Data for deep root carbon allocation of planted forest in the Loess Plateau

In this study, we hypothesize that deep root carbon (DRC) inputs in planted forests can be reliably inferred from forest age. To evaluate this hypothesis, we compiled forest age data and corresponding DRC input observations across the CLP. We first developed an empirical model that quantifies the relationship between stand age and stand-level DRC input density, and subsequently applied this model to assess the spatial distribution of DRC inputs using regional forest age dataset. Deep root biomass data used as train points in this study were collected from field sampling from 2016 to 2019. A roots auger (inner diameter 0.085 m) was used to sample the 37 pairs of sites (Supplementary data). Soil samples were collected at 0.20 m intervals down to the depth where the lack of roots was confirmed. The maximum sampling depth ranged from 5 to 30 m, with an average depth of 14.4 m. Each sample was washed carefully using 1 mm sieves to obtain fresh roots samples. After that, the fresh roots were oven-dried at 60 ℃ for 72 h and weighed to determine the roots dry weights. Finally, the roots sample after drying and weighing was selected, and the roots carbon content was determined by elemental analyzer after grinding, which was used for the subsequent calculation of carbon storage. In addition, adhering to the principle of sampling down to the maximum rooting depth, we identified ten literature sources that met the required criteria. Finally, our 37 paired sample sites include, to the extent possible, forest distributions of different tree species, climates, landforms and soil types throughout the Loess Plateau . Among them, 27 sampling points are used to construct the DRC input estimation model. The forest age data used in this study was collected from field sampling from 2016 to 2019. At each forest site, the forest age was determined based on tree rings from multiple trees and site survey from local farmers and field station staff. The forest age distribution map of the CLP was identified based on the LandTrendr algorithm according to the temporal change of the normalized burn ratio index after afforestation. Compared with ground truth data, the overall accuracy of the forest age distribution map was 89%, with a root-mean-square error (RMSE) of 2.14 years.