Effects of Vegetation Restoration On Soil Carbon Dynamics In Karst And Non-Karst Regions: A Synthesis of Multi-Source Data

20 Backgrounds A large-scale ecological restoration project has been initiated since 1990s in 21

southwest China, which is one of the largest areas of rocky desertification globally. However, the 22 different influences and potential mechanisms of vegetation restoration on soil carbon(C) 23 sequestration in karst and non-karst regions are still unclear. 24 Methods Based on field investigation and multi-source data synthesis, the mechanisms of soil 25 C sequestration were investigated to determine the most important variables affecting the rate of 26 soil C change (Rs) in southwest China. 27 Results Our results show significant differences in soil C sequestration between karst and non-28 karst regions with faster and longer C sequestration in karst regions, where Rs was approximately 29 31 % higher than in non-karst soils. And temperatures could be the primary factor inhibiting soil C 30 sequestration without precipitation. The total effect of nitrogen (N) on Rs was positive in both karst 31 and non-karst regions. suggested that soil C and nitrogen (N) content increased significantly following vegetation 65 carefully removed from the ground at each sampling point before sampling soil. Along the profile, 110 undisturbed soil samples were collected using a standard container (ring cutting sampler with 100 111 cm 3 in volume) for bulk density measurements at four layers of depth: 0-5, 5-10, 10-20 and 20-50 112 cm. We have recorded sampling depth, location and measured the physical and chemical properties 113 of soil in the laboratory. The complete synthesized dataset included 65 published articles (Table S2) 114 and data from 20 sampling sites (Table S3) covering both karst (228 observation points) and non-115 karst regions (548 observation points) (Fig. 1). Data relating to the karst vector distribution came 116 from the Karst Scientific Data Center (https://www.karstdata.cn/). Vegetation types selected for this 117 study included natural secondary forest, artificial forests, shrubland, grasslands, and cropland. In 118 order to increase the comparative study and detect more apparent trends in soil C, the soil layer was 119 divided into depths of 0-10, 10-20, 20-30, and 0-100 cm. The recovery time was divided into four 120 groups to depict more apparent trends: 1-6, 7-12, 13-18, and ≥ 19 years, and the type of vegetation 121 used for recovery was divided into five different groups: cropland to natural secondary forest (CN), 122 cropland to artificial forest (CA), cropland to shrubland (CS), cropland to grassland (CG), and 123 cropland to cropland (CC). 124 We used a "space for time" substitution approach, in which all of the data used in the analysis 125 were designed using a paired site method. We made the assumption that the soil conditions between 126 paired sites were similar prior to changes in land use (Don et al. 2011 To explore trends in vegetation productivity over the past two decades in the study area, the 143 spatial distribution and annual variation in gross primary productivity (GPP) and net primary 144 productivity (NPP) were assessed. GPP and NPP datasets were refined from MOD17A2 and 145 MOD17A3, respectively (http://files.ntsg.umt.edu/). The MODIS GPP and NPP datasets used in 146 this study were improved by Zhao et al. (2005) to reduce uncertainties from upstream inputs, and 147 are available at a spatial resolution of 1-km over annual intervals from 2000-2015. The GPP and 148 NPP data were used to analyze the temporal and spatial trends in these parameters using the per-149 pixel unary linear regression model. In cases where SOM was measured but SOC was not, a correcting factor of 0.58 was used to 153 convert organic matter into soil C. The relationships between SOM, SOC, and C stock were 154 calculated using the following formulas (Guo and Gifford 2002): 155 where SOC is soil organic C concentration (g kg -1 ); SOM is soil organic matter (g kg -1 ); Cs is soil 158 organic C stock (Mg ha -1 ); BD is bulk density (g cm -3 ) and D is the thickness of the soil (cm). 159 In cases where soil BD was not measured, two methods were used to estimate these data. The where Y describes the cumulative proportion of soil C stock from the soil surface to depth d (cm); β 176 is the relative rate of decrease in the soil C stock with depth; X 100 is the soil C stock in the upper 177 100 cm; d0 is the original soil depth available in individual studies; and 0 is the original soil C 178 stock. The global average depth distributions for C were used to calculate β (i.e., 0.9786) in these 179

equations. 180
The soil C sequestration and the rate of soil C change were calculated using the following 181 equations, respectively: 182 where ∆C S is the soil C sequestration (Mg ha -1 ), C LU2 represents the soil C stock at the 185 experimental sites and C LU1 is the soil C stock of the reference sites; is the rate of soil C change 186 (Mg ha -1 yr -1 ); and ∆Age represents the time since the change in land-use (yr). 187 Next, we explored the bivariate relationships between the selected soil properties, climate 188 variables, and elevation using Pearson's correlation coefficient. Data were log10 transformed to meet 189 normality assumptions prior to building the variable correlation matrix (Table 1). A 2-sample t-test 190 was then performed to analyze differences in soil C content between karst and non-karst regions.

Soil C dynamics under vegetation restoration 208
A total of 31.6 ×10 6 ha of land has been restored in the study area. Of this restored land, 13.6 209 ×10 6 ha is composed of karst land and 18 ×10 6 ha of on-karst land. (Fig. S4). Our results indicate 210 that soil C accumulation differs markedly between karst and the non-karst regions. SOC and N 211 contents were 58.4% and 75.5% higher in karst areas than non-karst areas. Soil C:N and N:P in karst 212 were higher compared with non-karst regions, and C:P ratio was the opposite. Both P and elevation 213 were 12.3% and 36.8% lower in karst areas than non-karst areas (Table 1). We found that karst has 214 higher rate of soil C change than non-karst region (i.e. the Rs of karst and non-karst were 6.17 Mg 215 ha-1 yr-1and 4.71 Mg ha-1 yr-1 in 0-100 cm, respectively.), and overall Rs irrespective of vegetation 216 type, with a mean rate of 1.24 Mg ha-1 yr-1, 0.73 Mg ha-1 yr-1, 0.68 Mg ha-1 yr-1 and 5.14 Mg ha-217 1 yr-1 in 0-10, 10-20, 20-30 and 0-100 cm, respectively (Fig. 2a). The sequestration of soil C in the 218 karst areas was always higher than in non-karst regions, regardless of which soil layers were 219 examined (Fig. 2b). Specifically, the 0-30 cm of topsoil accounted for more than 50 % of Rs, and 220 soil C sequestration in the total soil column (0-100 cm), and C accumulation is higher in karst than 221 it is in non-karst topsoil at 0-10, 10-20 and 20-30 cm depths (Fig. 2). 222 We found important changes in Rs associated with restoration age and vegetation type. Rs 223 generally increased with restoration age before decreasing, with similar temporal patterns observed 224 across the different soil layers (Fig. 3). However, this was only observed for the period of 13-18 225 years, and no differences were observed for the other time periods examined. The greatest Rs values 226 from karst and non-karst regions occurred at 13-18 years post-recovery time of (2.09, 1.62, 1.26, 227 and 9.18 Mg ha -1 yr -1 at a depth of 0-10, 10-20, 20-30, and 0-100 cm, respectively) and 7-12 years

Drivers of soil C sequestration 237
The SEM analysis revealed that Rs was mediated by different factors in karst versus non-karst 238 regions. Together, the predictor variables explained 43% and 37% of the spatial variation in Rs, were not significant in the karst region. The direct effects of MAT and N on Rs decreased from 0.33 257 to 0.30 and 0.52 to 0.50, respectively, in the non-karst regions (Fig. S6-S8). 258 259

Relationship between soil C dynamics and vegetation productivity 260
To explore the effects of the environment on soil C dynamics, interannual trends in temperature 261 and precipitation from 108 meteorological stations were gathered and correlated with regional 262 productivity. Although there were fluctuations in MAP, we observed no significant variation, 263 however, MAT significantly increase (Fig. S9). We observed no significant variation in vegetation 264 productivity, measured as GPP and NPP, between karst and non-karst regions (Fig. S10-S11). increasing rates of soil C accumulation when moving from cool temperate climates to subtropical 285 regions. They also found that soil organic C in tropical and subtropical zones was higher than that 286 in temperate zones. In our study, Rs and C sequestration in the 0-30 cm topsoil layer accounted for 287 more than 50 % of that the total 0-100 cm soil column (Fig. 2). These results provide evidence that 288 the upper soil levels sequester more C than deeper soils, which is largely in line with previous studies 289 conducted on this topic Jobbágy and Jackson (2000). 290 Generally, soil C accumulation shows a regular temporal pattern following vegetation Consistent with previous research, our study also found that the Rs showed an initial rise and then 296 a gradual return of soil C to pre-vegetation values coincident with vegetation restoration (Fig. 3). C sequestration than that of non-karst regions. Although Rs was overall higher in the non-karst 301 regions than in the karst regions during the early recovery period, the difference was not significant. 302 Rs was also significantly higher in karst regions than in non-karst regions during the 13-18 year 303 post-recovery period. Rs in the karst and non-karst regions were significantly different in CS and 304 CG compared with CN and CA (Fig. 4). Higher C sequestration from karst regions is likely a result 305 of the greater plant inputs and/or due to lower losses of C at high levels of plant diversity, and it is 306 well known that greater species diversity is linked to higher plant productivity, which likely results 307 in greater soil C sequestration (Prommer et al. 2020). In addition, cracked land surfaces often form 308 in the limestone-dominated calcium carbonate typically found in karst regions. This results 309 primarily from the dissolution of water, where plant roots are able to penetrate more deeply into the 310 soil through the cracks, which leads to a species-rich plant community and a more complicated We also noticed a net increase in soil C storage by CC over the previous two decades, and the 313 Rs of CC was significantly higher in karst regions than that in non-karst regions (Fig. 4). The 314 increase in soil C is likely due to the application of crop straw in these regions, as well as to different 315 fertilizer regimes used in the area. The phenomenon of increases in soil C is probably due to the 316 activities of crop straw application and fertilizer recommendation technique by the Agricultural 317 Ministry of China since the 1990s (e.g. no-till and rotation agriculture represents a relatively widely 318 factor inhibiting soil C sequestration in the study area, while precipitation has little to no effect. The results of our study also indicate that N, P, and BD play an important role in soil C sequestration. 360 In our model, the proportion of the total positive effect of N on Rs was relatively higher in both karst 361 and non-karst regions (Fig. 5) The soil in tropical and subtropical regions are often N-rich and P-limited, and this may lead 366 to P-limitation in the long-term deposition of N (Alvarez-Clare et al. 2013). According to our 367 synthesis, the C:N and N:P in karst was higher than in non-karst regions. This implies that soil 368 mineralization is lower in karst area with lower N limit than non-karst areas resulting in higher 369 potential soil carbon sequestration. Notably, the C:P ratio was relatively lower, which indicates that 370 the higher soil C sequestration in karst areas leads to lower P content, and the soil has higher P 371 availability. P may also be a key factor controlling ecosystem processes that depend on N saturation, 372 P addition can alleviate the limitation of soil C sequestration due to N saturation (Chen et al. 2016). 373

Chen et al. (2018) found that P limitation was more evident in non-karst forests than in karst forests, 374
because karst regions are more likely saturated with N. Our model suggests that compared to non-375 karst regions, Rs in karst regions is more often P-limited because of N saturation (Fig. 5), while non-376 karst regions have a relatively lower rate of change in N than in Rs. Therefore, Rs in non-karst 377 regions may be less negatively affected by P-driven N enrichment than that of karst regions (Fig. 6). 378

Conclusion 379
This study found a significant difference in the soil C sequestration between karst and non-380 karst regions, with faster and more persistent C sequestration in karst region. This finding is due 381 primarily to climate gradients and to the amount of N present within the soil. Both different patterns 382 of soil C dynamics following vegetation restoration in the karst and non-karst support the found that 383 climate gradients are largely controlled by topographic conditions, and that the increase in 384 temperature that has occurred over the past few decades in southwestern China may have led to 385 limit soil C sequestration in non-karst regions. In addition, P is the dominant factor limiting the use 386 of N in karst regions and then resulting in limitation of C sequestration. At the regional scale, climate 387 factors play an important role in carbonate dissolution in karst environment. And then it is concluded 388 that soil C storage could be led to intensify uneven increases due to combination of karst 389 environment and climate change in southwest China in future.   cm. The symbols *, **, and *** denote values where there are significant differences between the 590 karst and non-karst regions, at p < 0.05, p < 0.01, and p < 0.001, respectively. Different uppercase 591 letters denote a significant difference among the different restoration stages of the karst at p < 0.05. 592 Different lowercase letters denote a significant difference between the different restoration stages 593 in the non-karst regions at p < 0.05. Values above the bars represent the number of observations. 594 The error bars illustrate the standard errors (SE). 595 soil at 0-100 cm. The symbols *, **, and *** denote values where significant differences occur 598 between the karst and non-karst regions, at p < 0.05, p < 0.01, and p < 0.001, respectively. Different 599 uppercase letters denote significant differences among the different restoration stages of karst at p 600 < 0.05. Different lowercase letters denote significant differences between the different restoration 601 stages of non-karst at p < 0.05. Values above the bars represent the number of observations. The 602 error bars illustrate the standard errors (SE). 603  Spatial distribution of the observation sites in the dataset. All sites include multiple data entries. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.  Variation in the rate of soil C change over time within different soil layers after vegetation restoration for: (a) soil at 0-10 cm, (b) soil at 10-20 cm, (c) soil at 20-30 cm, and (d) soil at 0-100 cm. The symbols *, **, and *** denote values where there are signi cant differences between the karst and non-karst regions, at p < 0.05, p < 0.01, and p < 0.001, respectively. Different uppercase letters denote a signi cant difference among the different restoration stages of the karst at p < 0.05. Different lowercase letters denote a signi cant difference between the different restoration stages in the non-karst regions at p < 0.05. Values above the bars represent the number of observations. The error bars illustrate the standard errors (SE).

Figure 4
Variation in the rate of soil C change within different vegetation types at different depths after vegetation restoration for: (a) soil at 0-10 cm, (b) soil at 10-20 cm, (c) soil at 20-30 cm, and (d) soil at 0-100 cm. The symbols *, **, and *** denote values where signi cant differences occur between the karst and nonkarst regions, at p < 0.05, p < 0.01, and p < 0.001, respectively. Different uppercase letters denote signi cant differences among the different restoration stages of karst at p < 0.05. Different lowercase letters denote signi cant differences between the different restoration stages of non-karst at p < 0.05.
Values above the bars represent the number of observations. The error bars illustrate the standard errors (SE).   A conceptual diagram showing the differences mechanism in soil C sequestration processes between karst and non-karst areas. The diagram showing different lithology in karst and non-karst areas leading to different distribution of underground soil layers and aboveground vegetation community composition.
BD has a negative effect on soil carbon pool growth throughout the study area. MAP and P in karst areas offset negative impacts on soil carbon pool growth due to N saturation. Growth of soil C pools in nonkarst areas limited by temperature rise and N increase. MAT: mean annual temperature; MAP: mean annual precipitation; BD: soil bulk density; N: soil nitrogen; P: soil phosphorus. The upward, downward, and horizontal arrows near the ellipse represent increase, decrease, and no change of the corresponding variables, respectively. The plus and minus signs next to the arrows between the variables indicate the positive and negative effects, respectively

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.