The first topic in the course of this work was to analyse data of a sensor node in the OK Lab air quality network that uses a SDS011 low-cost PM sensor and a low-cost temperature/humidity sensor with respect to the humidity influence. This sensor node was located close to a governmental measurement station for air quality in Munich Lothstrasse, which we used as reference. Based on the PM and humidity data collected, we tried to correct the humidity influence on the low-cost sensor. The correction was based on a growth model fit to the ratio between low-cost sensor and reference data. The second topic was to compare the computationally corrected results to results obtained from a Twin-SDS measurement box containing one SDS011 equipped with a low-cost thermal dryer and another SDS011 without thermal dryer. The measurement box was operated during various field measurements and during a co-location measurement in Munich. Particularly, hygroscopic growth events and fog events were analysed and compared to the computational correction method. By accident also two Sahara dust events were captured and analysed. The results gave additional insight into error mechanisms related to the SDS011 low-cost sensor. For the co-location experiment, the constructed Twin-SDS measurement box was operated near the governmental measurement station in Munich, Lothstrasse. For the co-location measurement period, a comparison of the dryer efficiency to the professional equipment was done. The results revealed that the SDS011 low-cost sensor data can be corrected for hygroscopic growth of PM2.5 with a growth model obtained by the help of reference equipment and using humidity data of an accurate humidity sensor only under certain conditions. The computational correction is limited to one location and a short timeframe under the condition that the hygroscopic properties of the aerosol remain stable. Additionally, a highly accurate humidity sensor suitable for outdoor use is required. A low-cost indoor humidity sensor as used in the sensor network is not sufficient. In contrast to the computational correction, the use of a thermal dryer showed a high efficiency to correct the influence of hygroscopic growth and even fog events, independent of location or time, without the need of reference equipment once being calibrated. However, the control of the dryer requires the modification of the SDS011 sensor by adding a humidity sensor into the exhaust chamber. It also would make a potential network sensor node more bulky and more costly. Nevertheless, it could be shown that the use of a thermal dryer significantly improves the data quality of the SDS011 low-cost PM sensor and achieves an excellent matching to the reference equipment of a governmental measurement station.
Figures - uploaded by
Bernd LaquaiAuthor contentAll figure content in this area was uploaded by Bernd Laquai
Content may be subject to copyright.