Evaluation of spatial and temporal variation in fine root dynamics in a temperate mixed forest using a scanner method

ABSTRACT Understanding fine root phenology at the stand scale is crucial for elucidating how carbon and nutrient cycling in forest systems respond to climate change. This study aimed to reveal the spatio-temporal variations in the fine root phenology in a mixed temperate forest in Japan using scanner method. The spatio-temporal variation in the fine root areas was evaluated using two scales: among four plots in the stand and among four partitioned areas in the scanner image. Here, we hypothesized that root phenology would vary on both scales due to the mixture of species-specific phenologies, which means endogenous factors to have a larger impact than exogenous factors such as temperature and precipitation. The timing of the production peak varied among years, though it concentrated within a specific period of the year in all plots. In all plots, active root mortality dynamics were observed in the summer. These results suggested exogenous factors as stronger regulators of fine root production and mortality than endogenous factors. Root phenology within the scanner images is highly heterogeneous, suggesting the significance of broad observation surfaces, such as those acquired using the scanner method, in comprehending the representative phenology of root dynamics at a stand scale. This study revealed that fine root phenology had a synchronized pattern on the stand scale, even though high spatial variation existed in the scale size of the scanner. Differences in root phenology that were influenced by internal factors of the species were masked on the stand scale.


Introduction
Fine roots, defined as those with a diameter (φ) ≤2 mm (Finér et al. 2011), have higher production rates and faster turnover than coarse roots (McCormack et al. 2015). Changes in fine root phenology are one of the most sensitive indicators of the response of forest ecosystem toward climate change, including global warming (Rosenzweig et al. 2008;Yuan and Chen 2010). It is crucial to understand root phenology to clarify how carbon and nutrient cycling in forest ecosystems respond to climate change because fine root production accounts for 30-50% of net primary production in forest ecosystems (Jackson et al. 1997;Litton and Giardina 2008). Although the thermoresponse and phenology of aboveground plant parts, such as leaves or branches, have been extensively studied, little is known about root phenology (Radville et al. 2016b).
Fine root phenology can be studied using several indicators, including the initiation time, timing and frequency of peak and duration of root production, and timing of root mortality (Price and Hendrick 1998;McCormack et al. 2014;Radville et al. 2016b). McCormack et al. (2014) reported variable patterns of fine root production phenology, such as distributed, concentrated, and bimodal patterns. Previous studies have determined the changes in these indicators with respect to temperature (Pregitzer et al. 2000;Alvarez-Uria and Körner 2007), solar irradiance (Edwards et al. 2004), water stress (Joslin et al. 2000), and soil depth (Hendrick and Pregitzer 1996;Pregitzer et al. 1998).
Additionally, fine root phenology may also be species specific (McCormack et al. 2014;Radville et al. 2016b), which varies among forest types, such as evergreen or deciduous forests (Quan et al. 2010). These studies suggest that both exogenous and endogenous factors can affect fine root phenology.
Fine root phenology could highly vary both spatially and temporally within a forest. The reason is attributed to both environmental factors and species characteristics that determine the patterns of production and mortality. Several studies have evaluated fine root phenology at the stand scale in cool temperate and boreal forests (e.g. Quan et al. 2010;Fukuzawa et al. 2013;McCormack et al. 2014;Radville et al. 2016b); however, only limited studies are available for other regions, including warm-temperate forests. Furthermore, most of these studies determined fine root phenology in single-species stands, and only a few were conducted considering mixed stands. The variation of tree species in a mixed stand may cause high spatial and temporal variations in fine root phenology in space and time.
The minirhizotron method has been commonly used to investigate fine root phenology and its variation in forest stands (Radville et al. 2016b), but the image size or the observation area here is usually quite small (e.g. an image size of 16 mm × 18 mm; Minirhizotron BTC-100×, Bartz Technology Corporation, CA, California, U.S.A). Thus, it is difficult to determine representative patterns of root phenology at a forest stand scale and to observe root branching and distribution in the soil. In contrast, the optical scanner method (Dannoura et al. 2008) provides a larger image (e.g. 200 mm × 300 mm) and is more affordable when a commercial scanner is used. It is expected that the scanner method could be a cost-effective representative method for root phenology investigation in a forest stand producing reliable results with limited replicates.
In this study, we investigated the spatial variations in fine root phenology in a mixed warm-temperate forest in Japan, using an optical scanner. We evaluated the variation using two scales: among the plots within the stand and within the scanner images. We hypothesized that root phenology would differ across plots and within images both due to the diversity of tree species, which means the endogenous factors have a larger impact than the exogenous factors. Since it is known that fine root phenology is species specific (e.g. McCormack et al. 2014, Wang et al. 2020, the various tree species in the mixed stand could cause unclear pattern of the timing of fine root initiation and cessation. Also, the peak frequency of fine root production, and the duration of root production and mortality can vary because of the strong impact of endogenous factors in both of stand and local scales. On the other hand, the impact of exogenous factors, such as temperature and precipitation, could be small because temperature and moisture stress on fine root dynamics would be less in the warm-temperate climate, as observed in previous study conducted in a tropical forest (Endo et al. 2019). In this study, we further discussed the applicability of the scanner method in future studies on fine root dynamics in a forest.

Study site
This study was conducted at the Himeji Nature Sanctuary in Himeji City, Hyogo Prefecture, Japan (34°8'N, 134°6'E). The stand was a secondary forest growing broad-leaved trees such as jolcham oak (Quercus serrata), Oriental oak (Quercus variabilis), Japanese clethra (Clethra barbinervis), and East Asian eurya (Eurya japonica). In 2017, the stand density and mean stem diameter at breast height (DBH) were 0.33 trees m −2 and 11.2 cm, respectively (Table 1). We established four random replicate plots in the stand that had various tree species and sizes (Table 1). Each plot was a circle with 3 m radius, established within an area of 10 m × 10 m. Annual precipitation and mean annual temperature in Himeji city was 1,670 mm and 15.9°C in 2018 and 1,084 mm and 16.2°C in 2019. Daily maximum and minimum air temperature were 27.4°C and −5.7°C in 2018 and 28.7°C and −2.1°C in 2019, respectively (Japan Meteorological Agency, 2018-2019). The soil texture was clay loam derived from rhyolite (Wild Bird Society of Japan, 1992).

Scanning
Observations of fine root dynamics were performed at the center of the four plots. Images of the soil profile with a depth of 210 mm and a width of 297 mm were obtained using the optical scanner method (Dannoura et al. 2008). Soil profiles were scanned using a commercial scanner (GT-S 640, EPSON) at a resolution of 650 dpi with 48-bit colors. Scanning was performed from April 2017 to December 2019. The area of the woody roots on the scanned images was determined manually according to the procedure described by Kume et al. (2018). In this study, we extracted the details of all the roots from all images. Standing crop, growth, and mortality were defined as gross projected area (mm 2 cm-2 ), increase in area between two sequential images (mm 2 cm-2 ), and decrease in area between two sequential images (mm 2 cm-2 ), respectively.
We installed transparent containers in each plot in September 2016, and analyzed the data from January 2018 (one and a half years following installation) for 2 years (n = 24), to avoid any effect of soil disturbance for evaluating root phenology (Nakahata and Osawa 2017).

Data analysis
Root phenology was characterized based on a method modified from previous studies (McCormack et al. 2014;Radville et al. 2016b;Kou et al. 2019). First, we calculated the daily production and mortality (mm 2 cm− 2 day −1 ) by dividing the root area changes in each image by the number of days since the previous image was acquired. Monthly production and mortality (mm 2 cm− 2 month −1 ) were then calculated by accumulating daily production and mortality (Quan et al. 2010). The proportion of monthly root production and mortality was expressed as a percentage of total annual root production and mortality, respectively. The peaks in which we recorded twice as high as the average root production and mortality were defined as peak production (P p ) and peak mortality (P m ), respectively, and those that observed 1.5 times higher root production and mortality were defined as sub-peak production (p p ) and sub-peak mortality (p m ), respectively. We also calculated the sum of days when the daily root production and mortality exceeded the average root production and mortality across the year, which were defined as the active production (A p ) and mortality (A m ) times, respectively.
We compared monthly values of the fine root standing crop (mm 2 cm −2 ) between the four plots and between the four areas in each image where we divided it into four areas U ℓ (upper left), U r (upper right), L ℓ (lower left), and L r (lower right) ( Figure S1) using a nonparametric Friedman test. Similarly, we compared fine root production and mortality between the four plots and between the four areas (mm 2 mm− 2 month− 1 ). Here, we divided each value by the initial standing crop in each plot and each area to exclude the influence of fine root distribution. When a significant difference was observed, multiple comparisons were done using the Wilcoxon test and Bonferroni correction. Data were analyzed using R 3.5.2 statistical software.

Fine root dynamics in scanner images
The scanner method allowed us to observe the growth process of the root systems because the size of the scan image/area under observation was large enough to monitor the root connections. For example, while comparing the images of Plot 3 taken in September 2017 (Figures 1a,  1c) with those taken in October 2017 (Figures 1b, 1d), we observed the elongation of a second-order root and branching of several first-order roots. Furthermore, we identified the number of branches and their positions. Changes in root color from light brown to dark brown were observed during the growth process (Figures 1c, 1d).
The root system observed from June 2018 to September 2019 showed five instances of fine root production, and three of them disappeared after 9 or 10 months ( Figure 2).

Variation between plots
Among the four plots, mean annual standing crop was 15.5 ± 4 mm 2 cm −2 year −1 , mean annual production was 0.83 ± 0.14 mm 2 cm −2 year −1 , and mean annual mortality was 0.78 ± 0.11 mm 2 cm −2 year −1 (n = 8). There were significant differences in the mean monthly standing crop (mm 2 mm− 2 ), standardized monthly production and monthly mortality (mm 2 mm− 2 month− 1 ) between the four plots ( Table 2). The mean monthly standing crop was the smallest in Plot 3 and largest in Plot 4 (p < 0.05, Figure S2). The mean monthly production in Plot 3 was significantly higher than that in Plots 1, 2, and 4 (P < 0.05). The mean monthly mortality in Plot 3 was significantly higher than that in Plots 2 and 4 (P < 0.05).
The mean fine root production in the four plots showed different phenologies between the first and second years ( Figure 3a). Peaks were observed in spring and autumn in the first year, but in winter in the second year. The production phenology in each plot showed that the occurrence of peaks (P p ) and sub-peaks (p p ) was concentrated in May, June, and October in the first year (Figures 3(b-e)). In contrast, P p occurred in January or February in all plots during the second year. These results show that there is annual variation in fine root production patterns in this forest. The sum of active days (A p ) varied from 119 to 184 days in the first year and 121 to 212 days in the second year among the four plots, showing a double difference between them.
The mean fine root mortality of the four plots showed similar phenotypes between the first and second years (Figure 2f), in which peaks were observed in summer. The mortality phenology in each plot showed that most P m occurred between August and September in both years (Figures 3(g-j)). A m varied from 92 to 182 days and 122 to 153 days among the four plots in the first and second years, respectively. In the first year, the difference between the values nearly doubled, but it was less than half in the second year, showing high spatial variation only in the first year.

Variation within the scanner image
There were significant differences in standing crops among all the plots of four areas (Table 2), but with differences in the patterns of spatial distribution ( Figure S3). Significant differences in the amount of standardized root production and mortality (mm 2 mm− 2 month− 1 ) among the four areas were found in Plots 2, 3, and 4 (Table 2, Figure S4).
We found that the patterns of root production phenology differed among the four areas, especially in Plot 4 ( Figure 4, Figure S5-S7). In the first year, P p was observed twice in winter and autumn in the areas L ℓ and L r , but only once in summer in U ℓ (Figures 4(a-c)). P p and p p were found to occur several times intermittently in Ur (Figure 4d). In the second year, P p was found once in L ℓ, twice in U ℓ and L r , and three times in U r in different months (Figures 4(ad)). A p was approximately 180 days in U r in both the years, but only 89-120 days in L ℓ .
In the first year, the monthly mortality phenology revealed no P m in L ℓ , and one or two P m in the other areas (Figures 4(e-h)). In the second year, one P m was found in August or September in U ℓ , L ℓ , and L r areas, whereas two P m s in July and December in U r (Figures 4e, 4f, 4h). A m differed among the areas and was longest in the L ℓ area (215 days) and shortest in the U ℓ area (92 days) in the first year and longest in the L r area (184 days) and shortest in the L ℓ area (92 days) in the second year.

Discussion
In this study, we hypothesized that endogenous factors would have a larger impact than exogenous factors on fine  .001*** χ2 and df are the chi-square values and degrees of freedom, respectively. * p < 0.05, ** p < 0.01, *** p < 0.001. root phenology. However, our results revealed that peaks of root production in each plot were primarily concentrated in a specific period, which might again vary year by year, and those of fine root mortality were concentrated in summer. This suggested that exogenous factors could be stronger regulators of fine root production and mortality than endogenous factors, resulting in nearly synchronized production and mortality phenology across plots. Previous studies have also demonstrated a strong impact of local environments on fine root growth and distribution, masking the influence of internal factors on the species ( The variation in time of peak appearance in root production with respect to year could be attributed to the impact of prevailing temperature and precipitation (Quan et al. 2010;Schwieger et al. 2021). The study site had a rainfall of 1,671 mm in 2018, which was far above the average rainfall of 1,254 mm in 1991-2020 (Japan Meteorological Agency 2018). However, year 2019 was warmer in which the daily minimum temperature recorded the highest value in the history of observation (Japan Meteorological Agency 2019). Root dynamics exhibit plasticity and vary with multiple environmental factors (Erkan et al. 2018), which may have Figure 3. Temporal variation in fine root production and mortality in each plot. Black and white triangles denote the peak and sub-peak of production and mortality, respectively. The dashed line shows the annual mean fine root production and mortality. A p and A m refer to active production and mortality times. resulted in annual variation due to different regulators in these years.
In this study, the impact of endogenous factors on fine root phenology was not clear. Several studies have reported a relationship between above-and below-ground dynamics (Radville et al. 2016a) and different patterns of fine root phenology between evergreen and deciduous species (Fukuzawa et al. 2013). The multiple peaks and sub-peaks of root dynamics observed in this study might have been caused by the co-existence of evergreen and deciduous trees in each plot. For example, the increase in root production during winter might be attributed to the continuous photosynthetic activities of evergreen species. However, asynchronous aboveground and belowground dynamics have also been reported (Abramoff and Finzi 2014), and there is limited information available on the root phenology of temperate evergreen broadleaved species. Additional study on the patterns of root phenology in different tree species is necessary to understand the influence of endogenous factors over fine root phenological dynamics in a mixed forest. In this study, the active production time (A p ) showed high variation in both years, and the active mortality time (A m ) varied within the first year only among the four plots ( Figure 3). These indices were able to capture the span of the active root season, where concentrated patterns of production and mortality phenologies appeared for a shorter time than the patterns with broader or multiple peaks. Our results revealed that, even though the timing of peaks synchronized among the plots, the duration of peaks appearance was different, which might be caused by the mixture of species that differed among the plots. Previous study reported that the timing of peaks may be species specific  Figure S1 for a detailed explanation of area partitioning. The black and white triangles denote the peak and sub-peak of production and mortality, respectively. The dashed line shows the annual mean fine root production and mortality. A p and A m refer to active production and mortality times. (McCormack et al. 2014). If the timings of peak are slightly different due to the influence of the species, the duration of peaks appearance could vary depending on the species composition of the plot. We also found significant differences in standing crops, standardized values of root production, and mortality among the four plots (Table 2). These results suggest the influence of endogenous factors on fine root dynamics, which might have been caused by the differences in active times.
This study revealed significant differences in fine root production and mortality (Table 2) and the timing of peak production and mortality (Figure 4) among the four areas in the images in most of the plots. Both A p and A m values were highly variable among the four areas. These differences might have been caused by a decrease in the image area and an increase in the area ratio occupied by individual roots. Here, the phenology observed in the smaller images depended on the roots that may have its own dynamics. Since the phenology of particular roots can strongly depends on the traits of tree species and the surrounding soil microenvironment, large differences in root production and mortality caused by such specific root could be easily observed if the image size is small. This means that if a high number of replicates or a large observation area, including a substantial number of fine roots, are obtained, a representative phenology of fine roots in the stand scale could be determined. In this study, the larger observation area provided by the scanner method probably offered advantages for understanding root phenology in the mixed stand, in which the various local phenology of particular roots might have been masked.
Our results suggest that the scanner method can clearly show the process of root system development, timing of root elongation, locations and number of root branching points, changes in root color, and spatial distribution of individual roots, which is difficult to determine in other conventional methods (Figure 1, 2). In general branch-order analysis, for example, unit of root branches is sampled and then separated into individual root branches at each intersection (Pregitzer et al. 2002;Wada et al. 2019). With this method, it is difficult to observe the process of root system establishment. Scanner image analysis can be a breakthrough method for understanding the root structure, function, as well as the responses to the influence of endogenous and exogenous factors.
In conclusion, this study revealed that the patterns of spatial distribution of fine roots differed among the four areas of scanner size images in most plots; however, a synchronized root phenology pattern appeared on a stand scale. Our results showed that peaks of root production in the mixed forest are concentrated for a certain period that may again vary yearly, and those of fine root mortality are concentrated in summer. Differences in root phenology influenced by internal factors of the species were masked on the stand scale.