Using process-oriented model output to enhance machine learning-based soil organic carbon prediction in space and time.

Monitoring and modelling soil organic carbon (SOC) in space and time can help us to better understand soil carbon dynamics and is of key importance to support climate change research and policy. Although machine learning (ML) has attracted a lot of attention in the digital soil mapping (DSM) community for its powerful ability to learn from data and predict soil properties, such as SOC, it is better at capturing soil spatial variation than soil temporal dynamics. By contrast, process-oriented (PO) models benefit from mechanistic knowledge to express physiochemical and biological processes that govern SOC temporal changes. Therefore, integrating PO and ML models seems a promising means to represent physically plausible SOC dynamics while retaining the spatial prediction accuracy of ML models. In this study, a hybrid modelling framework was developed and tested for predicting topsoil SOC stock in space and time for a regional cropland area located in eastern China. In essence, the hybrid model uses predictions of the PO model in unsampled years as additional training data of the ML model, with a weighting parameter assigned to balance the importance of SOC values from the PO model and real measurements. The results indicated that temporal trends of SOC stock modelled by PO and ML models were largely different, while they were notably similar between the PO and hybrid models. Cross-validation showed that the hybrid model had the best performance (RMSE = 0.29 kg m-2), with a 19 % improvement compared with the ML model. We conclude that the proposed hybrid framework not only enhances space-time soil carbon mapping in terms of prediction accuracy and physical plausibility, it also provides insights for soil management and policy decisions in the face of future climate change and intensified human activities.

Soil carbon stocks in stable tropical landforms are dominated by geochemical controls and not by land use.

Soil organic carbon (SOC) dynamics depend on soil properties derived from the geoclimatic conditions under which soils develop and are in many cases modified by land conversion. However, SOC stabilization and the responses of SOC to land use change are not well constrained in deeply weathered tropical soils, which are dominated by less reactive minerals than those in temperate regions. Along a gradient of geochemically distinct soil parent materials, we investigated differences in SOC stocks and SOC (Δ14 C) turnover time across soil profile depth between montane tropical forest and cropland situated on flat, non-erosive plateau landforms. We show that SOC stocks and soil Δ14 C patterns do not differ significantly with land use, but that differences in SOC can be explained by the physicochemical properties of soils. More specifically, labile organo-mineral associations in combination with exchangeable base cations were identified as the dominating controls over soil C stocks and turnover. We argue that due to their long weathering history, the investigated tropical soils do not provide enough reactive minerals for the stabilization of C input in either high input (tropical forest) or low-input (cropland) systems. Since these soils exceeded their maximum potential for the mineral related stabilization of SOC, potential positive effects of reforestation on tropical SOC storage are most likely limited to minor differences in topsoil without major impacts on subsoil C stocks. Hence, in deeply weathered soils, increasing C inputs may lead to the accumulation of a larger readily available SOC pool, but does not contribute to long-term SOC stabilization.

Climate warming suppresses abundant soil fungal taxa and reduces soil carbon efflux in a semi-arid grassland.

Soil microorganisms critically affect the ecosystem carbon (C) balance and C-climate feedback by directly controlling organic C decomposition and indirectly regulating nutrient availability for plant C fixation. However, the effects of climate change drivers such as warming, precipitation change on soil microbial communities, and C dynamics remain poorly understood. Using a long-term field warming and precipitation manipulation in a semi-arid grassland on the Loess Plateau and a complementary incubation experiment, here we show that warming and rainfall reduction differentially affect the abundance and composition of bacteria and fungi, and soil C efflux. Warming significantly reduced the abundance of fungi but not bacteria, increasing the relative dominance of bacteria in the soil microbial community. In particular, warming shifted the community composition of abundant fungi in favor of oligotrophic Capnodiales and Hypocreales over potential saprotroph Archaeorhizomycetales. Also, precipitation reduction increased soil total microbial biomass but did not significantly affect the abundance or diversity of bacteria. Furthermore, the community composition of abundant, but not rare, soil fungi was significantly correlated with soil CO2 efflux. Our findings suggest that alterations in the fungal community composition, in response to changes in soil C and moisture, dominate the microbial responses to climate change and thus control soil C dynamics in semi-arid grasslands.

Soil carbon stocks and dynamics of different land uses in Italy using the LUCAS soil database.

In terrestrial biosphere, soil represents the largest organic carbon pool, and a small change of soil organic carbon (SOC) can significantly affect the global carbon cycle and climate. Land use change (LUC) and soil management practices coupled with climate variables can significantly influence the soil organic carbon stocks (SOC-S) and its dynamics; however, our understanding about the responses of SOC in different LUC's (e.g., cropland, grassland and forest land) to mitigate climate change is quite limited at country level like Italy. Thus, the aims of this study were which factors are affecting SOC dynamics in three LUC's over time across Italy; and their relevance in terms of SOC-S in the superficial layer of soil that significantly contributes to the climate change mitigation, using LUCAS soil database. To calculate the SOC-S, it is necessary to have soil bulk density (BD) which is not present in the LUCAS database. Thus, we estimate the soil BD using the pedotransfer function (PTFs); and results shows that the soil BD obtained from fitting of the PTFs were reasonable to estimate the SOC-S for different land use types (R2 ≥ 0.75). Overall, results showed that LUC's and soil management practices can significantly (p < 0.001) influences SOC dynamics and SOC storage from the soil and varied among LUC's but not for over time except grassland. Spatially, the mean SOC-S storage of the different LUC's was in the following order: forest land > grassland > cropland for both years 2009 and 2015. On the other hand, the SOC-S storage increased by 8.33% for cropland, 13.56% for forest land, and 29.79% for grassland during the year of 2009-2015, while SOC-S storage increased significantly (p < 0.001) in grassland over time but not for cropland and forest land which also follow the increasing trend but insignificantly. Our results also reveal that the SOC dynamics negatively correlated with MAT, and positively correlated with MAP for all land uses except forest land. Thus, this research indicates that LUC's and soil management practices coupled with climate variables can significantly influence SOC storage and its dynamics in the superficial layer of soil which have the potential capacity to mitigate climate change.

Soil carbon persistence governed by plant input and mineral protection at regional and global scales.

Elucidating the processes underlying the persistence of soil organic matter (SOM) is a prerequisite for projecting soil carbon feedback to climate change. However, the potential role of plant carbon input in regulating the multi-layer SOM preservation over broad geographic scales remains unclear. Based on large-scale soil radiocarbon (∆14 C) measurements on the Tibetan Plateau, we found that plant carbon input was the major contributor to topsoil carbon destabilisation despite the significant associations of topsoil ∆14 C with climatic and mineral variables as well as SOM chemical composition. By contrast, mineral protection by iron-aluminium oxides and cations became more important in preserving SOM in deep soils. These regional observations were confirmed by a global synthesis derived from the International Soil Radiocarbon Database (ISRaD). Our findings illustrate different effects of plant carbon input on SOM persistence across soil layers, providing new insights for models to better predict multi-layer soil carbon dynamics under changing environments.

Natural abundance of 13 C and 15 N provides evidence for plant-soil carbon and nitrogen dynamics in a N-fertilized meadow.

Natural abundance of carbon (C) and nitrogen (N) stable isotope ratios (δ13 C and δ15 N) has been used to indicate ecosystem C and N status and cycling; however, use of this approach to infer plant and microbial N preference under projected ecosystem N enrichment is limited. Here, we investigated natural abundance δ13 C and δ15 N of five dominant plant species, and soil δ15 N of microbial biomass and available N forms under N addition in a meadow steppe. Additional N, applied as urea, led to decreases in δ15 N of soil NO3- (δ15 Nnitrate , from 3.0 to 0.4‰) and increases in δ15 N of soil NH4+ (δ15 Nammonium , from -1.3 to 11‰) and dissolved organic N (δ15 NDON , from 8.5 to 15‰) that reflected increased net nitrification rates, a possible increase in NH3 volatilization, and greater availability of the three N forms. An overall increase in δ15 N of soil total N (δ15 NTN ) from 7.1 to 7.9‰ indicated accelerated and greater openness of soil N cycling that was also partially revealed by enhanced net N mineralization rates. Plant δ15 N, which ranged from -1.8 to 2.1‰, generally decreased with N addition, indicating a greater reliance on soil NO3- under N-enrichment conditions. Nitrogen addition decreased δ15 N of microbial biomass N (from 14 to 2.8‰), possibly because of a shift in preferential N form (DON to NO3- ), that indicated a convergence of plant and microbial preferential N forms and an increase in plant-microbial N competition. Microbes were thus more flexible than plants in the use of different forms of N. Addition of N decreased plant litter δ13 C, whereas plant species δ13 C remained unaffected, likely because of a shift in the abundance of dominant species with a greater proportion of biomass coming from δ13 C-depleted species. Enrichment factor (the difference in plant δ15 N relative to δ15 NTN ) of four nonlegume species was negatively related to soil inorganic N availability, net nitrification rate, and net N mineralization rate, and was proven to be a good indicator of ecosystem N status. Our study highlights the importance of natural abundance of 15 N as an indicator of plant-microbial N competition and ecosystem N cycling in meadow steppe grasslands under projected ecosystem N enrichment.

Tree planting in organic soils does not result in net carbon sequestration on decadal timescales.

Tree planting is increasingly being proposed as a strategy to combat climate change through carbon (C) sequestration in tree biomass. However, total ecosystem C storage that includes soil organic C (SOC) must be considered to determine whether planting trees for climate change mitigation results in increased C storage. We show that planting two native tree species (Betula pubescens and Pinus sylvestris), of widespread Eurasian distribution, onto heather (Calluna vulgaris) moorland with podzolic and peaty podzolic soils in Scotland, did not lead to an increase in net ecosystem C stock 12 or 39 years after planting. Plots with trees had greater soil respiration and lower SOC in organic soil horizons than heather control plots. The decline in SOC cancelled out the increment in C stocks in tree biomass on decadal timescales. At all four experimental sites sampled, there was no net gain in ecosystem C stocks 12-39 years after afforestation-indeed we found a net ecosystem C loss in one of four sites with deciduous B. pubescens stands; no net gain in ecosystem C at three sites planted with B. pubescens; and no net gain at additional stands of P. sylvestris. We hypothesize that altered mycorrhizal communities and autotrophic C inputs have led to positive 'priming' of soil organic matter, resulting in SOC loss, constraining the benefits of tree planting for ecosystem C sequestration. The results are of direct relevance to current policies, which promote tree planting on the assumption that this will increase net ecosystem C storage and contribute to climate change mitigation. Ecosystem-level biogeochemistry and C fluxes must be better quantified and understood before we can be assured that large-scale tree planting in regions with considerable pre-existing SOC stocks will have the intended policy and climate change mitigation outcomes.

Deep Learning Optimizes Data-Driven Representation of Soil Organic Carbon in Earth System Model Over the Conterminous United States.

Soil organic carbon (SOC) is a key component of the global carbon cycle, yet it is not well-represented in Earth system models to accurately predict global carbon dynamics in response to climate change. This novel study integrated deep learning, data assimilation, 25,444 vertical soil profiles, and the Community Land Model version 5 (CLM5) to optimize the model representation of SOC over the conterminous United States. We firstly constrained parameters in CLM5 using observations of vertical profiles of SOC in both a batch mode (using all individual soil layers in one batch) and at individual sites (site-by-site). The estimated parameter values from the site-by-site data assimilation were then either randomly sampled (random-sampling) to generate continentally homogeneous (constant) parameter values or maximally preserved for their spatially heterogeneous distributions (varying parameter values to match the spatial patterns from the site-by-site data assimilation) so as to optimize spatial representation of SOC in CLM5 through a deep learning technique (neural networking) over the conterminous United States. Comparing modeled spatial distributions of SOC by CLM5 to observations yielded increasing predictive accuracy from default CLM5 settings (R 2 = 0.32) to randomly sampled (0.36), one-batch estimated (0.43), and deep learning optimized (0.62) parameter values. While CLM5 with parameter values derived from random-sampling and one-batch methods substantially corrected the overestimated SOC storage by that with default model parameters, there were still considerable geographical biases. CLM5 with the spatially heterogeneous parameter values optimized from the neural networking method had the least estimation error and less geographical biases across the conterminous United States. Our study indicated that deep learning in combination with data assimilation can significantly improve the representation of SOC by complex land biogeochemical models.

Burning increases post-fire carbon emissions in a heathland and a raised bog, but experimental manipulation of fire severity has no effect.

Large amounts of carbon are stored in northern peatlands. There is concern that greater wildfire severity following projected increases in summer drought will lead to higher post-fire carbon losses. We measured soil carbon dynamics in a Calluna heathland and a raised peat bog after experimentally manipulating fire severity. A gradient of fire severity was achieved by simulating drought in 2 × 2 m plots. Ecosystem respiration (ER), net ecosystem exchange (NEE), methane (CH4) flux and concentration of dissolved organic carbon ([DOC], measured at the raised bog only) were measured for up to two years after burning. The response of these carbon fluxes to increased fire severity in drought plots was similar to plots burnt under ambient conditions associated with traditional managed burning. Averaged across all burnt plots, burning altered mean NEE from a net carbon sink at the heathland (-0.33 µmol CO2 m-2 s-1 in unburnt plots) to a carbon source (0.50 µmol m-2 s-1 in burnt plots) and at the raised bog (-0.38 and 0.16 µmol m-2 s-1, respectively). Burning also increased CH4 flux at the raised bog (from 1.16 to 25.3 nmol m-2 s-1 in the summer, when it accounted for 79% of the CO2-equivalent emission). Burning had no significant effect on soil water [DOC].

Human-induced and natural carbon storage in floodplains of the Central Valley of California.

Active floodplains can putatively store large amounts of organic carbon (SOC) in subsoils originating from catchment erosion processes with subsequent floodplain deposition. Our study focussed on the assessment of SOC pools associated with alluvial floodplain soils that are affected by human-induced changes in floodplain deposition and in situ SOC mineralisation due to land use change and drainage. We evaluated depth-dependent SOC contents based on 23 soil cores down to 3 m and 10 drillings down to 7 m in a floodplain area of the lower Cosumnes River. An estimate of 266 Mg C ha-1 or about 59% of the entire SOC stored within the 7 m profiles was found in the upper 2 m. Most profiles (n = 25) contained discrete buried A horizons at depths of approximately 0.8 m. These profiles had up to 130% higher SOC stocks. The mean δ13C of all deep soil profiles clearly indicated that arable land use has already altered the stable isotopic signature in the first meter of the profile. Radiocarbon dating showed that the 14C age in the buried horizon was younger than in overlaying soils indicating a substantial sedimentation phase for the overlaying soils. An additional analysis of total mercury contents in the soil profiles indicated that this sedimentation was associated with upstream hydraulic gold mining after the 1850s. In summary, deep alluvial soils in floodplains store large amounts of SOC not yet accounted for in global carbon models. Historic data give evidence that large amounts of sediment were transported into the floodplains of most rivers of the Central Valley and deposited over organically rich topsoil, which promoted the stabilization of SOC, and needs to be considered to improve our understanding of the human-induced interference with C cycling.

Long-term impact of conservation agriculture and diversified maize rotations on carbon pools and stocks, mineral nitrogen fractions and nitrous oxide fluxes in inceptisol of India.

Given the increasing scarcity of production resources such as water, energy and labour coupled with growing climatic risks, maize-based production systems could be potential alternatives to intensive rice-wheat (RW) rotation in western Indo-Gangetic Plains (IGP). Conservation agriculture (CA) in maize systems has been widely promoted for minimizing soil degradation and ensuring sustainability under emerging climate change scenario. Such practices are also believed to provide mitigation co-benefits through reduced GHG emission and increased soil carbon sequestration. However, the combined effects of diversified crop rotations and CA-based management on GHG mitigation potential and other co-benefits are generally over looked and hence warrant greater attention. A field trial was conducted for 5-years to assess the changes in soil organic carbon fractions, mineral-N, N2O emission and global warming potential (GWP) of maize-based production systems under different tillage & crop establishment methods. Four diversified cropping systems i.e. maize-wheat-mungbean (MWMb), maize-chickpea-Sesbania (MCS), maize-mustard-mungbean (MMuMb) and maize-maize-Sesbania (MMS) were factorially combined with three tillage & crop establishment methods i.e. zero tilled permanent beds (PB), zero-tillage flat (ZT) and conventional tillage (CT) in a split-plot design. After 5-years of continued experimentation, we recorded that across the soil depths, SOC content, its pools and mineral-N fractions were greatly affected by tillage & crop establishment methods and cropping systems. ZT and PB increased SOC stock (0-30 cm depth) by 7.22-7.23 Mg C ha-1 whereas CT system increased it only by 0.88 Mg C ha-1as compared to initial value. Several researchers reported that SOC & mineral-N fraction contents in the top 30 cm soil depth are correlated with N2O-N emission. In our study, global warming potential (GWP) under CT system was higher by 18.1 and 17.4%, compared to CA-based ZT and PB, respectively. Among various maize systems, GWP of MMS were higher by 11.2, 6.7 and 6.6%, compared that of MWMb (1212 kg CO2-eq. ha-1), MCS (1274 kg CO2-eq. ha-1) and MMuMb (1275 kg CO2-eq. ha-1), respectively. The results of our study suggest that CA and diversified crop rotations should be promoted in north-western IGP and other similar agro-ecologies across the globe for ensuring food security, restoration of soil health and climate change mitigation, the key sustainable development goals (SDGs).

Distinct responses of soil respiration to experimental litter manipulation in temperate woodland and tropical forest.

Global change is affecting primary productivity in forests worldwide, and this, in turn, will alter long-term carbon (C) sequestration in wooded ecosystems. On one hand, increased primary productivity, for example, in response to elevated atmospheric carbon dioxide (CO 2), can result in greater inputs of organic matter to the soil, which could increase C sequestration belowground. On other hand, many of the interactions between plants and microorganisms that determine soil C dynamics are poorly characterized, and additional inputs of plant material, such as leaf litter, can result in the mineralization of soil organic matter, and the release of soil C as CO 2 during so-called "priming effects". Until now, very few studies made direct comparison of changes in soil C dynamics in response to altered plant inputs in different wooded ecosystems. We addressed this with a cross-continental study with litter removal and addition treatments in a temperate woodland (Wytham Woods) and lowland tropical forest (Gigante forest) to compare the consequences of increased litterfall on soil respiration in two distinct wooded ecosystems. Mean soil respiration was almost twice as high at Gigante (5.0 µmol CO 2 m-2 s-1) than at Wytham (2.7 µmol CO 2 m-2 s-1) but surprisingly, litter manipulation treatments had a greater and more immediate effect on soil respiration at Wytham. We measured a 30% increase in soil respiration in response to litter addition treatments at Wytham, compared to a 10% increase at Gigante. Importantly, despite higher soil respiration rates at Gigante, priming effects were stronger and more consistent at Wytham. Our results suggest that in situ priming effects in wooded ecosystems track seasonality in litterfall and soil respiration but the amount of soil C released by priming is not proportional to rates of soil respiration. Instead, priming effects may be promoted by larger inputs of organic matter combined with slower turnover rates.

The Automated Root Exudate System (ARES): a method to apply solutes at regular intervals to soils in the field.

Root exudation is a key component of nutrient and carbon dynamics in terrestrial ecosystems. Exudation rates vary widely by plant species and environmental conditions, but our understanding of how root exudates affect soil functioning is incomplete, in part because there are few viable methods to manipulate root exudates in situ. To address this, we devised the Automated Root Exudate System (ARES), which simulates increased root exudation by applying small amounts of labile solutes at regular intervals in the field.The ARES is a gravity-fed drip irrigation system comprising a reservoir bottle connected via a timer to a micro-hose irrigation grid covering c. 1 m2; 24 drip-tips are inserted into the soil to 4-cm depth to apply solutions into the rooting zone. We installed two ARES subplots within existing litter removal and control plots in a temperate deciduous woodland. We applied either an artificial root exudate solution (RE) or a procedural control solution (CP) to each subplot for 1 min day-1 during two growing seasons. To investigate the influence of root exudation on soil carbon dynamics, we measured soil respiration monthly and soil microbial biomass at the end of each growing season.The ARES applied the solutions at a rate of c. 2 L m-2 week-1 without significantly increasing soil water content. The application of RE solution had a clear effect on soil carbon dynamics, but the response varied by litter treatment. Across two growing seasons, soil respiration was 25% higher in RE compared to CP subplots in the litter removal treatment, but not in the control plots. By contrast, we observed a significant increase in microbial biomass carbon (33%) and nitrogen (26%) in RE subplots in the control litter treatment.The ARES is an effective, low-cost method to apply experimental solutions directly into the rooting zone in the field. The installation of the systems entails minimal disturbance to the soil and little maintenance is required. Although we used ARES to apply root exudate solution, the method can be used to apply many other treatments involving solute inputs at regular intervals in a wide range of ecosystems.

Soil respiration and carbon loss relationship with temperature and land use conversion in freeze-thaw agricultural area.

Soil respiration (Rs) was hypothesized to have a special response pattern to soil temperature and land use conversion in the freeze-thaw area. The Rs differences of eight types of land use conversions during agricultural development were observed and the impacts of Rs on soil organic carbon (SOC) loss were assessed. The land use conversions during last three decades were categorized into eight types, and the 141 SOC sampling sites were grouped by conversion type. The typical soil sampling sites were subsequently selected for monitoring of soil temperature and Rs of each land use conversion types. The Rs correlations with temperature at difference depths and different conversion types were identified with statistical analysis. The empirical mean error model and the biophysical theoretical model with Arrhenius equation about the Rs sensitivity to temperature were both analyzed and shared the similar patterns. The temperature dependence of soil respiration (Q10) analysis further demonstrated that the averaged value of eight types of land use in this freeze-thaw agricultural area ranged from 1.15 to 1.73, which was lower than the other cold areas. The temperature dependence analysis demonstrated that the Rs in the top layer of natural land covers was more sensitive to temperature and experienced a large vertical difference. The natural land covers exhibited smaller Rs and the farmlands had the bigger value due to tillage practices. The positive relationships between SOC loss and Rs were identified, which demonstrated that Rs was the key chain for SOC loss during land use conversion. The spatial-vertical distributions of SOC concentration with the 1.5-km grid sampling showed that the more SOC loss in the farmland, which was coincided with the higher Rs in farmlands. The analysis of Rs dynamics provided an innovative explanation for SOC loss in the freeze-thaw agricultural area. The analysis of Rs dynamics provided an innovative explanation for SOC loss in the freeze-thaw agricultural area.

Global patterns and controls of soil organic carbon dynamics as simulated by multiple terrestrial biosphere models: Current status and future directions.

Soil is the largest organic carbon (C) pool of terrestrial ecosystems, and C loss from soil accounts for a large proportion of land-atmosphere C exchange. Therefore, a small change in soil organic C (SOC) can affect atmospheric carbon dioxide (CO2) concentration and climate change. In the past decades, a wide variety of studies have been conducted to quantify global SOC stocks and soil C exchange with the atmosphere through site measurements, inventories, and empirical/process-based modeling. However, these estimates are highly uncertain, and identifying major driving forces controlling soil C dynamics remains a key research challenge. This study has compiled century-long (1901-2010) estimates of SOC storage and heterotrophic respiration (Rh) from 10 terrestrial biosphere models (TBMs) in the Multi-scale Synthesis and Terrestrial Model Intercomparison Project and two observation-based data sets. The 10 TBM ensemble shows that global SOC estimate ranges from 425 to 2111 Pg C (1 Pg = 1015 g) with a median value of 1158 Pg C in 2010. The models estimate a broad range of Rh from 35 to 69 Pg C yr-1 with a median value of 51 Pg C yr-1 during 2001-2010. The largest uncertainty in SOC stocks exists in the 40-65°N latitude whereas the largest cross-model divergence in Rh are in the tropics. The modeled SOC change during 1901-2010 ranges from -70 Pg C to 86 Pg C, but in some models the SOC change has a different sign from the change of total C stock, implying very different contribution of vegetation and soil pools in determining the terrestrial C budget among models. The model ensemble-estimated mean residence time of SOC shows a reduction of 3.4 years over the past century, which accelerate C cycling through the land biosphere. All the models agreed that climate and land use changes decreased SOC stocks, while elevated atmospheric CO2 and nitrogen deposition over intact ecosystems increased SOC stocks-even though the responses varied significantly among models. Model representations of temperature and moisture sensitivity, nutrient limitation, and land use partially explain the divergent estimates of global SOC stocks and soil C fluxes in this study. In addition, a major source of systematic error in model estimations relates to nonmodeled SOC storage in wetlands and peatlands, as well as to old C storage in deep soil layers.

Predictability of the terrestrial carbon cycle.

Terrestrial ecosystems sequester roughly 30% of anthropogenic carbon emission. However this estimate has not been directly deduced from studies of terrestrial ecosystems themselves, but inferred from atmospheric and oceanic data. This raises a question: to what extent is the terrestrial carbon cycle intrinsically predictable? In this paper, we investigated fundamental properties of the terrestrial carbon cycle, examined its intrinsic predictability, and proposed a suite of future research directions to improve empirical understanding and model predictive ability. Specifically, we isolated endogenous internal processes of the terrestrial carbon cycle from exogenous forcing variables. The internal processes share five fundamental properties (i.e., compartmentalization, carbon input through photosynthesis, partitioning among pools, donor pool-dominant transfers, and the first-order decay) among all types of ecosystems on the Earth. The five properties together result in an emergent constraint on predictability of various carbon cycle components in response to five classes of exogenous forcing. Future observational and experimental research should be focused on those less predictive components while modeling research needs to improve model predictive ability for those highly predictive components. We argue that an understanding of predictability should provide guidance on future observational, experimental and modeling research.