Wireless sensor networks and fusion of contextual information for weather outlier detection

Weather stations are often expensive hence it may be difficult to obtain data with a high spatial coverage. A low cost alternative is wireless sensor network (WSN), which can be deployed as weather stations and address the aforementioned shortcoming. Due to imperfect sensors in WSNs context, provided raw data may be drawn in from of a low quality and reliability level, expectedly that is an emergence of applying outlier detection methods. Outliers may include errors or potentially useful information called events. In this research, forecast values as contextual information are utilized for weather outlier detection. In this paper, outliers are identified by comparing the patterns of WSN and forecasts. With that approach, temporal outliers are detected with respect to slopes of the WSNs and forecasts in the presence of pre-defined tolerance. The experimental results from the real data-set validate the applicability of using contextual information in the context of WSNs for outlier detection in terms of accuracy and energy efficiency.