Detecting and adjusting artificial biases of long‐term temperature records in Israel. (1st April 2018)

• Record Type: Journal Article
• Authors: Yosef, Yizhak; Aguilar, Enric; Alpert, Pinhas
• Is Part Of: International journal of climatology. Volume 38:Number 8(2018)
• Page Start: 3273
• Page End: 3289
• Publication Date: 2018-04-01
• Subjects: climate change -- East Mediterranean -- homogenization -- Israel -- trend analysis -- temperature; Climatology -- Periodicals; Climat -- Périodiques; Climatologie -- Périodiques; 551.605
• DOI: 10.1002/joc.5500
• Languages: English
• ISSNs: 0899-8418

Abstract

Abstract : Climate trends analyses are studied through the analysis of long‐term records which usually are compromised by artificial non‐climatic factors (e.g., station relocation, instrumental replacements, etc). The impact of these factors on the analysis must be assessed and corrected in a procedure called "Homogenization, " before computing any trends. An unbiased analysis is essential for the East Mediterranean climate change, which suffers from scarcity of long and reliable datasets. Here, for the first time, we address these problems by jointly applying some of the state‐of‐the‐art homogenization methods, to long‐term Israeli temperature records (TX and TN) at five different meteorological stations throughout the period of 1950–2011. All of the studied time series were found to be inhomogeneous, where instrumentational issues were responsible for almost 50% of the breaks. The most frequent adjustments range between [−0.6, +0.6] o C while some larger adjustments do not fall within the range of the [−1, +1] °C interval. The adjustment of these breaks is crucial because they introduce large errors that may lead to wrong conclusions about the estimated trends. The difference in the seasonal and annual trends between raw and homogenized series was analysed applying the Mann–Kendall test. The general annual trend differences before and after homogenization, fell within the range of [−0.12, 0.16] o C/decade. Based on the homogenized dataset, a highly significant positive trend was found for the annual TN ¯ with 0.15 °C/decade ( p = .002) whereas the TX ¯ trend was 0.10 °C/decade ( p = .051). In general, the maximum temperature trends are lower and less statistically significant than those for minimum temperature. The most pronounced seasonal trends were recorded for the summer, which was characterized by significant positive trends for TX ¯ (0.15 °C/decade) and TN ¯ (0.23 °C/decade), while the winter had mainly no significant positive trends. Abstract : This study stresses the necessity of adjusting inhomogeneities prior to any climate analysis, in addition to studying its effect on the trends, in a region where previous climate studies were solely based on inhomogeneous data. Some clear dissimilarities between raw (red) versus homogenized (blue) data were found in Israel's temperature database. The figure shows an example of how from a zero trend for the TX raw data, to a substantial increase of 0.16 °C/decade is achieved, after an adjustment is made in one station. Zefat station's (WMO No. 401530), raw (dashed red line) versus homogenized (solid blue line).