Comparison between chironomid-inferred mean-August temperature from varved Lake Żabińskie (Poland) and instrumental data since 1896 AD

• Chironomids in 130 samples were analyzed to reconstruct temperature since 1896 AD.
• Chironomid-inferred temperatures were very similar to the instrumental data.
• 83% of the chironomid-inferred temperatures had differences with instrumental data < RMSEP.
• Chironomids can reconstruct both the pattern of temperature changes and its amplitude.

Chironomids preserved in varved Lake Żabińskie (54°07′54.5″ N; 21°59′01.1″ E; 120 m a.s.l), northeastern Poland, were used to reconstruct mean-August air temperature since 1896 AD at annual (1949–2011 AD) and at 3–4 year resolutions (1896–1948 AD). This is one of very few studies using chironomids at such high temporal resolution, for a total of 130 sediment samples analyzed. To infer temperature a combined (Northeastern Canada and Poland) transfer function was developed. This transfer function had 112 lakes (50 from Poland, 72 from Canada) and 95 taxa. The mean-August-air temperature gradient was 23.5 °C. The coefficient of determination (r2jack) was 0.88, the root-mean-square error of prediction (RMSEP) was 1.30 °C and the maximum bias was 2.08 °C.The average mean August air temperature measured during the 1961–1990 AD reference period was 16.3 °C and the chironomid-inferred average temperature for the same period was 16.6 °C. For the 1981–2010 AD reference period, the temperatures were 1 °C warmer (i.e. measured = 17.3 °C; inferred = 17.6 °C). The differences of only 0.3 °C between the measured and the inferred temperatures during the reference periods were a first indication that the merged transfer function could provide accurate estimates of mean August air temperature.Chironomid-inferred mean August air temperatures since 1896 AD had strong relationships with instrumental data at near-annual resolution (rPearson = 0.74, pcorr < 0.01). The relationship increased on decadal scale (rPearson = 0.91, pcorr < 0.01) when the errors in varve chronology (−5 to 2 years) were considered.Although the RMSEP of the transfer function was higher than most changes recorded, the average difference between the instrumental and the inferred mean August air temperatures was 0.75 °C. The results of this study suggest that not only do chironomids reconstruct the right pattern in temperature changes (83% of the inferences recorded the same temperature increase or decrease than the instrumental data) they can also infer changes with the right amplitude (61% of the inferences had differences with the instrumental data below 0.7 °C). This study is one more example of the possible accuracy of chironomids to reconstruct climate, at least for the past 100 years and at this site.