Climate Change Assessment: Uncertainty Due To Missing Data In Temperature Time Series

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Thursday, 6 February 2014: 12:00 AM
Room C205 (The Georgia World Congress Center )
Luciano Massetti, Institute of Biometeorology/National Research Council of Italy, Firenze, Italy

Assessing future climate trends and global warming requires the integration of long and complete historical data series collected by meteorological networks with other types of data sources like mathematical models and paleoclimate data. Therefore the degree of uncertainty introduced by each source should be accurately quantified and incorporated in the final assessment. This study focuses on uncertainty introduced on climate trend analysis by the use of incomplete temperature time series recorded by weather stations. The estimation of global warming is based on the analysis of monthly average or extreme temperatures calculated on daily observation data collected by network of weather stations worldwide. These time series are often affected by gaps due to malfunctioning or poorly calibrated instrumentation. Furthermore, these networks are managed by public or private organization at national, regional or local level that might apply different data quality strategies. For this reason, data quality is a strategic issue for the reliability of this kind of studies and particularly missing data management strategy is a key aspect whose effects should be properly considered. Two approaches are commonly used: replacement or filling gap techniques and exclusion criteria based on the maximum acceptable number of missing values on the time series. A standard methodology is not yet defined and accepted worldwide for both approaches and the application of different methods to the same time series could even lead to different results. The first aim of this study is to assess the uncertainty of monthly average temperature calculation based on daily temperature records and its effects on the estimation of climate trend. Long and complete data series of some weather stations worldwide are selected for this analysis. For each series, several new series are generated simulating the presence of missing values. Random or consecutive values are dropped from the series and monthly average temperature of the incomplete series are calculated. The relationship between missing values and the difference between average values of the complete and incomplete data series are then analyzed. The second aim is to analyze the effect of the uncertainty due to missing data on the estimation of the trend of temperature time series. The same approach is applied to monthly time series. Errors due to missing daily data are simulated at different point of the monthly time series and sensitivity analysis is performed to quantify uncertainty of trend estimation caused by the amount of error and its position in the monthly series.