Responses to Climate Change of Maximum Latewood Density from Larix speciosa Cheng et Law and Abies delavayi Franch. in the Northwest of Yunnan Province, China

: Tree-ring density has been used for climate-response analysis and climate reconstruction for many species. However, our knowledge of wood density for the responses of different species to climate remains very limited and inconclusive. To determine the relationship between maximum latewood density (MXD) and climate for deciduous and evergreen coniferous species, MXD chro-nologies were developed from Larix speciosa Cheng et Law and Abies delavayi Franch. growing at 3200–3300 m a.s.l. in Gongshan county, northwestern Yunnan, in China. Significant positive correlations with late summer mean temperature were found for the MXD chronologies of both species. However, the highest correlation occurred in August–September for L. speciosa (r = 0.551, p < 0.01) and in September–October for A. delavayi (r = 0.575, p < 0.01), which may be associated with the physiological habits of trees. Linear model can describe relationships between late-summer temperature and MXD index for L. speciosa (MXD = 0.0506T 8–9 − 0.0509, R 2 = 30.3%) and A. delavay (MXD = 0.0317T 9–10 + 0.4066, R 2 = 33.0%). The composite chronology from the two species can reveal a late summer temperature (August−October) signal with the explained variance 32.2% for its response model. However, in dry areas and or at high altitudes close to upper tree line, the responses of wood densities to climate require further investigation for deciduous and evergreen coniferous species.

Some studies have investigated how tree ring density varied with elevation [16,[47][48][49].The results indicated that the mean absolute MXD values varied with elevation in the central Scandinavian mountains [47], and a positive influence of warm summers on high elevation and a positive influence of summer precipitation on low elevation MXD in the Tatra Mountains in Slovakia [16].However, the influences of inner heredity from tree species cannot be ignored, although the growth increment and climate relationships of treerings are significantly more dependent on their site environments [50,51].The European project named GenTree (http://www.gentree-h2020.eu,accessed on 16 January 2021) tries to provide systematic evaluations of phenotypic and genomic variation among tree species under different circumstances for all of Europe [52].Only a few studies have investigated the inter-species comparison of wood density responses to climate, which appears to vary regionally.For example, in the dry conditions of northern Mexico, the inter-annual variation of the minimum density for Pinus lumholtzii B. L. Rob.& Fernald showed a stronger water availability signal because of its ability to inhabit drier sites compared to Pinus durangensis Martínez [53].Similar relationships with climate were found for the inter-annual variation in MXD of Abies veitchii Lindl.and Abies mariesii Mast.at 1900 m a.s.l. on Mount Norikura, central Japan [48].In China, MXD from L. gmelinii and Pinus sylvestris L. var.mongolica Litv.had a similar correlation with July−August mean temperature in the Da Hinggan Mountains [54].However, an understanding of the growth-climate-species interactions for tree-ring density is currently limited, and an investigation of the relationships between tree-ring density among species, in particular under the same growing conditions, is necessary.
The northwest of Yunnan Province is located in the south of the Hengduan Mountains.Abundant conifer species are present because of the complex terrain, which has significant potential for dendroclimatological studies [55].There are mixed forests of Larix speciosa Cheng et Law and Abies delavayi Franch.in the north of Gaoligong Mountain [56], which provides the same growing environment for an assessment of climatic sensitivity of tree-ring density from different species.The L. speciosa is a deciduous conifer species, which prefers to a sunny, warm, and well-drained environment at 2500-3800 m a s l altitude [57].Meanwhile, A. delavayi is an evergreen species mainly distributed in warm and shady slopes with elevation ranges from 3000 to 4300 m a s l. [58].
A warmer growing season is beneficial for tree-ring radial growth and thickening of the cell wall of L. speciosa in 3200-3300 m a.s.l., with positive correlations to May−August temperature for tree-ring width and to August−September temperature for its MXD [57,59].Different to L. speciosa, warm and dry air conditions in spring limit stomatal aperture and then reduce photosynthesis rates, slowing its radial growth of stem for A. delavayi [58].For instance, the annual basal area increment of A. delavayi above 3600 m a s l. showed a decreasing trend since the 1950s under the warming and drying climate [60].As lower relative humidity in spring makes light isotope of A. delavayi lost easily, the tree-ring stable oxygen isotope (δ 18 O) yields a negative response to spring relative humidity.However, the MXD of Abies delavayi var.motouensis in 4000 m a.s.l, which is a variant of A. delavayi, still strongly respond to August-September temperature with correlation coefficient of 0.79 [61].It is uncertain that their different climate responses of tree-ring growth were caused by species difference, because they had different growing environment.
Here, to compare their similarities and differences in response to climate, we collected tree ring cores from L. speciosa and A. delavayi in a mixed forest to develop MXD chronologies and conduct response analysis.The results will provide scientific information for future regional large-scale climate reconstruction studies for selected species.

Study Area
The study area was located in the southeastern Tibetan Plateau, in the Yunnan Province of Southwestern China.The Nujiang, Lantsang, and Jinsha Rivers traverse from north to south (Figure 1A).The floral distribution changes with elevation, and the vegetation is sub-alpine coniferous forest, such as Tsuga dumosa (D.Don) Eichler, L. speciosa, and A. delavayi at 2700-3500 m a.s.l.[57].According to the meteorological records of Gongshan (98.67°E, 27.75° N, 1595.7 m a.s.l.), Deqin (98.88°E, 28.45° N, 3488.0 m a.s.l.), and Weixi (99.28°E, 27.27° N, 2321.5 m a.s.l.) stations, the mean annual temperature is 10.6 °C, with a mean temperature 17.5 °C in July (the warmest month) and 3.1 °C in January (the coldest month).The mean annual precipitation is 1094.3mm, with two rainy peaks in March-April and June-August (Figure 1B).

Tree-Ring Data
Tree-ring increment cores from L. speciosa and A. delavayi were collected from a southeast-facing slope (98.477°E, 27.786° N, 3200-3300 m a.s.l.) in Gongshan county, northwestern Yunnan in November, 2012 and 2019.Clumping Fargesia praecipua Yi grows under the canopy (Figure 1C), resulting in less sunlight and relatively high humidity in the forest.The soil surface is covered with fallen leaves and branches, forming a relatively thick layer of litter.Ninety-six increment cores were sampled from 50 living Larix trees, and 108 cores were sampled from 55 living Abies trees, and 51 cores from 26 Larix trees and 51 cores from 27 Abies trees were selected for densitometric analyses (Table 1).These did not decay and were without spiral growth and scars [62].

Climate Data
The climate data from the Gongshan, Deqin, and Weixi stations were obtained from the China Meteorological Data Service Centre (https://data.cma.cn/,accessed on 5 January 2021), and included monthly mean maximum temperature (Tmax), monthly mean temperature (Tmean), monthly mean minimum temperature (Tmin), and monthly precipitation (P).In mountainous areas, the temperature and precipitation vary with the altitude [63].Therefore, the sampling site and its nearest station Gongshan may have a different climate due to a nearly 1600 m elevational difference.To find a better representation of climate at a sampling site, monthly temperature and precipitation data from the Climate Research Unit (CRU TS4.04, with the spatial resolution of 0.5° × 0.5° https://catalogue.ceda.ac.uk/, accessed on 31 December 2020) at two grid points (98.250°E/27.750°N and 98.750° E/27.750°N) near the sampling site were used for the growth responses analysis (Table 1).The CRU TS4.04 data were interpolated from meteorological station data.However, most station records started after 1950 in China, meaning this dataset span 1951-2019 has a higher quality than data from 1901-1950 [64].Thus, we only analyzed climate responses during 1951-2019.C/T 1 : cores/trees of sampling, C/T 2 : cores/trees for densitometric analyses.

Chronology Development
The tree-ring cores were processed according to the standard procedures for densitometric analysis [65] and were visually cross-dated with a microscope.The TRW in each year was measured using the LINTAB measurement system, with a precision of 0.01 mm.The examination and validation of raw TRW were executed by the COFECHA program [66].The tree-ring cores were then cut into subsections at lengths of 3-4 cm.The subsections were mounted on support blocks and the wood fiber angles were measured using a Dendroscope and cut into 1.0 mm slices with a twin-bladed Dendrocut.The resin was extracted after 72 h with a water bath of 80 °C.The slices were kept at a constant room temperature of 20 °C and 50% relative humidity for 12 h before X-ray photography was taken.The grey-scale values of X-ray film were measured by the DENDRO-2003 tree-ring workstation.The density cross-dating was determined by comparing the consistency of the variations of raw tree-ring width series, which were measured by the DENDRO-2003 densitometer and the LINTAB measurement system [45].
The MXD chronologies of L. speciosa and A. delavayi were developed by using the ARSTAN program [67].Each series of tree-ring MXD measurements was fitted with a 67% cubic smoothing spline at a 50% cutoff frequency to remove the non-climatic trends from the age, size, and effects of stand dynamics [68].Each value in the detrended series was calculated as the ratio of tree-ring original value to the corresponding spline curve value in a given year.Index series of all samples were averaged to form a mean chronology using a robust mean value function [67].All the measurements of the two species were entered into a file and then run with the ARSTAN program to develop the composite chronology using the same detrending and standardizing method.The mean correlation coefficient between trees (Rbt), signal-to-noise ratio (SNR), and expressed population signal (EPS) was used for a statistical characteristic analysis of chronology.The Rbt represents the chronology signal strength, the SNR expresses the strength of the observed common signal among the trees [48], and EPS indicates the representativeness of a subsample to the entire sample.The threshold of 0.85 determines the trusted period of the chronology [69].Standard chronologies (STD) and residual chronologies (RES) were utilized for comparing climate signals in different chronologies and determining which was more appropriate for climate-growth analysis.

Climate-Growth Response Analysis
The Pearson's correlation coefficients of MXD and the climate variables (Tmin, Tmean, Tmax, and P) from September of the previous year to October of the current year were calculated.To test whether the correlation with climate was influenced by variations in the low-frequency domain of MXD chronologies, the correlation coefficients between the chronologies and the climate data after the first order differences were calculated.A spatial field correlation analysis was performed between the MXD chronologies and the gridded climate data.A hyperbolic tangent function (Z) and test statistic (µ) were induced to estimate the significance of the differences in the correlations with contemporaneous climate variable between species [70].A correlation with climate during the same period from different species was considered to be identical when the absolute value of µ was lower than 1.96.A linear regression model was developed to estimate specific statistical relationships between MXD and climate parameters, and the slopes were used to estimate the responsive sensitivity to climate.

Statistical Parameters of Chronologies
The 519-year MXD chronology of L. speciosa covered the time period 1501-2019 A.D. (Figure 2A) and the 447-year MXD chronology of A. delavayi was for the time period 1573-2019 A.D. (Figure 2B) (Table 2).The higher first-order autocorrelation (AC1) of L. speciosa STD chronology indicated that L. speciosa had more climate memory than A. delavayi.Beginning in 1638 (1580) for the Larix STD (RES) chronology, the series could be considered reliable with sufficient numbers of samples because the EPS reached 0.85 with 18 (13) cores (Figure 2A,D,H,I).The STD (RES) chronology of A. delavayi began in 1732 (1678) with 23 (13) cores when the EPS exceeded 0.85, which was later than that of the L. speciosa STD (RES) (Figure 2B,E,H,I).The Rbt of both species ranged from 0.2-0.3,showing low common signals between trees.The mean sensitivity of the two MXD chronologies was 0.035 and 0.05, indicating small differences in MXD of two adjacent rings.This was similar to the results of other studies in this region [59,69].In addition, the RES of MXD for the two species showed a higher signal-to-noise ratio (SNR) compared with STD, indicating more climatic signal in the RES chronologies.The composite chronologies showed a high AC1 for STD and a low ACI for RES, which was the same as for L. speciosa.The EPS of composite chronology reached 0.85 in 1669 for STD and 1637 for RES with 35 and 20 cores, respectively (Figure 2C,H-K) (Table 2).However, the composite chronology displayed a lower Rbt and the variance explained by the first eigenvector (PC1) than each individual species chronology, which may have resulted from the mixture of different species MXD.However, the composite chronology still retained a strong climate signal according to their high SNR values.
Similarities between low-frequency MXD series (15-year moving average) for the two species were apparent at the decadal scale (Figure 2A

Responses of Abies and Larix MXD to Climate
Figure 4 shows the results of the climate response analysis.The L. speciosa MXD displayed positive correlations with the mean temperature during August-September (r = 0.551, p < 0.01) (Figure 4A,D,G).It also was more sensitive to mean maximum temperature (r = 0.499, p < 0.01) than minimum temperature (r = 0.369, p < 0.01) from August to September.However, L. speciosa MXD showed a negative response to the current September precipitation (r = −0.273,p < 0.05) (Figure 4J).For A. delavayi MXD, it showed positive correlations with mean temperature of April (r = 0.286, p < 0.05) and September-October (r = 0.575, p < 0.01) (Figure 4B,E,H).In addition, A. delavayi MXD positively correlated with the precipitation in current February (r = 0.289, p < 0.05), and negatively related to precipitation in April (r = −0.387for first difference, p < 0.05) (Figure 4L).Although similar correlations with mean temperature during late summer for both species, the strongest responsive time was August-September for L. speciosa while September-October for A. delavayi.These were also statistically significant for the first-difference data for both species, respectively (r = 0.538 and r = 0.719, p < 0.01).The results indicated that there was a difference about the response time between the two species.The absolute values of statistic (µ) of less than 1.96 during the growing season also suggested the accordance of responses to temperature for the tree-ring density of the two species (Figure 4E,F,I,L).

Statistic Model of Climate-Growth Relationships for MXD
Despite the different response seasons from the MXD of the two species, there was similar climatic sensitivity between them.For L. speciosa MXD, its relationship to August-September mean temperature was more suitable to a linear model, with MXD changes of 0.0506 sigma unit per centigrade (Figure 5).While its distribution to September-October mean temperature is relatively discrete, only with explained variance 14.1% for a linear fit.For A. delavayi MXD, the linear model described its relationships to September-October mean temperature better, with a slope of 0.0317 sigma unit per centigrade.Composite chronologies also retained a late summer (August-October) temperature signal, with the explained variance 32.2% of the response model.T8-9: mean temperature of August-September, T9-10: mean temperature of September-October, and T8-10: mean temperature of August-October.

Tree Growth-Climate Relationship
The strong response to late-summer temperature of tree growth revealed the MXD characteristic, reserving climate information of the growing season during the process of cell formation [16,44,71,72].It indicated that the wall thickness of latewood cells was likely influenced by temperature on high altitude mountains, because high-temperature conditions favored more deposited cell wall material [17,73].In other alpine areas of the southeastern Tibetan Plateau, the MXD of coniferous species was also affected by late summer temperature [12,13,69].
However, the formation time of maximum latewood density for L. speciosa was earlier by one month than A. delavayi, which may be associated with the physiological habit of trees.A. delavayi is an evergreen coniferous species, for which the optimum temperature for photosynthesis is 10-25 °C.However, L. speciosa is a deciduous tree species, and its optimal photosynthetic temperature is 15-25 °C [74].The maximum temperature of this site in October ranges from 6.5 to 14.7 °C, which is better for A. delavayi.Previous studies have shown that the most temperature-related month of MXD for L. gmelinii is earlier than Pinus sylvestris var.mongolica [54].The MXD chronologies of the two species correlated differently to other reconstructions in nearby areas.For the correlation with the reconstructed July-October mean temperature in Yulong County, the L. speciosa MXD chronology was lower than A. delavayi (Figure 6A-D), which may be induced by the evergreen coniferous species spruce used in the reconstruction [69].Furthermore, the climates of Gongshan and Yulong are similar, which may be another reason.The nearest station data may not well represent the climate of the mixed forest due to an elevational difference.Therefore, CRU TS4.04 data were employed to analyze annual variability of two species MXD and late-summer temperature (Figure 6E-G).The MXD of A. delavayi and Picea asperata Mast.were less sensitive to climate warming in the recent twenty years than L. speciosa.A non-significant correlation was found between the MXD chronologies of the two species and the reconstructed August-September mean temperature in Sygera Mountain using mean latewood density because the sampling sites were located in different climate regions [12].The MXD chronology of A. delavayi correlated significantly with the current April Tmax and Tmean (Figure 4), while L. gmelinii did not.Studies indicate that nonstructural carbohydrates, produced both in the early and late growing season, are needed for the formation of latewood cell walls, because there were not enough photosynthates to form tissues in late summer [75].L. gmelinii is just beginning to bud in April, and therefore, it cannot perform photosynthesis to produce photosynthate.Conversely, being an evergreen conifer species, A. delavayi has a stronger photosynthesis and accumulate more carbohydrates.
A negative correlation was also found between MXD chronology and precipitation for L. speciosa in the current September because of lower temperatures on rainy days [69].This can be confirmed by the negative coefficient of mean temperature and precipitation as −0.416 (CRU TS 4.04).The partial correlation coefficients of RES with precipitation for L. speciosa was −0.091 (p > 0.05), which indicated that temperature is an important factor for MXD.The negative correlation that was found between MXD chronology and precipitation for Abies in April was because the high moisture content in soil can cause enlargement of earlywood cells [17,76,77], resulting in less carbohydrates reserved for the growth of latewood cells.In addition, the positive correlation between A. delavayi and precipitation in February could be caused by increasement in snow depth.For one hand, the snow cover acts as a reservoir for soil and trees during the growing season [78].For the other, months with more precipitation have a high relative humidity.A wet condition in February can prevent the reduced leaf photosynthesis due to an excessive transpiration [58]. A. delavayi normally photosynthesize with foliage to accumulate carbohydrate for later growth [79,80].

Potentials and Limitations of MXD for Different Species for Climate Reconstruction
The composite chronology not only retained the August temperature information of L. speciosa, but also contained October temperature signals of A. delavayi.The spatial correlation reflected that MXD for both L. speciosa and A. delavayi contained the August-October mean temperature signal in the southeastern Tibetan plateau (Figure 7).Therefore, if there are insufficient replications for one site, developing composite chronologies of different species is suggested, which can reflect the climate response of the tree-ring density for an entire forest well.Different results would be obtained if sampling sites were in arid areas, whereby, the xylem growth for different species may respond to drought differently.On the lowest elevation of the southern Black Forest, drought may lead to cell differentiation of Fagus sylvatica L. to stop earlier and damage growth more severely compared to Abies alba Mill and Pinus sylvestris L. [81].The minimum wood density of P. lumholtzii is more sensitive to drought than P. durangensis in Sierra Madre Occidental because P. lumholtzii can inhabit drier sites [53].Therefore, different responses to climate for the wood density of deciduous and evergreen coniferous species may occur under drought conditions because of their physiological properties.This issue requires further investigation.

Conclusions
In this study, an assessment of the climate signals of MXD data for L. speciosa and A. delavayi growing under the same conditions in the northwest of Yunnan Province, was presented.It was found that the MXD chronologies of two species were positively responsive to late summer mean temperature.However, there were differences in the response span to climate between their MXD chronologies.The most sensitive period was August-September for L. speciosa, and September-October for A. delavayi., which may be associated with the deciduous properties of trees.While composite MXD chronology from two species in moister areas can reveal signal of the August-October mean temperature, which meant that tree-ring density from different species in moister areas could produce a new series for reconstruction.Meanwhile, the question of whether their wood densities respond to climate differently requires further verification for deciduous and evergreen coniferous species under dry conditions.Furthermore, our sampling site is near the center of forest ecological amplitude, where good environmental conditions limit the response of tree growth to climate.Tree-ring growth in a mixed forest close to upper tree line is more sensitive to temperature, and there may be a different responsive amplitude between species.Some low-frequency differences occurred in 1735-1770, 1829-1882, and 1938-1964 for L. speciosa and A. delavayi.More evidence is needed to verify that the same difference exists in other tree species.There is also a question to answer that how to qualitatively and quantitatively estimate the impact of multiple-species origin on climate reconstruction and how to eliminate such influence.Institutional Review Board Statement: not applicable.
Informed Consent Statement: not applicable.

Figure 1 .
Figure 1.Geography of the study area.(A) The location of sampling site within the northwestern Yunnan, China region.(B) Climate diagram of the mean values of the meteorological stations in Gongshan, Deqin, and Weixi.P, monthly precipitation; Tmax, monthly mean maximum temperature; Tmin, monthly mean minimum temperature; Tmean, monthly mean temperature.(C) Images of the L. speciosa and A. delavayi mixed forest.

Figure 2 .
Figure 2. MXD Chronologies of L. speciosa and A. delavayi.(A-C) The standard chronologies.(D-F) The residual chronologies.(G) The sample depth of chronologies.(H-K) The running inter-series correlation (rbar) and the expressed population signal (EPS) were centered on a 50-year window with a 25-year step length.The dashed line is 0.85 cutoff.The gray shaded vertical areas indicated the low-index intervals of 15-year moving average of chronologies.

Figure 3 .
Figure 3. Moving correlations centered on a 50-year window with a 1-year step length between L. speciosa MXD and A. delavayi MXD during 1701-2019.

Figure 4 .
Figure 4. Correlation between MXD of different tree species and CRU TS 4.04 climatic variables during different periods (1951-2019).Correlation between (A) Tmax, (D) Tmean, (G) Tmin, (J) P, and RES chronologies of L. speciosa.Correlation between (B) Tmax, (E) Tmean, (H) Tmin, (K) P, and RES chronologies of A. delavayi.Significance of the different correlations with (C) Tmax, (F) Tmean, (I) Tmin, and (L) P between the MXD of the two species.The absolute values of statistic (µ) less than 1.96 (dashed line) indicates the same correlation with climate for MXD of the two species.

Author
Contributions: methodology, G.D.; validation, G.D.; formal analysis, G.D.; resources, M.L. and X.S.; data curation, G.D.; writing-original draft preparation, G.D.; writing-review and editing, M.L. and Z.H.; visualization, G.D.; supervision, M.L.; project administration, M.L.; funding acquisition, M.L.All authors have read and agreed to the published version of the manuscript.Funding: This research was funded by National Natural Science Foundation of China (grant number 41977391, 41630529 and 41571194).

Table 1 .
Details of sampling site, meteorological stations, and grid points in the northwestern Yunnan Province.