Open AccessField and Laboratory Evaluation of the CS655 Soil Water Content Sensor
Author(s) -
Caldwell Todd G.,
Bongiovanni Tara,
Cosh Michael H.,
Halley Chelsea,
Young Michael H.
Publication year - 2018
Publication title -
vadose zone journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.036
H-Index - 81
ISSN - 1539-1663
DOI - 10.2136/vzj2017.12.0214
Subject(s) - water content , attenuation , soil water , calibration , soil science , mean squared error , environmental science , hygrometer , remote sensing , meteorology , mathematics , geology , physics , optics , geotechnical engineering , statistics , humidity
Core Ideas The CS655 measures two‐way travel time and attenuation of a low‐frequency EM pulse to derive K a , EC, and SWC. Laboratory and field evaluations for five central Texas soils found the sensor robust but RMSE could be reduced by a soil‐ or site‐specific function. The factory calibration has a cross‐over point at K a ∼15, with low bias below and high bias above. The TxSON network mean was unaffected by calibration but the variance was significantly reduced. Soil moisture sensors infer volumetric soil water content (SWC) from other properties of the bulk porous media. The CS655 water content reflectometer is a relatively new, low‐frequency electromagnetic sensor that determines relative permittivity ( K a ) using the two‐way travel period and voltage attenuation of the applied signal along two 12‐cm rods. This measured attenuation is quadratically related to bulk electrical conductivity (EC). Along with an onboard thermistor, the CS655 allows a more robust correction of propagation time and K a , which its predecessors, the CS615 and CS616, lacked. However, with new sensors it is necessary to quantify their practical accuracy in the field. Here, we present an overview of the CS655 sensor and an evaluation under both laboratory and field conditions, using five surface soils (0–10‐cm depth) in the laboratory and gravimetric samples collected in the field. Overall, a site‐specific calibration using a two‐term linearization of the SWC– K a function reduced the root mean square error (RMSE) of the factory‐derived SWC of 0.073 and 0.043 m 3 m −3 during batch and infiltration experiments, respectively, to 0.025 and 0.028 m 3 m −3 . Results further indicate that a soil‐specific calibration additionally reduced the RMSE to <0.02 m 3 m −3 . Field evaluation across the Texas Soil Observation Network found that calibration reduced the variance across the network but did not affect the arithmetic mean or the RMSE against gravimetric sampling, which remained ∼0.05 m 3 m −3 regardless of the SWC– K a –EC function applied. At the regional scale, a global calibration is sufficient.