Land Use Types and Geomorphic Settings Reflected in Soil Organic Carbon Distribution at the Scale of Watershed

Soil organic carbon (SOC) is vital to soil ecosystem function and it plays a key role in carbon cycling in the terrestrial ecosystem. The spatial pattern of SOC stock (SOCs) is affected by specific geomorphic settings and land-use types at the scale of watershed. Nevertheless, the distribution of SOCs with fluvial landform regimes and land use types was not sufficiently elucidated in the semi-humid riparian ecosystem in north China. In this study, 103 soil plots were sampled and spatial auto-correlation method was adopted to detect the spatial pattern of SOCs in the Changhe watershed that was located at the boundary of the Loess Plateau and the Taihang Mountains. The results showed that SOCs in the Changhe watershed varied from 18.03 Mg ha−1 to 21.51 Mg ha−1 and it was in the order: grassland > forestland > cropland > construction land. SOCs varied with geomorphic settings, among which, the altitude exerted more influence on the distribution of SOCs than the aspect and the slope. In terms of the spatial pattern of SOCs, 17 plots with higher SOCs collectively distributed in the west of the watershed and that with lower SOCs (19 plots) concentrated in the midlands. This indicated that the upland had higher SOCs while the lowland had lower values. Overall, land use type and geomorphic settings (especially the altitude) should be considered when estimating the SOC sequestration in warmer and wetter watershed in north China. With regard to the implications for land use management, reforestation could elevate the SOCs. Moreover, no-tillage and returning crop straw to cultivated soils could be efficient approaches to elevate soil carbon sequestration and soil productivity.


Introduction
The dividing crest separates most of biotic and abiotic ecosystem elements from both sides of the watershed.Thus, watershed is characterized as an independent ecological system [1].From this point of view, watershed is regarded as an optimal geographic unit to observe ecological processes [2,3].Generally, watershed landscape consists largely of the riparian scenery which exerts substantial influence on geomorphic, hydrologic and ecological processes of the watershed [4,5].Moreover, the hydro-geomorphic regime of the riparian ecosystem underpin the spatial distribution of the material and energy throughout the watershed [6].For example, the morphology and runoff characteristics of the watershed not only dictate alluvial aquifer recharge and subsurface hydrology regime [7,8], but determine the allocation of aboveground soil properties [9,10].It is more challenging to simulate the distribution of soil properties in the watershed than in flat areas because water and heat condition vary river and the channel length is 14.5 km.Changhe watershed has an area of 113.19 ha.The dominant topography is mountain and hill with an average elevation of 880 m.It has a temperate, continental monsoon climate with an average annual precipitation of 628 mm and an annual evaporation of 1700 mm.The temperature averages 11 • C. The soil type is characterized by cinnamon soil.Moreover, the land use types are dominated by cropland (6863.99ha; 60.62% of the total area), grassland (2133.7 ha; 18.84%), construction land (1320.89ha; 11.67%) and forest land (893.25 ha; 7.89%) (Figure 1c).In addition, the construction land was used for rural building.
Changhe watershed has a higher cultivated rate due to the excellent water and heat regimes.To date, more than 60 percent of the land is cultivated for agricultural purposes.The cropland is predominantly terrace, which protects the soil from water erosion.Nevertheless, areas with rugged terrain were implemented with the reforestation engineering for decades.The cropping system is "three crops, 2 years" with the rotation of Zea mays and Triticum aestivum.The tree species are Populus Linnaeus, Robinia pseudoacacia Linnaeus and Pinus tabulaeformis Carrière.and herbaceous community mainly consists of Setaria viridis Beauv, Artemisia annua Linnaeus and Artemisia gmelinii.1c).In addition, the construction land was used for rural building.Changhe watershed has a higher cultivated rate due to the excellent water and heat regimes.To date, more than 60 percent of the land is cultivated for agricultural purposes.The cropland is predominantly terrace, which protects the soil from water erosion.Nevertheless, areas with rugged terrain were implemented with the reforestation engineering for decades.The cropping system is "three crops, 2 years" with the rotation of Zea mays and Triticum aestivum.The tree species are Populus Linnaeus, Robinia pseudoacacia Linnaeus and Pinus tabulaeformis Carriè re. and herbaceous community mainly consists of Setaria viridis Beauv, Artemisia annua Linnaeus and Artemisia gmelinii.

Soil Sampling
The soil samples were collected in July 2016.One-hundred-and-three soil sampling sites were distributed as a grid with a distance of 1 km between the two adjacent sites (Figure 1c).The coordinate, altitude, slope, and aspect of each soil sampling sites were recorded when sampling.Five random sampling locations were identified within each sample site and soil samples were collected at 0-20 cm depths of each location after litter and fermentation being removed.All of the soil from each plot was mixed to make the composite sample.After removing all visible roots and fresh litter material, soil samples were air-dried and sieved through a 100-mesh for the analysis of

Soil Sampling
The soil samples were collected in July 2016.One-hundred-and-three soil sampling sites were distributed as a grid with a distance of 1 km between the two adjacent sites (Figure 1c).The coordinate, altitude, slope, and aspect of each soil sampling sites were recorded when sampling.Five random sampling locations were identified within each sample site and soil samples were collected at 0-20 cm depths of each location after litter and fermentation being removed.All of the soil from each plot was mixed to make the composite sample.After removing all visible roots and fresh litter material, soil samples were air-dried and sieved through a 100-mesh for the analysis of SOC, nitrogen (N), phosphorus (P), potassium (K) and pH.In addition, soil samples for bulk density (BD) were obtained using stainless steel cylinders (5.0 cm diameter and 5.0 cm high).

Analyses of Soil Properties
The method of dry combustion was applied to determine the SOC [35].pH was determined in H 2 O suspension (soil: water ratio of 1:2.5) with a pH Meter (FE20K, Mettler Toledo, Zurich, Switzerland).BD and field water capacity were calculated by the oven-dried method [36].Available nitrogen (AN) was determined using the zinc-cadmium reduction method [37] and available phosphorus (AP) was determined using the Olsen method [38].Moreover, available potassium (AK) was measured using a flame photometric method [37].Finally, soil coarse fragments were analyzed using Longbench Mastersizer 2000 laser particle-size analyzer (Malvern Instruments, Malvern, England).

Calculation of SOCs
SOCs was calculated using Equation ( 1) where S d was the SOCs (Mg ha −1 ); C c was the content of SOC (g kg −1 ); BD represented the soil bulk density (g cm −3 ); D was the soil layer thickness (cm); and α was the proportion of coarse fragments > 2 mm in soils.

Spatial Auto-Correlation Method
Spatial auto-correlation method was utilized to detect whether the distribution of SOCs was clustered or random via ArcGis (Ver9.3,ESRI, Redlands, CA, USA).The global spatial auto-correlation reflected the spatial correlation among soil carbon stocks at the scale of the watershed, while local spatial auto-correlation further revealed the relationship of SOCs among geographic units.Global spatial auto-correlation was expressed by the global Moran s I index and local spatial auto-correlation by local Moran s I index.Moran s I index ranged from −1 to 1.If Moran s I = 0, it indicates a randomness spatial pattern.If Moran s I > 0, it denotes a positive spatial correlation, while Moran s I < 0 means a negative correlation.In addition, the significance of spatial relationship was detected by Z(I).Z(I) > 1.96, indicating a significant auto-correlation among SOCs [39].
The local spatial auto-correlation could be represented by the local indicators of spatial association (LISA) map.In the map, sites marked with High-High (HH) and Low-Low (LL) indicated it was where the higher or lower SOCs values were clustered, respectively.Whereas the High-Low (HL) and Low-High (LH) types indicated negative spatial auto-correlations.
The calculation of global Moran s I index was: The local Moran s I index was: where N was the amount of the objects, X i denotes the observed value, X indicates the mean value of X i and W(i,j) was the spatial connection matrix between i and j.
One-way ANOVA was adopted to compare the mean values of soil physicochemical properties among land-use types in SPSS 16.0 (SPSS Inc., Chicago, IL, USA).The relationship between SOC stock and terrain factors (elevation, aspect and slope) were detected by boosted regression trees (BRT) analysis using modified "GBM" packages [40] in R 3.3.2.All statistical significance was determined at p < 0.05.

Physicochemical Properties of Changhe Soils
Generally, the cropland had highers amount of AK, AP and AN than other land use types (Table 1), among which AK varied from 37.84 to 44.99 mg kg −1 , AP from 27.12 to 35.91 mg kg −1 and AN from 0.72 to 0.78 g kg −1 .Changhe soils were characterized as alkaline with pH fluctuating around 8. No significant differences were observed in BD and field water capacity among land use types (p > 0.05): BD had a range of 1.27 to 1.35 Mg m −3 and field water capacity ranged from 16.02 to 16.59%.Nevertheless, significant difference (p < 0.05) was obtained in clay content among soils: Construction land had the highest fraction (31.45%), whilst forestland had the lowest (23.15%).

The Distribution of SOCs with Land Use Types and Geomorphic Settings
SOCs varied with land-use types and geomorphic settings.Grassland (21.51 Mg ha −1 ) had the highest SOCs, followed by forest land (20.88 Mg ha −1 ) and cropland (19.27Mg ha −1 ).Construction land (18.03 Mg ha −1 ), however, had the least amount of SOCs (p < 0.05) (Table 2).In terms of the relationship between SOCs and geomorphic settings, an increasing trend of SOCs with altitude was observed: Areas with elevation >920 m stored more SOCs than that with 740-800 m and 800-860 m.Moreover, higher SOCs sequestrated where the slope was 11-15 • and >15 • whereas the lower fractions stored where the slope was 0-5 • and 5-10 • .In addition, the north and southwest aspect of the watershed had higher fraction of SOCs than other aspects (Figure 2).In terms of the influence exerted by geomorphic factors to SOCs, the elevation explained 43.1% of the SOCs variance, followed by the slope (42.9%) and the aspect (14.1%) according to BRT (Figure 3).

The Distribution of SOCs in the Changhe Riparian Ecosystem
The global Moran′s I indexes of SOCs was 0.26 and Z(I) was 3.55, indicating that the SOCs had significant auto-correlation in spatial and the SOCs had a clustered distribution in the Changhe watershed (Table 3).
Generally, spatial patterns of SOCs were associated with geomorphic settings (Figure 4).There were 17 High-High (HH) plots that collectively distributed on the west of the watershed, while 19 Low-Low (LL) plots were on the midland (along the Changhe channel).These indicated that plots with higher SOCs intensively distributed in the west of the watershed while that with lower values concentrated in the midland.From the point of altitude-related gradients, the upland had a higher stock of soil carbon while the lowland had lower values.3).
Generally, spatial patterns of SOCs were associated with geomorphic settings (Figure 4).There were 17 High-High (HH) plots that collectively distributed on the west of the watershed, while 19 Low-Low (LL) plots were on the midland (along the Changhe channel).These indicated that plots with higher SOCs intensively distributed in the west of the watershed while that with lower values concentrated in the midland.From the point of altitude-related gradients, the upland had a higher stock of soil carbon while the lowland had lower values.

The Distribution of SOCs in the Changhe Riparian Ecosystem
The global Moran′s I indexes of SOCs was 0.26 and Z(I) was 3.55, indicating that the SOCs had significant auto-correlation in spatial and the SOCs had a clustered distribution in the Changhe watershed (Table 3).
Generally, spatial patterns of SOCs were associated with geomorphic settings (Figure 4).There were 17 High-High (HH) plots that collectively distributed on the west of the watershed, while 19 Low-Low (LL) plots were on the midland (along the Changhe channel).These indicated that plots with higher SOCs intensively distributed in the west of the watershed while that with lower values concentrated in the midland.From the point of altitude-related gradients, the upland had a higher stock of soil carbon while the lowland had lower values.

Discussion
The SOC is not only vital to the carbon cycling in the terrestrial ecosystem [23,41] but is a key element for improving the process of soil formation (supporting service) [42] and for the rejuvenation of chemical and physical soil properties at watershed [20].Nevertheless, SOC is sensitive to the environment because it is mainly from litter, plant roots, microbial residues, soil animals and their excreta [24,43,44].Authors paid attention to detecting the mechanism of SOC accumulation [45].Some documented that nitrogen enrichment resulted in SOC sequestration [46].Others argued that SOC rejuvenation could be due to soil acidification [47].
Despite the mechanism of SOC sequestration remaining to be explored, the authors agreed that land-use types dictated the sequestration of SOC [48].This was confirmed by our findings: Grassland (21.51 Mg ha −1 ) and forestland (20.88 Mg ha −1 ) had a higher amount of soil carbon than construction (18.03Mg ha −1 ) and cropland (19.27Mg ha −1 ) in the Changhe Watershed.This indicated that grassland and forestland had a stronger carbon sequestration ability than other land-use types.The discrepancy of SOCs among land-use types was coupled with previous results [31] and it could be associated with several Generally, much of the SOC was derived from plants and particularly their roots [49].Grass and forest ecosystems had more net primary production (NPP) and more organic material input into soils and less human-derived disturbance, which resulted in the enrichment of SOCs [41,50].According to Bojko and Kabala [51], the mountain pine soils had 7% of total organic carbon and grass soils had 3%.Both of them were higher than the 1.6% of SOC in arable soils.Just in contrast, active human-induced disturbance to soil layer gave rise to the mineralization of organic carbon in cropland and construction land [52,53].It had a long history of cultivation and high cultivation rate in Changhe watershed and the "three crops, 2 years" land management also exerted high stress to soils, which resulted in the loss of SOCs.To rejuvenate soil carbon in cropland, conservation cropping management, such as no-tillage could be expected.In addition, straw was used to be burnt in the farmland after crops were harvested in the Changhe region.Thus, returning crop straw to cultivated soils could be an effective approach to elevate soil carbon sequestration and soil productivity.
Geomorphic settings directly affect biogeochemical cycles and other biological processes in the riparian ecosystem [5,51].Authors found that soil salinity migrated from upstream to downstream within the period of 1982-2015 in Sangong River watershed [18].Others found that greater abundance of obligate riparian taxa and increasing structural importance of shrub and tree species appeared in the downstream direction where increased moisture availability was observed [5].In terms of the SOCs, spatial heterogeneity with the higher altitude (>920 m a.s.l.) had higher values while the areas with altitude of 740-800 m had lower values in the Changhe watershed riparian ecosystem.It was confirmed by the result of the BRT: The upland had higher stock of soil carbon while the lowland had lower values.These observations could be coupled with the previous conclusion as well: The SOCs increased with altitude gradients.Zhao et al. found that SOCs was significantly correlated with altitude at the level of p < 0.05 [31].Similarly, Leifeld et al. suggested that SOC increased from 0.75 to 2.10 mg/g as altitude increased by 100 m [54].Some suggested that higher elevation gave rise to lower temperature, which restrained the deformation of SOC [51,55,56].For example, in Karkonosze Mountains, Czech Republic, the mean temperature was 7.0-8.0• C at 400-500 m a.s.l., while it decreased to 1.3-2.0• C at 1250-1450 m a.s.l., and the precipitation increased from 650-700 to 1400-1500 mm, accordingly.Consequently, the total organic carbon increased from 1.8% (<500 m a.s.l.) to 5.4% (>1250 m a.s.l.) [51].Generally, geographic settings affected the distribution of SOCs in two ways: (1) Through the effect of temperature and (2) through the land-use types.Actually, land-use types were determined by geomorphic setting to some extent.As mentioned above, the forestland and grassland were the dominant land-use types at rugged terrains with higher elevation, while the cropland and construction land distributed in lowland with flat terrain.
The geomorphic characteristic in Changhe watershed resulted in another intriguing observation: SOCs increased with slope gradients [51,57].Actually, areas with higher slope were covered with wood and grass due to the reforestation engineering that was implemented several years ago, which resulted in the observation mentioned above.In terms of the relationship between SOC and aspect, no significant difference was observed.Comparably, authors found that the aspect had different effects on the distribution of SOCs, for example, Zhao et al. found that sunny aspect stored more SOC in sloped cropland while the shady aspect had more SOCs in grassland [31].Compared to watershed in semi-arid Loess Plateau areas that had sparse vegetation and fragmented geomorphic settings, Changhe riparian ecosystem had higher vegetation canopy (it had ~30% forest and grass land) and less soil erosion.Consequently, SOC sequestration and spatial distribution were different.For example, the SOCs of 18.03 Mg ha −1 to 21.51 Mg ha −1 in Changhe watershed was higher than that in the semi-arid Loess Plateau [58].Specifically, just as mentioned above, watershed was an independent ecological system and it was an optimal geographic unit to observe ecological processes such as carbon cycling.To date, the spatial pattern of soil carbon in watershed was accessible.However, the detection of the transported carbon in the process of carbon cycling genuinely frustrated the authors.E et al. suggested that rainfall events played an important role in carbon cycling and 2.7 kg ha −1 year −1 carbon was observed lost from soils in Yangjuangou watershed [59].Nevertheless, advance experiments were expected to develop to detect the process of carbon cycling in detail in watershed.Overall, altitude had a bigger influence on the distribution of SOCs than slope and aspect, which was confirmed by other studies [31].This indicated that altitude could be a fundamental factor to estimation the spatial pattern of SOCs in the present study area.

Conclusions
Soil organic carbon was important to soil ecosystem function and it was affected by land use types and geographic settings in Changhe watershed riparian ecosystem.SOCs varied from 18.03 Mg ha −1 to 21.51 Mg ha −1 in Changhe watershed, among which, the grassland and forestland had higher SOCs than cropland and construction land.Spatially, SOCs differed with slope, aspect and especially the altitude.From the point of altitude-related gradients, the upland had higher values of SOC than the lowland.Overall, land-use type and terrain factors should be considered when estimating the storage of soil carbon in warm and wetter watershed.With regards to the implications for land use management, reforestation could elevate the SOC storage.No-tillage and returning crop straw to cultivated soils could be effective approaches to elevate soil carbon sequestration and soil productivity.More specifically, more advanced experiments were expected to make the carbon cycling understood at the scale of the watershed.

Figure 1 .
Figure 1.Location and landscape of the Changhe Watershed.(a) Geographic location of Changhe watershed; (b) elevation of the study area; (c) land-use types and soil sampling sites in Changhe Watershed; (d,e) Landscape of the study area.

Figure 1 .
Figure 1.Location and landscape of the Changhe Watershed.(a) Geographic location of Changhe watershed; (b) elevation of the study area; (c) land-use types and soil sampling sites in Changhe Watershed; (d,e) Landscape of the study area.

Figure 3 .
Figure 3. Influence of the elevation, slope and aspect to SOCs via BRT in the Changhe riparian ecosystem.

Figure 3 .
Figure 3. Influence of the elevation, slope and aspect to SOCs via BRT in the Changhe riparian ecosystem.4.1.The Distribution of SOCs in the Changhe Riparian Ecosystem The global Moran s I indexes of SOCs was 0.26 and Z(I) was 3.55, indicating that the SOCs had significant auto-correlation in spatial and the SOCs had a clustered distribution in the Changhe watershed (Table3).Generally, spatial patterns of SOCs were associated with geomorphic settings (Figure4).There were 17 High-High (HH) plots that collectively distributed on the west of the watershed, while 19 Low-Low (LL) plots were on the midland (along the Changhe channel).These indicated that plots with higher SOCs intensively distributed in the west of the watershed while that with lower values concentrated in the midland.From the point of altitude-related gradients, the upland had a higher stock of soil carbon while the lowland had lower values.

Figure 3 .
Figure 3. Influence of the elevation, slope and aspect to SOCs via BRT in the Changhe riparian ecosystem.

Figure 4 .
Figure 4.The distribution of SOCs in the Changhe watershed.In the map, plots marked with High-High (HH) and Low-Low (LL) indicates the plots were where higher or lower SOCs was clustered.

Table 1 .
Summary statistics for soil physicochemical in the Changhe Watershed.
Values are in the form of: Means ± Standard Errors (SE).Values with significant differences (ANOVA, LSD post-hoc test, p < 0.05) are distinguished by the different letters.

Table 2 .
Summary statistics of SOCs in the Changhe riparian ecosystem.Values with significant differences (ANOVA, LSD post-hoc test, p < 0.05) are distinguished by different letters.

Table 2 .
Summary statistics of SOCs in the Changhe riparian ecosystem.Values with significant differences (ANOVA, LSD post-hoc test, p < 0.05) are distinguished by different letters.

Table 3 .
Moran′s I Index of soil carbon in the Changhe watershed.

Table 3 .
Moran s I Index of soil carbon in the Changhe watershed.

Table 3 .
Moran′s I Index of soil carbon in the Changhe watershed.