Dendroclimatic transfer functions revisited: Little Ice Age and Medieval Warm Period summer temperatures reconstructed using artificial neural networks and linear algorithms

Abstract

Abstract. Tree-rings tell of past climates. To do so, tree-ring chronologies comprising numerous climate-sensitive livingtree and subfossil time-series need to be "transferred" into palaeoclimate estimates using transfer functions. The purpose of this study is to compare different types of transfer functions, especially linear and nonlinear algorithms. Accordingly, multiple linear regression (MLR), linear scaling (LSC) and artificial neural networks (ANN, nonlinear algorithm) were compared. Transfer functions were built using a regional tree-ring chronology and instrumental temperature observations from Lapland (northern Finland and Sweden). In addition, conventional MLR was compared with a hybrid model whereby climate was reconstructed separately for short-and long-period timescales prior to combining the bands of timescales into a single hybrid model. The fidelity of the different reconstructions was validated against instrumental climate data. The reconstructions by MLR and ANN showed reliable reconstruction capabilities over the instrumental period (AD 1802(AD -1998. LCS failed to reach reasonable verification statistics and did not qualify as a reliable reconstruction: this was due mainly to exaggeration of the low-frequency climatic variance. Over this instrumental period, the reconstructed low-frequency amplitudes of climate variability were rather similar by MLR and ANN. Notably greater differences between the models were found over the actual reconstruction period (AD 802-1801). A marked temperature decline, as reconstructed by MLR, from the MeCorrespondence to: S. Helama (samuli.helama@helsinki.fi) dieval Warm Period (AD 931-1180) to the Little Ice Age (AD 1601-1850), was evident in all the models. This decline was approx. 0.5 • C as reconstructed by MLR. Different ANN based palaeotemperatures showed simultaneous cooling of 0.2 to 0.5 • C, depending on algorithm. The hybrid MLR did not seem to provide further benefit above conventional MLR in our sample. The robustness of the conventional MLR over the calibration, verification and reconstruction periods qualified it as a reasonable transfer function for our forest-limit (i.e., timberline) dataset. ANN appears a potential tool for other environments and/or proxies having more complex and noisier climatic relationships.

A palaeotemperature record for the Finnish Lakeland based on microdensitometric variations in tree rings

“…The wamth around the turn of the first millennium AD was seen closely synchronous with the MXD and TRW evidence from northern Fennoscandia (Grudd, 2008;Esper et al, 2012) and northern Finland (Helama et al, 2009a(Helama et al, , 2009b(Helama et al, , 2010a(Helama et al, , 2010b(Helama et al, , 2012b. The northern MXD (Esper et al, 2012) and TRW (Helama et al, 2010a) reconstructions showed their warmest 250-year periods in AD 816-1065 and AD 932-1181, respectively.…”

“…Long instrumental summer temperature records were also available in adjacent areas from St. Petersburg (Jones and Lister, 2002) 1801-1804, 1992-1994 and 2001. Summer temperature reconstructions from northern Finland and northern Fennoscandia were available over the study period, based on tree-ring width (Helama et al, 2009a(Helama et al, , 2009b(Helama et al, , 2010a and MXD data (Esper et al, 2012), respectively. These records were filtered using a spline function (Cook and Peters, 1981) with the rigidity of 50-years (50% frequency cut-off) and compared with the new palaeoclimate evidence by visual inspection and Pearson correlations (r).…”

“…Tree-ring chronologies and temperature time-series provided the independent and dependent data, respectively. Tree-ring chronologies often contain high autocorrelation (Fritts, 1976) exceeding the autocorrelation observed in climatic time-series and, consequently, the transfer function may benefit from using tree-ring values of several years (t -n, t, t + n) as independent variables in regression, in order to estimate the climate in year (t) (Helama et al, 2009a). Here we used multiple linear regression with tree-ring values in years t -1, t, and t + 1.…”

“…Standardizing the MXD data and calibrating the obtained MXD chronologies against the instrumental summer temperature records yield a new palaeotemperature curve that helps us to throw light on late Holocene climate evolution in the district where no such reconstruction yet exist. Contrasting the new palaeoclimate information with existing tree-ring based summer temperature reconstructions from northern Finland and Fennoscandia (Helama et al, 2009a(Helama et al, , 2009b(Helama et al, , 2010aEsper et al, 2012) enables the north-south comparison of temperature variations on annual resolution. It outlines the spatial palaeoclimate variations in the region where previous palaeoclimate research have highlighted notable climatic shifts through the Medieval Climate Anomaly to Little Ice Age (Grudd et al, 2002;Helama et al, 2005bHelama et al, , 2009aHelama et al, , 2009bHelama et al, , 2010aGunnarson et al, 2011;Esper et al, 2012) during the past thirteen centuries.…”

“…These chronologies originate from regions near the timberline where tree growth is predominantly governed by summer temperature. In consequence of their cold habitat, the tree-ring width chronologies provide excellent proxy records of past temperature variability (Grudd et al, 2002;Helama et al, 2002Helama et al, , 2009aHelama et al, , 2009bHelama et al, , 2010aHelama et al, , 2010bHelama et al, , 2012bLinderholm and Gunnarson, 2005). In many regions, however, an improved dendroclimatic signal of summer temperature can be attained using X-ray based microdensitometric series of maximum latewood densities (MXD), rather than using annual widths of the rings (Briffa et al, , 2002a(Briffa et al, , 2002b.…”