Daily climate data reveal stronger climate-growth relationships for an extended European tree-ring network

Highlights

• Day-wise aggregated correlations were on average higher by 0.071

• Greater differences were calculated for precipitation and SPEI data.

• Day-wise aggregated correlations on average use fewer days than monthly.

• Correlations from daily and monthly approaches are not significantly different.

Abstract

An extended European tree-ring network was compiled from various sources of tree-ring data from Europe, northern Africa and western Asia. A total of 1860 tree-ring chronologies were used to compare correlation coefficients calculated with aggregated day-wise and month-wise mean temperature, sums of precipitation and standardised precipitation-evapotranspiration index (SPEI). For the daily approach, climate data were aggregated over periods ranging from 21 to 365 days. Absolute correlations calculated with day-wise aggregated climate data were on average higher by 0.060 (temperature data), 0.076 (precipitation data) and 0.075 (SPEI data). Bootstrapped correlations are computationally expensive and were therefore calculated on a 69.4% subset of the data. Bootstrapped correlations indicated statistically significant differences between the daily and monthly approach in approximately 1% of examples. A comparison of time windows used for calculations of correlations revealed slightly later onset and earlier ending day of the year for the daily approach, while the largest differences between the two approaches arise from window lengths: Correlations calculated with day-wise aggregated climate data were calculated using fewer days than the monthly approach. Differences in the onset and ending dates of periods for the daily and monthly approaches were greater for precipitation and SPEI data than for temperature data.

Introduction

In dendroclimatology, various tree-ring proxies are usually compared to gridded or observed station climate data with monthly resolution to analyse climate-growth relationships (Cook and Kairiukstis, 1992). Monthly climate data are more easily accessible, available for most land territories and have longer time spans than daily data, but at the cost of accuracy, particularly when dealing with precipitation data (Hofstra et al., 2009; Yin et al., 2015). All monthly data, whether they are gridded or from station observations, are derived from daily climate station observations, which are the raw climate products, and then aggregated into monthly datasets. In addition to the many daily observations available from the KNMI Climate Explorer (https://climexp.knmi.nl/start.cgi), various reforecast project collaborations have resulted in high quality gridded daily data, such as E-OBS gridded daily datasets for Europe (Cornes et al., 2018), Berkeley Earth temperature datasets (http://berkeleyearth.org) and various datasets provided by the National Oceanic and Atmospheric Administration of the United States (https://www.esrl.noaa.gov/psd/data/gridded/tables/daily.html).

Daily climate data is well integrated into various process-based models, such as the VS model (Anchukaitis et al., 2006; Shishov et al., 2016) and MAIDENiso (Danis et al., 2012). Some previous dendroclimatological studies have used daily climate data. Land et al. (2017) reported increasing correlations between ring widths and precipitation if heavy precipitation events are excluded from the precipitation data. Their study showed that the annual radial growth of oak trees is mainly affected by daily precipitation sums of less than 10 mm. Schönbein et al. (2015) reconstructed summer precipitation based on subfossil oak tree-ring data and daily precipitation records from southern Germany, while Pritzkow et al. (2016) combined the earlywood vessel area of Quercus robur and daily temperature data from northern Poland to reconstruct minimum winter temperatures back to 1810. Climate-growth relationships using daily climate data have been calculated by various authors (e.g. Castagneri et al., 2015; Liang et al., 2013; Sanders et al., 2014; Sun and Liu, 2016). One of the first software programs for dendroclimatological studies based on daily climate data was CLIMTREG, provided by Beck et al. (2013), while Jevšenak and Levanič (2018) recently presented the dendroTools R package, which is designed for the R environment (R Core Team, 2019) and provides various options for analysis of climate-growth relationships on daily and monthly scales.

Combining tree-ring networks with gridded climate data can provide comprehensive spatio-temporal information related to tree growth and climate sensitivity. Compiled large-scale tree-ring networks have already been used for various purposes, e.g. to analyse climate-growth associations for northern hemisphere tree-ring width records (St. George, 2014); to evaluate the climate sensitivity of model-based forest productivity estimates (Babst et al., 2013); to identify climatic drivers of global tree growth (Babst et al., 2019); to characterise relationships between climate, reproduction and growth (Hacket-Pain et al., 2018); to simulate radial tree growth with the VS-Lite model on a global scale (Breitenmoser et al., 2014); to assess global tree-mortality (Cailleret et al., 2017); and to quantify the drought effect on tree growth as a measure of vitality (Bhuyan et al., 2017). Zhao et al. (2019) analysed representatives and biases of tree-ring records in the Global Tree-Ring Databank (ITRDB), identified priority sampling areas and corrected identified issues, while Babst et al. (2018) discussed challenges and opportunities related to tree-ring networks. No tree-ring network has so far been used to analyse climate-growth relationships for daily data and to compare daily and monthly climate-growth relationships. To do so, an extended European tree-ring network was established using freely available data from various sources and combining these data with gridded daily climate data, i.e. E-OBS daily data on a 0.1-degree regular grid.

In this study, I compare climate-growth correlations calculated from aggregated daily and monthly data of mean temperature, sums of precipitation and standardised precipitation-evapotranspiration indices (SPEI). Climate data with daily resolution enable greater flexibility in the analysis of climate-growth relationships and provide higher explained variance in calibration models for climate reconstructions. In areas where the time period related to the climate signal starts/ends near the 15th day of the month, a daily approach should provide significantly greater differences between correlations calculated from day-wise and month-wise aggregated climate data. An important benefit of using a daily approach is the possibility to study changes in time windows over time. While the temporal stability of monthly data usually enables the study of only the changes in correlation coefficients over time, a daily approach enables the study of changes in temporal windows over time as well. Hypothetically, this information could be used to model the divergence of climate-growth relationships (Loehle, 2009) and changes in growing season patterns (Linderholm, 2006). Finally, studying climate growth correlations using day-wise aggregated climate data could improve our understanding of the climate signal in tree rings and enable us to more accurately predict future growth under different climate scenarios. The goal of this study is to highlight the advantages of using daily rather than monthly data and, at the same time, expose possible caveats related to the daily approach.

The paper is structured as follows: in section 3.1 I give a general description of the extended European tree-ring network, while correlations calculated with day-wise and month-wise aggregated climate data are compared in sections 3.2 Comparison of correlations calculated with day-wise and month-wise aggregated climate data, 3.3 Comparison of bootstrapped correlation coefficients. The time periods related to the calculated correlation coefficients for the daily and monthly approach are compared in section 3.4. Finally, in section 3.5 the potential applications and future extensions of the daily approach are discussed. In the conclusions the main results are summarised, and possible caveats of the daily approach are discussed.