Abstract: Background, aim, and scope Due to high resolution, accurate dating and extensive geographical distribution, tree rings are playing important roles in paleoclimate and ecological studies. There are three proxies of tree ring, including tree-ring width (TRW), tree-ring stable isotopes and tree-ring density. Various climatic information can be obtained from these independent tree-ring proxies. In subtropical China (SC), TRW is generally limited by temperature and therefore used to reconstruct temperature history, while tree-ring δ18O (δ18Otree) is proved to be an ideal indicator of hydroclimatic variation. Previous study on single tree-ring proxy (TRW or δ18Otree) in Guilin, subtropical southwest China failed to be used for climate reconstruction owing to the weak climatic signals contained in tree rings, which may cause a waste of resources. Therefore, it is urgent to find a better solution. This paper attempts to explore whether combing TRW and δ18Otree will help to increase the climate signals in tree rings. Materials and Methods Based on previous work, TRW and δ18Otree records of Pinus massoniana Lamb. during 1939–2014 in Guilin are obtained. Pearson correlation analysis is utilized to calculate the relationship between tree-ring index and hydroclimatic factors, i.e. relative humidity (RH) and standardized precipitation-evapotranspiration index (SPEI). Multiple linear regression method, using TRW and δ18Otree as independent variables, is adopted to determine the most suitable climatic factor and period for future reconstruction. The stability of the regression model is evaluated by split calibration-verification test. The statistical parameters include correlation coefficient (r)，explained variance (R2 ), reduction of error (RE) and coefficient of efficiency (CE). Results Simple correlation analysis revealed that TRW significantly correlated with the June-October mean SPEI (r=0.32, p<0.05), while δ18Otree indicated significantly negative relationship with the mean SPEI from current May to current October (r=-0.61, p<0.01). TRW also showed significant correlation with the previous February-November mean RH (r=0.55, p<0.01) and the current April-July mean RH (r=0.49, p<0.01). δ18Otree indicated high correlation with the August-October mean RH, with r value of -0.52 (p<0.01). Discussion April-November is the growing season of Pinus massoniana Lamb. in Guilin, the amount of effective soil moisture during these months is vital to tree growth. Therefore, it is reasonable that growing-season hydroclimatic variation directly influences the TRW and δ18Otree records. However, single tree-ring proxy is not enough to be successfully used for hydroclimatic reconstruction so far, even though the TRW and δ18Otree records are sensitive to hydroclimatic variation. It is worth noting that the results of multiple linear regression model indicated much high r values and explained variance (R2 ) when combing TRW and δ18Otree together. The r value and R2 for modeling the May-November mean SPEI was 0.667 (p<0.01) and 44.6%, respectively, and for modeling the April-October mean RH was 0.636 (p<0.01) and 40.4%, respectively. The r value and R2 of the multi linear regression models greatly enhanced compared with the simple correlation analysis that based on single proxy. Moreover, the multi-proxy-based reconstruction tracks the instrumental hydroclimatic data very well. The reconstructed May-November mean SPEI has good skill in simulating the hydrological variation in a large field around the studying site. Conclusions It shows that the multiple linear regression model based on TRW and δ18Otree is more suitable for hydroclimatic reconstruction (SPEI, RH) in Guilin than that based on single tree-ring proxy. In comparison, May-November mean SPEI is a better choice for reconstruction. Recommendations and perspectives This study provides a new way for the future dendroclimate reconstruction in the subtropical regions of China, which will be a good guidance for climate investigation and forest management.
Keywords: Pinus massoniana Lamb.; Multiproxy; Tree-ring width; δ<sup>18</sup>O; Hydroclimatic reconstruction