| dc.description.abstract | The main objective of this research was to examine the potential benefits that are associated with the integration of formal and informal data. This was studied according to the integration of amateur weather station data derived from the WOW-NL application and formal KNMI data (temperature measurements). It is known that the integration of these types of sensor data can improve the spatiotemporal resolution of formal data sets. Hence, it was examined to what extent this was true for this specific case. However, the use of informal data requires caution since there is a considerable lack of quality control and validation standards. As a result, it was imperative to conduct substantial pre-processing and an elaborate data assessment before the WOW-NL data could be integrated with the KNMI data. In order to determine the quality of WOW-NL data, its observations were compared with reliable temperature estimates. Accordingly, these were derived from interpolations based on formal KNMI data. However, this required the selection of the best interpolation method for temperature measurements with a 10 minute temporal resolution. An extensive inter-comparison showed that Tin Plate Splines is the best performing interpolation method in this regard. The assessment of the WOW-NL data showed that the WOW-NL stations generally observe higher temperatures than were estimated. Furthermore, the WOW-NL data also contains a substantial amount of gross errors and outliers that had to be removed prior to the data integration. Besides that, the WOW-NL observations deviate most from the estimates during the day in summer. Station attributes that were derived from the metadata did not show notable patterns in this regard. In addition, there were quite some stations that only showed considerable deviation from the interpolations when the predicted temperatures were above approximately 20 °C. Since amateur weather stations are known to have weak radiation shields which cause them to overheat, it is likely that these patterns are the result of radiation bias. When the gross errors and outliers were removed, the data could be integrated. This was done according to three different integration scenarios. These include: (1) threating the WOW-NL data as equal compared to the KNMI data, (2) using the WOW-NL data only as a secondary predictive variable, and (3) making corrections for solar radiation to the WOW-NL data.
The first scenario showed that the integrated data improved temperature interpolations for the Netherlands in both October and January. However, in August the integrated data did not improve the interpolations. Equally, the second integration scenario showed that the integrated data improved interpolations considerably for October and January. For August, the improvements were negligible. The third scenario showed that corrections could be made according to incoming solar radiation. However, the corrections only resulted in a marginal improvement over the original WOW-NL data for the whole study period. When the same relation was modeled for exemplary warm days in August, more substantial corrections could be made, and solar radiation had more predictive power. Finally, the integrated datasets could be used to make maps with an increased spatiotemporal resolution.
Finally, this research concludes that the integration of the WOW-NL data and KNMI data can improve the spatiotemporal resolution of meteorological data and maps. However, the extent to which this is true is highly dependable on the time of the year and the data integration method. | |
