不同去趋势方法对树轮气候信号识别的影响

树木生长受到气候因子、随年龄增长的内在生长趋势、环境干扰和其他扰动信号的影响。目前存在不同的去趋势方法对树木年轮进行去趋势以识别树木生长中的气候信号。以往的研究多基于单个方法识别树轮气候信号,而不同去趋势方法识别的树轮气候信号可能会有一定的差别。为了对比不同去趋势方法对树轮气候信号识别的影响,我们基于国际年轮数据库网站获取中国西部地区68个点的树轮宽度数据,采用最常用的"signal-free"方法（SsfCrn）、线性和负指数函数法（std）、67%样条函数法（spline）、firedman方法、以及基于经验模式分解去趋势方法（EEMD）5种去趋势方法分别建立树轮年表,并对比分析同一地点的不同年表对气候响应的异同。结果表明：不同去趋势方法得到的年表对温度、降水以及相对湿度等气候因素的响应具有明显差异。其中,SsfCrn去趋势方法建立的年表对温度（月平均温、月最低温、月平均最低温）响应中相关最高的样点在所有样点中占比最高;EEMD去趋势方法建立的年表对降水量、相对湿度和月最高温响应中相关最高的样点在所有样点中占比最高;firedman去趋势方法建立的年表对月平均最高温响应中相关最高的样点在所有样点中占比最高。研究结果表明SsfCrn,EEMD和firedman方法在识别树轮气候信号方法具有一定的优势。在不同研究区域中,不同去趋势方法建立的年表对不同气候条件响应有差异,因此选择不同的去趋势方法识别树木生长趋势,分析哪种方法可以更好的反应气候变化对树木生长的影响显得尤为重要。

Influences of different detrending methods on identifying the climate signals from tree-ring analysis

Tree radial growth is influenced by climate factors such as temperature, precipitation and relative humidity, its intrinsic growth trend (increments of age), environmental disturbances, and unexplained variability of other signals. Tree ring chronologies were obtained using detrending method to remove the intrinsic growth trend of trees to retain climate signals in tree rings. Currently, there are many detrending methods which can be used to develop tree ring chronologies. Climate signals from tree rings were mainly identified using one detrending method in most previous studies, however, the identified climate signals can be different from the changes in detrending methods. To compare influences of different detrending methods on identifying the climate signals from tree ring analysis, we obtained the width data of the tree ring of 68 sites in the western region of China from the international tree ring database. The commonly used "signal-free" method (shorted as SsfCrn), linear and negative exponential function method (shorted as std), 67% spline function method (shorted as Spline), firedman method, and ensemble empirical mode decomposition-based (shorted as EEMD) detrending method were used to establish tree ring chronology, respectively. The similarities and differences of how tree growth responses to climate with different tree ring chronologies were compared. The results showed that there were significant differences in how trees responded to climate factors such as temperature, precipitation and relative humidity, based on the tree ring chronology obtained by different detrending methods. The chronology had higher correlation with temperature (monthly average temperature, monthly lowest temperature, and monthly average lowest temperature) when using SsfCrn detrending method than other detrending methods in most sites of the study region. The chronology had higher correlation with precipitation, relative humidity, and monthly maximum temperature when using EEMD-based detrending method than other detrending methods in most sites of the study region. The chronology had higher correlation with average monthly maximum temperature when using firedman detrending method than other detrending methods in most sites of the study region. Our results suggested that the chronologies developed with SsfCrn, EEMD and firedman methods are more suitable to detect growth-climate relationship than other methods. Tree ring chronologies obtained with different detrending methods responded differently to the same climatic factors and the same tree ring chronology had different responses to different climatic conditions in different research areas. Therefore, choosing proper detrending methods to identify tree growth trends is particularly important to detect the influences of climate changes on tree growth.