<p>Infrared gas analyzers (IRGAs) are commonly used in Eddy Covariance (EC) system and are used for, in particular, the ecosystem water cycle. However, they suffer from a measurement drift of absolute concentrations with time, leading to the increasing bias of readings. It is recommended in the manuals to do a factory calibration once every 1-2 years (e.g., LI-6262) or user calibration when considerable drift occurs (e.g., LI-7000). However, our experience shows that a significant drift can occur within a few days already. At our semi-arid EC site of Yatir Forest (31&#730;20'N, 35&#730;03'E, Israel), we are measuring a vertical air humidity profile (absolute humidity, C<sub>w</sub> in mmol&#215;mol<sup>-1</sup>, and relative, RH, %),&#160; to study the VPD regime within the canopy and to analyze dew formation events, which requires highly accurate RH measurements, however accurate RH measurements are difficult to achieve.</p><p>Air humidity in Yatir is measured by three different instruments: (1) LI-7000 close-pass IRGA above the canopy for EC flux calculations; (2) LI-6262 close-pass IRGA with inlets in 4 different heights from above the ground up to the sonic height, used for humidity profile measurements; (3) Rotronic HC2S3 air humidity (RH) and temperature (T) sensor above the canopy. Both IRGAs are placed within a temperature-controlled box, and calibrated for zero and span with N2, dew point generator and laboratory standard gases every 1-2 weeks. The Rotronic sensor has very low drift and does not require calibration, but is assumed to be less accurate, especially under high and low RH.</p><p>To achieve highly accurate measurements on daily time scale we propose a correction routine that rely on the stability of the RH probe, and the accuracy of the IRGAs after calibration. Every time the IRGA is calibrated, a correction-1 to the RH probe is produced. Between calibrations, the trends in the drifting IRGAs data are corrected (correction-2) to the interpolated stable RH probe data.</p><p>For the flux measurements, the mean absolute Cw error before correction was 1.0 mmol&#215;mol<sup>-1</sup>, which translates under average temperature of 25<sup>&#176;</sup>C and RH of 50% to errors of RH, VPD and dew point of 3.0%, 93.5 Pa and 0.9<sup>&#176;</sup>C, respectively. Following our correction procedure, reduced the error to 0.5 mmol&#215;mol<sup>-1</sup>, which decreased the errors in RH, VPD and dew point under the same conditions to 1.5%, 47 Pa and 0.4<sup>&#176;</sup>C, respectively. For the humidity profile, Cw error after correction decreased from 1.9 mmol&#215;mol<sup>-1</sup> to 0.5 mmol&#215;mol<sup>-1</sup>, which decreased the errors in RH, VPD and dew point under the same conditions by 4.1%, 131 Pa and 1.2<sup>&#176;</sup>C, respectively.</p><p>We will describe the method in more detail and demonstrate its application to our field measurements.</p>