Long-term Trends of Land Surface Temperature over Europe derived from a daytime normalized AVHRR Time Series

Kurzfassung

For monitoring conditions repeatedly over large areas, satellite derived Land Surface Temperature (LST) has become an indispensable tool. However, to make climate relevant statements and quantify the impact of land surface variables over long time, we need sensors that are available for more than 30 years. The AVHRR is the only sensor providing spatially and temporally continuous, daily measurements for 40 years. The TIMELINE project of the German Aerospace Center (DLR) aims at the generation of a homogeneous multi-decadal time series from AVHRR data (3 different AVHRR sensors on 16 NOAA platforms) over Europe and North Africa [1]. However, different observation times of the AVHRR sensors affect the observed temperature and therefore the AVHRR LST time series [2]. In addition, the orbital drift effect, resulting in a slowly changing observation time during the lifetime of one sensor is influencing the time series. In the past, several methods have been developed to account for the effect of varying acquisition times on the AVHRR LST time series. An important requirement for these methods is that they preserve the actual trends in LST. The methods can be classified into physical and statistical methods. While statistical models use the regression between the LST anomalies and the corresponding sun zenith angle (SZA) anomalies, the physical methods model the daily temperature circle at the given location. The statistical models have already shown accurate results; however, they are computationally expensive and their performance depends a lot on the data availability of the respective pixel. Therefore, they have been only tested for single pixels [3-5]. The physical models have been already applied to larger areas, however their diurnal LST cycle models are largely generalized and therefore linked to high uncertainties [6]. In this study we have combined our AVHRR LST time series with geostationary SEVIRI LST data. The daytime normalization is performed via correction terms, which are derived individually for every pixel from a typical diurnal cycle model build on the SEVIRI observations [7]. The resulting dataset provides daily maximum LST in 1 km resolution for the last 40 years for Central and Western Europe. The time series was already compared to the CCI LST time series for the years 1996-2016 and to air temperature measurements for the whole period showing a high accordance. Furthermore, we validated the product, by comparing observations from different NOAA platforms on the same day. Here, we show the first analysis of long-term LST trends in 1 km resolution over Europe. The trends were calculated by conducting the Mann-Kendall significance test and calculating the Theil Sen slope. Only significant trends with a significance level p < 0.1 were analysed. Furthermore, we classified the trends regarding land cover and elevation. Thus, we are presenting a unique dataset, allowing a wide range of applications in the context of climate and land surface change.