux plates was converted to soil heat flux using manufacturer-supplied calibration relationship.

| SOIL HYDROLOGICAL DATA
Soil moisture was measured using CS616 (Campbell Scientific, Inc.) water content reflectometers. Soil moisture probes consisted of two parallel stainless steel rods that measure dielectric permittivity of the surrounding medium. The CS616 sensor averaged water content over the entire length of the sensor. Soil moisture probes were installed vertically in the soil profile to measure soil moisture at 0-5, 0-30, 30-60, 60-90, and 90-120 cm. All probes were connected to a data logger that recorded the sensor output (wave period, mS) at 30-min intervals.
Soil moisture sensors typically require soil-specific calibration to provide accurate volumetric soil moisture measurements (Western et al., 2004). To undertake the calibration, undisturbed soil samples collected from the field sites in a metal tube were fully saturated in the laboratory. The saturated soil samples (with soil moisture sensors inserted) were placed in the temperature-controlled chamber for accelerated drying during which gravimetric soil moisture contents were measured for calibration. Similar approach had been used for the soil moisture sensor calibration (Rüdiger et al., 2010;Seyfried & Grant, 2007;Western & Seyfried, 2005).
The calibration coefficients were obtained from a curve fit of the laboratory period measurements and gravimetric measurements. The power function shown in Figure 4 indicates the best curve fit between period measurements and the gravimetric measurements with an R 2 value of .95 as shown in Figure 4. Different calibration relationships were obtained for different soil depths.
The in-field accuracy of CS616 soil moisture probes was checked by comparison with time domain reflectometer (TDR) measurements collected during our field visits. The root-mean-square difference between calibrated CS616 soil moisture measurements and TDR  FIGURE 3 Schematic diagram of soil moisture and micrometeorological sensors installed at study sites where X is the NIR I incident reading (μmol·s −1 ·m −2 ); Y is the Red I incident reading (μmol·s −1 ·m −2 ); Z is the ratio sensitivity of reflected NIR:Red; NIR R(nA) is the reflected reading in nanoamps; and Red R(nA) is the reflected reading in nanoamps.
The output of each sensor was measured at 5-min intervals.

| Radiative surface temperature
Ground-based radiative surface temperature was measured using an Apogee (SI-111) infrared radiometer. The sensor was equipped with a radiation shield and measured target thermal radiance in the 8-to 14-μm atmospheric window. This atmospheric window reduces the effects of water bands below 8 μm and above 14 μm. The sensor measured radiation emitted from the target, which was then converted to temperature using the Stefan-Boltzmann constant and an assumed surface emissivity of 1.0. Errors associated with sensor body temperature were corrected using manufacturer-supplied calibration coefficients. Radiative surface temperature was measured at 5-min intervals and then averaged to produce 30-min interval data ( Figure 2c). Midday radiative surface temperature was compared with MODIS 8-day 1-km resolution temperature data (MOD11A2). The root-mean-squared error difference between ground-based radiative surface temperature and MODIS land surface temperature was 4.3°C.
5 | DATA QUALITY AND APPLICATIONS

| Data quality control
Data collected from experimental sites were visually inspected to identify errors. Errors associated with field activities were identified from daily photographs and removed from the dataset. Surface reflectance and radiative surface temperature data were compared with MODIS and CROPSCAN data. Variations in evapotranspiration were cross-checked with net radiation and air temperature. Errors or data gaps in meteorological data at one site were filled with meteorological data from the other site. Energy balance closure in the Study Sites 1 and 2 was 0.76 and 0.84, which indicates good energy budget closure during the study period.
Soil moisture data were checked by comparison with soil moisture from other layers and rainfall data. Soil moisture from 0 to 30 cm was also compared with TDR measurements collected during field visits.
Rainfall data collected from tipping bucket rain gauges were shown to record more rainfall by a mean of 11% when compared with collecting rain gauge measurements recorded during field visits.

| Applications
The Dookie hydrometeorological and spectral dataset can be used for various purposes. The dataset is useful for understanding the sensitivity of Evapotranspiration (ET) to root-zone soil moisture in agriculture landscapes (Akuraju, Ryu, George, Ryu, & Dassanayake, 2013, 2017 and how this relationship might manifest in remote sensing data. This dataset is also suitable for potential evapotranspiration calculations and land surface modelling. Surface reflectance data collected along with photos would be useful for understanding vegetation dynamics of different crops. Hydrometeorological and spectral datasets are well suited for validation and testing of remote sensing ET and Soil moisture (SM) products. Soil moisture data could be useful for developing models to estimate or validate surface and root-zone soil moisture based on optical and thermal remote sensing. Although the authors continue to analyse and utilize this dataset, it is available to other researchers to use.

| Data availability
The data that support the findings of this study are openly available in figshare at https://melbourne.figshare.com/projects/Dookie_hydro-meteorological_dataset_2012-2014/61451. Data are shared under a Creative Commons attribution licence (CC BY) and must be appropriately cited.

| CONCLUSIONS
A comprehensive dataset including meteorological, soil moisture, surface flux, surface temperature, and spectral measurements over two rain-fed agricultural fields in Victoria, Australia, has been described.