RECONSTRUCTION OF INCORRECT SENSOR DATA: FROM FAILURE TO SUCCESS

Abstract

Experimental research on crop performance or climate control in greenhouses requires accurate measurement of climate conditions. Despite the use of calibrated sensors, afterwards it may appear that the collected climate data are incorrect. Then the question arises whether we need to repeat the experiment or whether we can reconstruct the correct climate data from the incorrect measurements. In this paper we show how correct data of temperature and air humidity could be reconstructed from incorrect data. In a large research project on Botrytis development and energy use in the greenhouse cut flower gerbera, measuring boxes were installed in 12 commercial greenhouses. The measuring box consisted of a temperature sensor, an air humidity sensor and a CO2 sensor and on top a PAR sensor. Before the start of the experiment all sensors were taken out the measuring box for calibration by the supplier of the sensors. Subsequently the measuring box containing the sensors was tested at different temperatures and air humidity?s in a climate chamber. Nevertheless the temperature and air humidity data gathered in the greenhouses appeared to be erroneously. Temperature was overestimated up to 1-2°C while relative air humidity was underestimated by 10-20%. Detailed analysis afterwards showed that the climate in the measuring box was affected by irradiance, mainly due to insufficient ventilation of the measuring boxes. The measurement error depended on the irradiance and the rate of change in climate conditions. To analyze this error, simultaneous measurements with certified measuring devices and the measuring box were performed in a greenhouse under a wide range of dynamically varying conditions. The measured temperature was determined by a first order effect of the temperature of the greenhouse air as well as a first order effect of irradiance. Similarly measured air humidity showed first order relations to humidity of the greenhouse air and to irradiance. The data with the certified sensors were used to estimate the parameters of the first order relations. Subsequently these relations were used successfully to reconstruct the temperature and air humidity from the incorrect data gathered at the 12 commercial growers. After correction the accuracy of the achieved temperature was approx. +/- 1.0°C, the relative humidity approx. +/- 3% similar to the technical specifications of the sensors.