Three soil moisture sensor types were used in this study, including the Decagon EC-5 (Decagon Devices, Pulman, WA), the Vegetronix VH400 (Vegetronix, Inc., Riverton, UT), and the Watermark 200ss (Irrometer Company, Inc., Riverside, CA) (Figure 1). These sensors were selected for their relatively low cost, which makes them affordable for most growers. Current purchase prices for these sensors are around US $110 for the Decagon EC-5, US $40 - 56 (depending on cable length) for the Vegetronix VH400, and US $33 for the Watermark 200ss.

The Decagon EC-5 sensor determines volumetric water content (VWC) by measuring the dielectric constant of the media using capacitance/frequency domain technology [14] . The sensor manufacturer claims that the 70 MHz frequency of the EC-5 sensor minimizes salinity and textural effects, making it accurate in almost any soil or soilless media. The rated accuracy of the sensor is ± 3% VWC in most mineral soils with a salinity level of up to 8 dS/m, and ±1% - 2% VWC with soil-specific calibration. The rated accuracy is also ±3% VWC for rockwool (0.5 to 8 dS/m) and potting soil (3 to 14 dS/m). The sensor requires an input voltage of 2.5 to 3.6 VDC at 10 mA and produces an output voltage that is proportional to the soil moisture and to the input voltage. It is designed to operate within the temperature range of −40˚C to +50˚C and requires a measurement time of 10 ms.

The Vegetronix VH400 sensor measures the dielectric constant of the soil using transmission line techniques [15] . The sensor has an internal voltage regulator and operates with an input voltage of 3.5 to 20 VDV, requiring an input current of less than 7 mA. It produces an output voltage in the range of 0 to 3 VDC, which can be measured with a regular multi-meter, a data logger, or a microcontroller. The output voltage is related to the water content of the soil. The sensor is also sold as a water level sensor, which explains why it is externally labeled similar to a measuring ruler. The sensor was designed to be insensitive to salinity, operates within the temperature range of −40˚C to +85˚C, and requires a measurement time of 400 ms after power on to get a stable output.

The Watermark 200ss sensor [16] is a solid-state electrical device that senses electrical resistance and is commonly used to measure soil water potential, rather than soil moisture. The sensor has a pair of corrosion-resistant electrodes in

the shape of two concentric rings. The space between the two rings is filled with a gypsum material to provide a buffer against changes in salinity. The electrodes are imbedded within a granular matrix (a material similar to very fine sand) that creates the bulk of the instrument. The granular matrix is enclosed in two layers, the inner layer is a filter-like material that allows water exchange between the granular matrix and the soil, while the outer layer is a perforated stainless steel frame that provides rigidity and maintains the shape and physical integrity of the sensor. Two ABS plastic green caps are installed at each end of the sensor (Figure 2).

The electrodes are connected to two lead wires (AWG 20) that attach to the data logger. When installed in the soil, water is exchanged between the soil and the granular matrix to reach equilibrium. Since water is an electrical conductor, the resistance between the electrodes decreases as soil moisture increases. The sensor requires application of an electrical current (AC rather than DC) and the output is an electrical resistance (in units of ohm). The electrical resistance correlates to soil water potential (negative pressure), which is usually expressed in units of centibars (cb) or kilopascal (kPa) (cb = kPa), with a rated range of measurement from 0 to 239 cb. Guidelines for using Watermark sensors for irrigation scheduling have been developed by Irmak et al. [17] [18] .

In this study, hand-held devices were used to measure the output of the three types of sensors (Figure 3). The output from the EC-5 sensors was measured using a ProCheck sensor readout device (Decagon Devices, Pulman, WA), which outputs the VWC using the factory calibration for mineral soils. The output of the VH400 sensors (in volt) was measured with a FLUKE 117 electrician multi-meter (Fluke Corporation, Everett, WA). The output of the Watermark 200ss sensors (in cb or kPa) was measured using a Watermark Meter (Irrometer Company, Inc., Riverside, CA), which is an alternating current (AC) resistance bridge meter specifically designed to read the Watermark sensors.

Four soil texture types that are typical of South Carolina farms were used in this study. Physical properties of these soils are shown in Table 1. Soil texture for each soil type was measured in the laboratory using the Hydrometer method [19] . The soil bulk density (BD) was also measured using a disturbed soil sample. Estimates of the volumetric water content (VWC) at field capacity (FC) and permanent wilting point (PWP) were obtained from Saxton and Rawls [20] . The four soil types used in this study all have a high proportion of sand, which is typical of South Carolina soils, and represent the left lower corner of the soil texture triangle (Figure 4).

Before the start of the experiment each soil was oven-dried at 105˚C for 24 hours using a model 230F IsoTemp oven (Fisher Scientific, Loughborough, UK). Then, for each soil type, three square plastic pots were filled with a 700 cm³ volume (V_soil) of oven-dried soil. The mass of each pot was measured using a precision laboratory balance with a maximum capacity of 2200 g and accuracy of 0.01 g (Kern & Sohn GmbH, Balingen, Germany). Each pot was placed on top of an aluminum plate and the three sensor types were installed inside each pot (Figure 5).

The weight of the container (W_c), the weight of the three sensors (W_s), and the dry weight (W_dry) of the whole setup (container + dry soil + sensors) were recorded (in grams). The soil was compacted around the sensors to have good contact. After installing the sensors, water was added to wet the soil to field capacity (FC) and the wet weight (W_wet) (container + wet soil + sensors) was recorded. The amount of water added was estimated by multiplying the FC value by the volume of oven-dried soil (V_soil). The soil was then allowed to dry at room temperature until the soil was air-dried. During this period, the output of each

sensor and the W_wet were recorded once a day at about 9:00 am. The measured W_wet was used to determine the gravimetric soil VMC (%) as:

The daily gravimetric soil VMC values for each soil type were compared with the sensor readings using linear and non-linear curve-fitting methods. Statistical analyses, including linear regression and analysis of co-variance (ANCOVA) were conducted using the statistical software R, version 3.0.2 [21] . ANCOVA was conducted to test the effect of soil type on the performance of the different soil moisture sensors. This test is normally used to compare two or more regression lines by testing the effect of a categorical factor on a dependent variable while controlling for the effect of a continuous co-variable. The ANCOVA was performed fitting the following model:

where, aov = analysis of variance, y = dependent variable (sensor output), cv = covariance variable (gravimetric VWC), and F is the categorical factor variable (soil texture). Linear regression analysis was used to fit the linear response of the EC-5 and VH400 sensors to changes in soil water content. The fitted linear models were of the form:

where, X = independent variable (gravimetric VWC). The curvilinear response of the Watermark 200ss sensors, however, was fitted using Sigma Plot (version 12.5) (Systat Software, Inc., San Jose, CA) using a three-parameter exponential decay model as:

where, y_o, a, and b are empirical constants fitted to the data. In this equation, y_o represents the minimum value; a represents the y intercept, and b represents the slope or degree of curvature of the line.