The dynamics of tree fine roots and their drivers are crucial for understanding forest carbon cycling. The scanner method, a non-destructive approach for observing fine roots in large soil profile images, has been used to evaluate fine-root dynamics. Traditionally, roots in these images were extracted manually, a process that required significant time and effort. Recently, an automated root segmentation tool based on deep learning, A Root Auto Tracing and Analysis (ARATA), has been developed to extract fine roots from scanned images. We previously showed that transfer learning to the ARATA model using site-specific images of the target stand greatly improves the performance of automated root extraction. In the present study, we examined the relationships between fine-root dynamics and climatic factors as well as stem growth phenology, using the optimized root extraction model trained with site-specific images.
　This study was conducted in a 120-year-old Chamaecyparis obtusa stand at the Kota monitoring site, Aichi, Japan. Scanning was performed once monthly in five transparent boxes over two years, and fine roots were automatically detected using an optimized model. After the root extraction, we quantified the daily net change in the total root length per imaged area (root length density, RLD), which includes both growth and mortality. We also measured air temperature, throughfall, and stem circumference on the same dates as fine-root image acquisitions. We used linear models to test the effects of air temperature, throughfall, and stem growth on daily net change in RLD. The models were selected based on the Akaike Information Criterion (AIC). In the C. obtusa stand, fine-root phenology exhibited clear seasonal patterns with active growth in autumn and pronounced mortality in summer. Linear model analysis indicated that this seasonal variation was primarily explained by air temperature, with a negative effect. The analysis also showed that no clear temporal synchrony was observed between root and stem growth phenology. Notably, peaks in fine-root growth of C. obtusa lagged behind periods of active stem growth. Advances in automated image analysis are expected to promote integrated analyses of fine-root dynamics and environmental controls across diverse forest ecosystems.