Evaluating the Evolution of Soil Erosion under Catchment Farmland Abandonment Using Lakeshore Sediment

Abstract

Discriminating the potential sources contributing to lacustrine sediment is helpful for decision-making for catchment soils and lake management strategies within lake-catchment systems. Using a sediment fingerprinting approach from the multivariate mixing model, the spatiotemporal sources of geology and land use were identified in a small agricultural catchment in southwest China. Results showed that sediment accumulation rates (SARs) were estimated to range from 0.002 to 0.065 g cm⁻² a⁻¹ (mean 0.015 ± 0.016 g cm⁻² a⁻¹), which has a positive correlation with instrumental Indian Summer Monsoon (ISM) precipitation. Time-integrated sources were divided into four zones in combination with the changes in SARs, which were qualitatively and quantitively interpreted by particle size, and precipitation, and historical land use polies over the past ~160 years. Spatially, Quaternary granite (QG) in geology and channel bank (CB) in land use were the dominant contributors to the lakeshore sediment, respectively. Two relatively higher contributions of abandoned land (AL) to lakeshore sediment were found during the periods of 1930s–1950s and post-1990s, which originated from the dual impacts of topographical factors of slope gradient and elevation, and socioeconomic factors of the gap of farmer’s expenditure to income. The results illustrate that restricting the farmland to be abandoned would be useful for reducing the soil erosion within the lake-catchment system.

1. Introduction

Soil erosion is a severe environmental issue that impacts sustainable land management in the land-lake interface [1]. Understanding the relationship between sustainability and soil erosion is critical for tacking the global challenges, such as climate change, food security, and biodiversity loss [2]. The transition of agricultural land to farmland abandonment at a basin scale provides a potential threat to magnify the magnitudes of soil erosion within the catchment [3], thereby resulting in increasing fluxes to lakes [4,5]. Variation in land use types within the lake-catchment system is of great importance to understanding lake environmental changes [6,7]. Lakes are a sink that receive the erosional materials from river catchments via the processes of soil erosion, delivery, and accumulation [8,9,10]. Lacustrine sediments, in which stable, continuous, and high-resolution records are contained, can be regarded as excellent archives for identifying the erosional sources under diversified land use patterns over time [7,11]. Discerning the dominant sources of sediment is important for improving the sustainability of soil management in river catchments and reducing the negative effects of downstream increased siltation in lakes [1].

Sediment fingerprinting is an effective method that can identify the relative contributions from each potential source to sediment loads [12,13,14,15,16]. The fundamental theory of this method is to compare the diagnostic physicochemical characteristics of sediment with their corresponding sources for passing three constraints of mass-conservation, Kruskal–Wallis (KW) test, and a stepwise discriminant function analysis (DFA), using a multivariate mixing model [17]. Of these physical–chemical proxies, fallout radioactive nuclides (¹³⁷Cs, excess ²¹⁰Pb) are often selected as the primary fingerprinting properties, based on their accurate estimation of sediment chronology dating [18,19]. Combined with other physical–chemical proxies, such as stable isotope, major/minor geochemical elements, biomarkers, and organic/inorganic nutrient concentrations, a compositing fingerprinting technique provides an opportunity to quantify synthetic contribution of different sediment sources, whilst potentially identifying the long-term change in sediment delivery [14,20,21,22,23,24,25,26]. This approach has been applied successfully to agricultural catchments with land use and land cover change (LUCC) [15,17].

Farmland abandonment was defined as cultivated land idle and unused for more than one year [27] and was a widespread process of LUCC in the world [28]. With the complex and multi-dimensional factors of population, geography, history, economy, society, policy, and environmental pollution [29], more and more farmland was abandoned over the past century, which caused a series of impacts on global ecosystem recovery and biodiversity, rural sustainable development, and food security [30,31,32,33,34]. According to statistics, approximately 1.5 million square kilometers were abandoned globally [19]. The published works on farmland abandonment were focused on all continents on earth with the exception of Antarctica [35,36,37,38,39,40]. Meanwhile, the regions of farmland abandonment in the previous studies were mainly located in the mountainous and hilly areas, such as the central Himalayas [41,42], Alps [43], Carpathians [44], and Andes [45]. Mapping, modeling, driving forces, and risk assessment of farmland abandonment in Europe were systematically identified and discussed in comparison with other continents [29,46,47]. Currently, approximately 3–4% of farmland in Europe could still be abandoned in the upcoming decades [38]. More recently, the studies of farmland abandonment have also been conducted in developed countries of other continents, such as the United States [48], South Korea [34], Japan [46,47], and Australia [35]. Natural factors (topography and landform) and human influences (aging population, urbanization, industrialization, markets, and agricultural levels), may all enhance farmland abandonment [35,37,45,47,52]. However, relatively few studies were conducted in the mountainous and hilly regions of developing countries, including China [39].

China has a major land area of 9.6 million square kilometers, approximately 70% of which are distributed in the mountainous regions. During the period of 2000–2010, approximately 28% of croplands in the hilly and mountainous areas were abandoned (including the areas of Grain-for-Green), owing to weak agricultural infrastructure, remote distances, and barren land [39,49]. Both geographical and socio-economic factors are important driving factors for influencing the farmland abandonment [50]. For instance, sloping farmland (>25°) is prone abandonment due to soil degradation by extreme erosion, therefore, the national policy encouraged the farmers to return sloping farmland to forest in mountainous lands [49]. Compared with the incomes of farmers working in cities undergoing rapid urbanization in China, the average net incomes per farmer for cultivating cropland is relatively lower, and therefore, farmers are willing to seek high-income jobs in cities instead of farming in the countryside [39]. Numerous studies have qualitatively concluded the driving mechanisms of farmland abandonment in mountain city of Chongqing [49,51], Three Gorges Reservoir Area [52], Yunnan-Guizhou Plateau [53], and Tibet Plateau [54]. However, there is a lack of insight quantitatively evaluating the proportion of farmland abandonment contributing to downstream sediments at a basin scale in the plateau lake-catchment system.

Lake Fuxian, a deep plateau lake, is situated in southwestern China. With the rapid development of socioeconomic factors within the catchment, the spatiotemporal LUCC was changed during the period of 1974–2014 [55]. Meanwhile, the trophic status of the lake was found to be changed along with the variation in sediment accumulation rates (SARs) in Fuxian Lake over the past ~100 years [56]. In order to further improve the water quality of the inflow rivers, different pollution control zones within the catchment were also established on the basis of the spatial variation in topography and land-use patterns within the catchment [57]. Serious soil erosion was focused on the east-west banks and some regions of the south bank of the Fuxian Lake based on digital elevation model (DEM) and remote sensing (RS) monitoring method [58,59], but there is a lack of investigation of the spatiotemporal sources of geology and land use types contributing to the lacustrine sediment. Such information can offer valuable insights for designing effective management of land use at a basin scale within the catchment of highland lakes.

In this context, the West-East Luju River catchment on the south bank of the deep plateau Fuxian Lake in southwestern China was selected as the study area, and the objectives of this study were: (1) to calculate the SARs in the lakeshore sediment by radionuclide dating (²¹⁰Pb_ex and ¹³⁷Cs), (2) to apportion and differentiate the spatiotemporal contribution of the potential sources between geology and land use types to the lakeshore sediment using a sediment fingerprinting approach, and (3) to estimate the evolution of farmland abandonment under the changes in SARs, particle size, precipitation, and historical land use policies over the past ~160 years.

2. Materials and Methods

2.1. Study Area

The West-East Luju River catchment (102°50′–102°52′ E, 24°18′–24°22′ N) is situated south of Lake Fuxian, a deep plateau lake in southwestern China (Figure 1a). It has a relatively high elevation from 1612 to 2351 m (a.s.l) and an area of ca. 30 km² (Figure 1b). With the impact of Indian monsoon climate, the mean annual temperature and mean rainfall, mainly from May to October, are 15.5 °C and 879 mm, respectively. High precipitation was distributed in the southwestern part of the catchment due to the dual impacts of monsoon and topography. The mean flows of West and East Luju River are 1.39 and 1.03 m³ s⁻¹, respectively [60]. The major soil types within the catchment were formed by the albitization and hydroponically slaking process of parent materials, including Udic Ferrosols (SLC) and Acidi Luvisols (SL) [61].

The mixed land-uses (Figure 1c) within the catchment are abandoned land (AL) with ruderal weeds which are used for grazing livestock (Figure 2a), woodland (WL) with vegetation of pinus yunnanensis (Figure 2b), and cultivated land (CL) with corn (Figure 2c), respectively. Two main channels of the West-East Luju River were distributed in the catchment (Figure 1). The upland geology of the catchment comprised Devonian dolomite (DD), Carboniferous sandstone (CS), and Silurian pebble and sandstone (SPS), whereas Permian basalt (PB) and Quaternary granite (QG) were distributed in the lower catchment, respectively (Figure 1d). Spatially, DD and PB for geology, and WL for land use were mainly distributed in the periphery of the catchment, but QG, SPS, AL, and WL were mainly distributed in the interior of the catchment, respectively (Figure 1d).

At present, farmland abandonment is a common phenomenon in the study area, because local landowners gave up cultivating their cropland but sought work in cities in order to obtain a high economic income. Due to a lack of reasonable management for farmland abandonment, the herbivores can freely enter the abandoned farmland to find food (Figure 2a).

2.2. Sampling

In 2020, representative surface soil samples within the Luju river catchment were collected from 39 sites, which included 11 samples from CS, 5 from DD, 5 from PB, 9 from QG, and 9 from SPS for geological types, and 14 samples from AL, 6 from channel bank (CB), 8 from CL, and 11 from WL for land use types, respectively (Figure 1c,d).

At each sampling site, three or four individual surface (0–2 cm) soil samples were collected at approximately 5 m intervals, using a small, clean plastic trowel for different land use types (Figure 2). For CB sampling, eroding channel reaches were selected for collecting potential source samples that represent the fluvial erosion on sediment storage (Figure 2d). Each individual trio of samples was thoroughly mixed and stored in plastic bags to form a single composite sample. To identify the temporal sources of the affiliated catchment contributing to the lacustrine sediment, a typical sediment core was synchronously collected in the southern lakeshore of Lake Fuxian (Figure 1b). The sediment core was taken to a depth of 34 cm and sectioned at 2 cm intervals.

2.3. Laboratory Analysis

2.3.1. Sample Pre-Treatment

Composite soil and sediment samples were either oven-dried at 40 °C or freeze-dried, depending on the water content. All the dried sediment samples were disaggregated using a mortar and pestle and then sieved to <2 mm for analysis of fallout radionuclides (¹³⁷Cs, ²¹⁰Pb_total, and ²²⁶Ra). After gamma counting, the scatheless samples were taken for analysis of particle size and then sieved to <0.075 mm (200 mesh) to meet the requirement of the X-ray Fluorescence (XRF) for analysis of oxides and minor elements.

2.3.2. Gamma Counting

Approximately 8 g of sediment samples were firstly packed into 8 mL centrifuge tubes then sealed by black insulating tapes for 21 days to reach a radioactive equilibrium between ²²⁶Ra and its daughter ²²²Rn. The ²¹⁰Pb_total, ²²⁶Ra (²¹⁴Pb), and ¹³⁷Cs activities were then determined simultaneously using a high-resolution, low background, low energy HPGe γ-spectrometry system (GWL-120-15, ORTEC, Oak Ridge, TN, USA), which had a 62 % relative detection efficiency. The samples were counted for 40,000 s, with a precision of 5% at the 95% confidence level. The resolution of the spectrometer was 2.25 keV for 1.33 MeV γ-rays from ⁶⁰Co. The ²¹⁰Pb_total, ²²⁶Ra (²¹⁴Pb), and ¹³⁷Cs were measured from their γ-ray emissions at 46.5 keV, 351.9 keV, and 661.6 keV, respectively. The unsupported ²¹⁰Pb (²¹⁰Pb_ex) was calculated by subtracting the ²²⁶Ra-supported ²¹⁰Pb activities (²¹⁴Pb) from the ²¹⁰Pb_total. The detector efficiency was obtained by analysis of standard soil samples with known activity and calibrated by dual institutions of the International Atomic Energy Agency (IAEA) and the Institute of Atomic Energy, Chinese Academy of Science (CAS). The gamma counting analysis was performed at School of Geographical Sciences, Nanjing Normal University.

2.3.3. Particle Size Analyses

To eliminate carbonates and organic matter, the samples for particle size analysis were firstly treated with 5% HCl and 10% H₂O₂ successively over 24 h, and then dispersed with (NaPO₃)₆ by ultrasonication for 15 min. Grain sizes of each sample were measured by a laser optical particle-size analyzer (Mastersizer-2000, Malvern, UK), with a measurement ranging from 0.02 to 2000 μm. Malven’s quality standards of QA3002 glass beads were used to assess the stability and repeatability of the instrument. The distributions of particle size fractions were categorized into three groups: clay (<4 μm), silt (4–63 μm), and sand (>63 μm). The particle size analysis was performed at the State Key Laboratory of Lake Science and Environment, Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences (CAS). The analytical errors for all the measurements were less than 5% at the 95% confidence level.

2.3.4. Measurement of Oxides and Minor Elements

Approximately 5 g of soil and sediment samples were set in the center of 40-mm diameter backed pellets encircled with ca. 10 g of powder boric acid and then pressed by a dynamoelectric HERZOG hydraulic press. Samples were then measured for oxides and minor elements by XRF with a Noton XL3T 950 He GOLDD+ XRF Analyzer mounted in a laboratory test stand with He gas purging to permit measurement of light elements.

Analyses were validated using the application of an instrument monitor of certificated reference material (GSS+GSD-09). Three sets of triplicates were produced using three randomly selected samples, these pellets were used to analyze the accuracy and precision of the spectrometry (Axios Max, PANalytical, Almelo, The Netherlands) instrument to gain an understanding of the accuracy and precision of the data set produced. The measurements of elements included 7 oxides and 25 minor elements (Na₂O, MgO, Al₂O₃, SiO₂, K₂O, CaO, Fe₂O₃, P, S, Ti, V, Cr, Mn, Co, Ni, Cu, Zn, Ga, As, Br, Rb, Sr, Y, Zr, Nb, Mo, Ba, La, Ce, Hf, Pb, and Th). The element analysis was performed at the School of Geographical Sciences, Nanjing Normal University. The analytical errors of <5% at the 95% confidence level were conducted for all the measurements.

2.4. Sediment Dating

Basically, the total ²¹⁰Pb activity was composed of supported ²¹⁰Pb (²¹⁴Pb) activity and excess or unsupported ²¹⁰Pb activity, and the latter provided the basis for establishing the sediment chronology [62]. When using the ²¹⁰Pb dating methods, a very important assumption is that the equilibrium between supported ²¹⁰Pb activity and ²²⁶Ra activity should be taken into consideration. However, since the release of ²²²Rn (a portion of ²³⁸U decay sequence) to the atmosphere also provides the sources of fallout ²¹⁰Pb (²¹⁴Pb_ex) in sediments, this assumption may be invalid [63,64]. The values of total ²¹⁰Pb and ²²⁶Ra obtained for the individual 2 cm depth increments of the sediment core are exhibited in Figure 3a,b, and the equilibrium between the supported ²¹⁰Pb and ²²⁶Ra would be expected. However, the negative values of ²¹⁰Pb_ex below 10 cm (Figure 3c) indicated that the values of supported ²¹⁰Pb may be overestimated by assuming equilibrium with ²²⁶Ra, which is similar to the previous study of [64]. According to the explanation for radioactive equilibrium between ²²⁶Ra and its daughter ²²²Rn [64], the supported ²¹⁰Pb should be corrected by the emanation coefficient (EC). Referring to the detailed description of EC in the previous study [64], the EC of 0.30 was selected to avoid the presence of negative values of ²¹⁰Pb_ex (Figure 3d). Hence, the constant rate of supply (CRS) model of ²¹⁰Pb_ex [62] has been used widely to calculate the sediment chronology and provides the best dating results [65]. The calculation formulas can be exhibited as follows:

$$t=\frac{1}{\lambda }\ln \left(\frac{A_h}{A_0}\right)$$

(1)

where t is the age in years, λ is the decay constant for ²¹⁰Pb (0.03114 a⁻¹), A₀ (Bq m⁻²) is the inventory of ²¹⁰Pb_ex in the sediment core, and A_h (Bq m⁻²) is inventory of ²¹⁰Pb_ex in each depth increment of the sediment. Based on the sedimentary chronology, the sediment accumulation rates (SARs) can be descripted as:

$$SARs=\frac{\partial Md}{\partial t}$$

(2)

where Md is the cumulative mass depth, it is equal to the sediment mass per unit area above a certain depth, and t is the interval time between a certain layer and the initial layer.

2.5. Sediment Source Apportionment

To successfully apply the sediment fingerprinting technique, three constraints must be satisfied [17]. First, a mass-conservative test was used to ensure all of sediment sample concentrations (mean, minimum, and maximum) of tracers fell within the range of source soil concentrations within a watershed [16]. On the contrary, those tracer properties in sediment falling outside the range of source values will be removed [15]. Second, the nonparametric KW was used to identify the ability of each fingerprinting property based on the null hypothesis that tracer properties exhibit no significant differences between source categories [15,66]. Third, a stepwise DFA based on minimization of Wilks’ lambda was applied to the fingerprinting properties that passed the KW to further identify the optimum set of fingerprinting properties that provide discriminatory power between sediment and sources [16,67].

A multivariate mixing model was used to estimate relative contributions from AL, CB, CL, and WL in land use types and form CS, DD, PB, QG, and SPS in geological types to sediment in this study [68]. The relative contributions from each source to sediment were determined by minimizing the following objective function:

$$f={{\displaystyle \sum _{i=1}^n\left\{\left[C_{ssi}-\left({\displaystyle \sum _{s=1}^m{P}_s{S}_{si}}\right)\right]/C_{ssi}\right\}}}^2$$

(3)

where n = the number of fingerprinting properties; m = the number sediment sources; C_ssi = the concentration of fingerprinting property (i) in the sediment samples; P_s = the relative percentage contribution from the source category; S_si = the median concentration of fingerprinting property (i) in the source category.

Two linear boundary conditions were satisfied by using the multivariate mixing model iterations to ensure that the relative source contributions from each source were non-negative (Equation (4)) and that these contributions from the different sources of sediment sum to unity (Equation (5)).

$$0\le P_s\le 1$$

(4)

$${\displaystyle \sum _{s=1}^m{P}_s}=1$$

(5)

The goodness-of-fit (GOF) provided by mixing model was evaluated by comparing the actual fingerprint property concentrations measured of the samples with the values predicted by the optimized mixing model. The GOF (%) can be calculated by the following modified objective function [69], viz.:

$$GOF=1-\left\{\frac{1}{n}{\displaystyle \sum _{i=1}^n\left|C_{ssi}-\left({\displaystyle \sum _{s=1}^m(S_{si}{P}_s)}\right)\right|/C_{ssi}}\right\}$$

(6)

A Monte Carlo approach was used to assess the uncertainties associated with the optimized sediment source contributions predicted by the mixing model. Cumulative normal distributions were constructed based on the mean and standard deviations of the measurements for each fingerprint property at each site using a random number generator. The procedure involved 1000 iterations. The confidence limits (95%) for the relative contribution of each soil sample to each sample were the results of the repeat iterations.

3. Results

3.1. Sediment Chronology and Accumulation Rates

Vertical distributions of ²¹⁰Pb_total, ²²⁶Ra, ²¹⁰Pb_ex, and ¹³⁷Cs in the sediment core are shown in Figure 3a–e. After correcting for ²¹⁰Pb_ex by EC of 0.30, the ²¹⁰Pb_ex activity in lakeshore sediment showed a fluctuated trend with depths (Figure 3d). The wave-like distribution of the ²¹⁰Pb_ex activity in the sediment core suggests variations in accumulation rates were impacted by the terrestrial material supply [70]. Unlike ²¹⁰Pb_ex, the appearance of detectable activity of ¹³⁷Cs at a depth of 20 cm indicates the year 1954 (Figure 3e), the date at which ¹³⁷Cs fallout was first widely deposited from the atmospheric nuclear testing [71]. No unambiguous peak of ¹³⁷Cs for 1963 was observed in the sediment core because of the long-term deposition of catchment-derived ¹³⁷Cs, which migrated from the up-stream catchment [72].

To overcome the uncertainty of the single ²¹⁰Pb_ex dating result, the results of ²¹⁰Pb_ex Correction-Constant Rate of Supply (C-CRS) dating approach were adjusted based on the individual ¹³⁷Cs time marker in the sediment core [56]. The C-CRS model gave results of ~160 years of sediment deposition, from the year 1858 to 2020. Based on the estimated chronology of ²¹⁰Pb_ex and ¹³⁷Cs (Figure 4a), the time-dependent sediment accumulation rates (SARs) of the core were described in Figure 4b. The SARs in the lakeshore sediment have a range of 0.002 to 0.065 g cm⁻² a⁻¹, with an average value of 0.015 ± 0.016 g cm⁻² a⁻¹. An extremely high value region appeared in the early-1990s.

3.2. Particle Size in Catchment Soils and Lakeshore Sediment

The percentages of each particle size (clay, silt, and sand) in catchment soils and lakeshore sediment are shown in Figure 5. For catchment soils, the fine fractions with particle size < 63 μm (clay + silt) were predominant, accounting for 90.3%. By contrary, there is a high percentage of sand (>63 μm) in lakeshore sediment, with a value of 46.6%. The fractions of particle size indicated that the driving mechanism of soil particle movement in catchment soils is different from lakeshore sediment.

3.3. Selection of Optimum Fingerprinting Properties

In this study, a total of 33 potential fingerprint properties were used to identify the sources under four different land use types (AL, CL, WL, and CB) and geological types (CS, DD, PB, QG, and SPS) contributing to the lakeshore sediment. Of these fingerprint properties, 23 fingerprint properties passed the first constraint that all of sediment sample concentrations (mean, minimum, and maximum) of tracers fell within the range of land use and geological sources soil concentrations, and ten fingerprint properties (i.e., Na₂O, Al₂O₃, P, S, V, Ni, Ga, Br, Zr, Hf) did not pass the constraint (

Table 1. Geochemical soil source tracers which passed (P) each constraint for input into the fingerprint optimization procedure.

Property	Source Samples			Sediment Samples			Sediment Mean Inside a	Sediment Samples Inside b	H-Value of Land Use (Geology)	p-Value of Land Use (Geology)
	Mean	Mix	Max	Mean	Mix	Max
Na2O	0.11	0.05	0.34	0.52	0.29	0.65	P
MgO	2.40	0.84	7.46	3.53	2.50	4.23	P	P	8.641 (13.625)	0.029 * (0.004 *)
Al2O3	19.48	11.18	27.66	10.99	9.37	14.47
SiO2	44.26	33.15	64.54	53.67	40.19	57.96	P	P	7.121 (10.237)	0.064 (0.030 *)
K2O	2.36	1.10	4.19	2.24	1.62	2.43	P	P	0.802 (5.983)	0.857 (0.204)
CaO	4.17	0.31	15.63	8.51	6.87	14.76	P	P	7.618 (10.715)	0.050 * (0.023 *)
Fe2O3	11.93	4.49	19.02	6.27	4.94	9.86	P	P	4.753 (8.765)	0.192 (0.064)
P	2274.19	765.00	6801.90	1653.22	677.80	5087.90	P
S	404.24	135.60	648.00	296.56	164.30	673.20	P
Ti	13,800.17	4138.90	28,260.20	9373.46	7583.50	14,985.50	P	P	1.412 (6.794)	0.721 (0.147)
V	273.46	116.20	451.30	149.45	111.00	272.30	P
Cr	145.78	72.50	263.50	122.84	85.20	195.10	P	P	4.436 (3.294)	0.220 (0.527)
Mn	978.95	289.20	2192.90	1084.03	746.80	1782.50	P	P	2.052 (1.353)	0.578 (0.865)
Co	30.13	6.60	65.60	19.95	9.10	49.40	P	P	2.130 (2.709)	0.561(0.625)
Ni	74.17	30.90	117.50	31.45	20.90	56.90	P
Cu	105.46	25.70	233.60	54.88	37.50	100.00	P	P	1.026 (10.796)	0.805 (0.022 *)
Zn	173.87	61.60	307.50	196.64	158.30	233.30	P	P	5.867 (10.300)	0.116 (0.029 *)
Ga	24.76	10.50	37.60	12.87	9.50	18.00	P
As	22.34	7.20	42.60	12.51	8.40	20.60	P	P	10.381 (7.568)	0.012 * (0.108)
Br	3.23	0.80	7.40	1.55	0.70	2.70	P
Rb	117.75	65.50	190.70	83.58	66.40	90.90	P	P	1.992 (14.909)	0.591 (0.002 *)
Sr	65.83	35.90	151.60	70.89	60.70	104.20	P	P	18.558 (24.576)	0.000 * (0.000 *)
Y	47.01	20.00	83.60	32.86	27.00	41.60	P	P	0.766 (1.225)	0.864 (0.886)
Zr	329.85	112.10	500.30	524.21	336.00	669.30	P
Nb	30.21	7.10	53.40	19.89	16.90	23.70	P	P	3.217 (9.599)	0.370 (0.041 *)
Mo	2.20	0.10	5.20	1.49	0.40	2.40	P	P	1.855 (8.922)	0.620 (0.059)
Ba	260.12	77.10	492.10	387.78	257.30	455.60	P	P	14.412 (15.165)	0.001 * (0.001 *)
La	57.67	34.10	86.30	50.91	44.80	66.30	P	P	1.726 (2.156)	0.651 (0.727)
Ce	119.45	56.40	160.70	90.93	77.40	121.60	P	P	4.206 (4.716)	0.242 (0.325)
Hf	8.40	3.20	12.20	13.29	8.10	17.80	P
Pb	61.01	24.00	234.60	65.48	55.50	74.90	P	P	4.421 (18.701)	0.222 (0.000 *)
Th	13.63	5.30	24.80	8.42	5.60	10.60	P	P	13.532 (21.206)	0.002 * (0.000 *)
CO3	13.33	8.26	18.43	12.85	11.08	15.74	P	P	1.036 (6.926)	0.804 (0.138)

^a Mean sediment concentration within the range of source category mean values. ^b All sediment sample concentrations were within the range of source sample values; in this example, constrains 1 and 2 were applied using lakeshore sediment. * Significant at p ≤ 0.05, means the element is not detected in sources but detected in sediments, units of weight in % (Na₂O, MgO, Al₂O₃, SiO₂, K₂O, CaO, Fe₂O₃), and the other are parts per million (ppm).

). A total of 6 fingerprint properties (i.e., MgO, CaO, As, Sr, Ba, Th) in land use sources and 11 fingerprint properties (i.e., MgO, SiO₂, CaO, Cu, Zn, Rb, Sr, Nb, Ba, Pb, Th) passed the second constraint of KW (p ≤ 0.05) with the H-values ranging from 7.618 to 18.558 in land use and from 9.599 to 24.576 in geology, respectively (Table 1). The third constraint of a stepwise DFA was subsequently undergone to create optimum composite fingerprint properties (As and Sr) in land use types, and (CaO, Rb, Sr, Th) in geological types for discriminating the potential sources (

Table 2. The optimum composite fingerprints for discriminating individual sediment source types in the West-East Luju River catchment.

Sources Types	Step	Fingerprint Property Selected	Wilks’ Lambda	Cumulative Source Type Samples Classified Correctly (%)
Land use	1	As	0.420	83.7
	2	Sr	0.749	100.0
Geology	1	CaO	0.126	78.4
	2	Rb	0.506	93.7
	3	Sr	0.804	99.9
	4	Th	0.996	100.0

). Regardless of land use and geological types, they correctly distinguished 100% of the source samples that were used to characterize each of the four land use source types (AL, CL, WL, and CB) and five geological source types (CS, DD, PB, QG, and SPS) (Table 2).

3.4. Spatiotemporal Sediment Source Apportionment

As shown in Figure 6, the spatial source apportionment of lakeshore sediment between land use and geology were calculated (Equation (3)) by comparing the potential sources with surface (0–2 cm) sediments. Spatially, unmixing of the lakeshore sediment against geology reveals that QG was the dominant sediment contributor source within the catchment with a proportion of 99.9%, followed by CS (5.5%), respectively (Figure 6a). When evaluated against land use source signatures (Figure 6b), CB was shown to be the dominant contributor (70.6%), followed by AL (29.4%), and CL (8.3%), respectively. Basically, DD, PB, and SPS for geology, and WL for land use, provided much smaller contributions to the lakeshore sediment. Evaluation of the temporal sediment sources was implemented by these unmixing sections of the sediment core, and CB was recognized as a dominant source in the whole sediment core, with an average contribution of 88.6%. Combined with those zones of changing of SARs, four different stages were divided in this study (Figure 7).

3.5. Results of Uncertainty in Sediment Fingerprinting

According to Equation (6), the GOF values calculated for the mixing model runs were associated with apportioning the sources of the lakeshore sediment samples collected from land use types and geological types. The estimates of the mean GOF were 0.62 in land use and 0.64 in geology, respectively. Both of mean GOF values did not exceed 80%, the simulation between the mixing model and the Monte Carlo were similar, indicating the acceptable estimation of the uncertainty of analysis using the inmixing model.

4. Discussion

4.1. Responding of Natural and Anthropogenic Factors to Changing of SARs

Changes in natural factors and anthropogenic activity are important drivers influencing the SARs in Lakes [73,74]. Generally, as the important natural factor, precipitation is often regarded as the dominant driving force for accelerating catchment erosion and sediment deposition [1,75]. The positive correlation between SARs and instrumental Indian Summer Monsoon (ISM) precipitation [76] (y = 83.321 ln(x) + 1560.5, R² = 0.5312, p < 0.05) (Figure 8) also supported our conclusion. However, only a few particle sizes fall within the range of the lakeshore sediment (Figure 5), indicating that precipitation was not the sole factor for influencing the changes in SARs. Alternatively, anthropogenic activities, such as variations in land use types should also be important factors for producing diverse contributions to the down-stream deposited sediments [11,77]. Compared with the changes in time-integrated source appointment of AL, CB, CL, and WL (Figure 7), particle sizes (d₅₀, clay, silt, and sand), precipitation in Chengjiang (Yuxi meteorological station) from 1951 to 2019, and instrumental ISM precipitation over the past ~160 years, four zones were partitioned for interpreting the dual impacts of natural factor and anthropogenic activity on changing of SARs (Figure 9).

Zone A (ca. 1860s to 1930s). Relatively lower SARs were observed in lakeshore sediment during this time frame, which is similar to our previous studies on the changes in SARs in lakeshore sediments [56]. Before the 1910s, Lake Fuxian was in a state of nature, with relatively little human disturbance existing within the catchment [56]. This result was also verified by the relatively low fractions of d₅₀ and sand in this study (Figure 9). Since then, a slowly upward trend in SARs was found in the study, which may be due to land reclamation in the early 20th century. Excessive reclamation of land accelerated soil erosion, thereby resulting in high SARs. In addition, the “Yongtian regime” was the important permanent tenancy system in which the landlord had the ownership of the land for collecting rent but farmer only had the use-right of the land for paying rent [78]. The private ownership of land made the farmers select land with good soil fertility for cultivating, whereas the land with poor quality was abandoned. And thus, the abandonment of farmland resulted in a lack of land management, and thereby causing soil erosion contributing to the sediments.

Zone B (1930s–1950s). There is a small peak for SARs in 1950s, which corresps to the high value of d₅₀ and sand at the same time (Figure 9). On one hand, it originates mainly from the impact of higher precipitation from ISM. On the other hand, Lake Fuxian and its catchment began to suffer from anthropogenic impacts from the early 1930s [56]. Large-scale exploitation of natural resources has led to increasing rates of weathering, erosion, transportation, and accumulation [79]. Furthermore, China began to collectivize its agriculture from around the early 1950s, which led to the ownership of landlords being transferred to collectivization [80]. The pattern of mutual-aid teams cannot improve the enthusiasm of agricultural activity due to the unresolved land ownership.

Zone C (1950s–1990s). An abnormal high value of SARs was found in the early-1990s, but there is a relatively lower value of d₅₀ in the same period (Figure 9). The result implied that the impact of anthropogenic activity on SARs reached a high level during this period. Meanwhile, similar research found that the unresolved land ownership resulted in the abandonment of the traditional agricultural landscape in Slovakia under the collectivization of agriculture during the period of the 1950s–1980s [81]. The not-active attitude for cultivating led to the farmland to be abandoned. Combined with the sediment records of biological and geochemical proxies, the latest studies also found that anthropogenic activity was the key factor in influencing the changes in the recent lake environment of Lake Fuxian [82,83].

Additionally, the changes in land use from 1974 to 2015 in Fuxian Lake catchment were interpreted by the previous research stating that there is an increasing trend in AL and construction land (Figure 10a,c), but a decrease in CL and WL (Figure 10b,d) [55]. In China, the greatest reform for land was the birth of the household-contract responsibility system in 1978, after which land has been contracted to every household up to now [80]. The policy greatly improved the enthusiasm of the farmer for crop cultivation, but the agricultural tax was still needed to pay for the local government. In most cases, poor agricultural productivity with the increased slope and elevation of farmland resulted in agricultural incomes slightly higher or lower than the agricultural tax. Consequently, many farmers returned their farmlands to the government to avoid having to pay the agricultural tax, which led to the farmland of the hilly and mountainous regions to be abandoned.

Zone D (1990s–2020s). There is a relatively lower value of SARs, which is in consistent with our previous study [56]. In order to reduce water and soil erosion into the lake and improve the level of the environment protection in the whole lake-catchment system, governmental land reform projects, such as “Grain-for-Green”, was implemented after the late-1990s [84]. The huge project of “Grain-for-Green” led to the forest coverage in Fuxian Lake catchment rising from 33.2 to 40.0% by 2017 (Yunnan Provincial Bureau of Statistics), which prevented soil loss within the catchment being eroded and transported into the lake [85]. The particle sizes in lakeshore sediment are dominated by fine and silt fractions after the 2000s (Figure 9), indicating weak soil erosion occurring in the catchment. Moreover, the closure of Gehe River that connected Fuxian and Xingyun Lakes (Figure 1a) in 2000s resulted in the catchment scales being transformed from the Fuxian and Xingyun Lake catchment to the giant Fuxian lake, thereby resulting in higher values of SARs and nutrient inputs before 2000 [86].

According to the latest research [87], the government has abolished the agricultural tax and subsidized money for the framers since 2006 to improve farming enthusiasm for farmers. However, due to the impacts of the market-oriented economy and rapid urbanization in China, young and middle-aged farmers wish to select gainful employment in urban areas to obtain a higher income, better life, and high-quality education for children, instead of farming in countryside [39,88]. The number in the agricultural labor force has been shrinking rapidly from 363 million in 2003 to 241 million in 2013 [39], and a declining trend in the area was still identified to change from 20.23 million during the period of 2005–2010 to 11.42 million during the period of 2020–2030 [89]. The decline in agricultural labor forces has been recognized as the key driver resulting in cropland to be abandoned, particularly in the hilly and mountainous regions of China [39,55]. According to the model prediction of farmland abandonment from the latest study [39], nearly two-thirds of croplands would be abandoned by 2030, which could significantly exacerbate the future challenges and produce a potential threat to the maintenance of China’s food security and sustainability.

4.2. Interpreting of Spatiotemporal Source Apportionment in Geology and Land Use

For spatial and temporal sources of land use, the primary lakeshore sediment contributor was CB, indicating fluvial erosion was the dominant driver for delivering the material to the lake. Numerous previous fingerprinting investigations conducted in UK, Australia, and US indicated that sediments from channel sources, such as removal of sediment through processes of fluvial erosion, dominate for transporting sediment within riverine systems (>50%) [15,16,18]. For geological sources, QG was the dominant contributor to the lakeshore sediment. According to the map of geology (Figure 1d), QG was distributed in the middle and lower reaches of the West-East Luju River, which facilitated the land use and geological source to be formed by fluvial erosion.

Interestingly, two relatively higher contributions of AL to lakeshore sediment were found during the periods of 1930s–1950s and post-1990s, which are in consistent with the uptrend trend in AL and construction land after 1990s (Figure 10a,c). Also, the onset time of human disturbance on the lake and its catchment agrees with our previous research on nutrient accumulation and burial in lacustrine sediments [56,86]. Other geochemical proxies, such as n-alkane homologues, carbon isotopes (¹³C_org), inorganic/organic carbon, and black carbon in lacustrine sediment also magnified that human activity is of importance in influencing the environmental changes in lakes [85,86].

Regardless of spatial and temporal scales, CB was the dominant source contributing to sediment, but the inputs of anthropogenic sources were important during the two periods. For instance, the proportion of 29.4% in AL at the surface indicated farmland abandonment derived from human activity played a vital role in providing the potential sources to lakeshore sediment.

The relationships between particle size and the optimum fingerprint properties (As, Sr, CaO, Rb, Th) in soils and sediments are shown in

Table 3. Relationships between particle size and elements in soils and sediments.

	<4 μm	4–63 μm	>63 μm	As	Sr	CaO	Rb	Th
<4 μm	1	0.036	−0.723 **	0.663 **	−0.360 **	−0.320 *	0.294 *	0.648 **
4–63 μm		1	−0.716 **	0.195	0.092	−0.271 *	0.495 **	0.126
>63 μm			1	−0.597 **	0.188	0.410 **	−0.547 **	−0.540 **
As				1	0.093	−0.431 **	0.377 **	0.583 **
Sr					1	0.143	−0.299 *	−0.456 **
CaO						1	−0.479 **	−0.548 **
Rb							1	0.736 **
Th								1

* Correlation is significant at the 0.05 level (2-tailed). ** Correlation is significant at the 0.01 level (2-tailed).

. Most fingerprint properties showed a positive correlation with particle size (<63 μm), which indicated tracer elements adsorbed easily with the fine fractions of particle size < 63 μm (clay + silt) [15]. The predominant fraction of particle size of <63 μm (clay + silt) of 90.3% in soils and 53.4% in lakeshore sediments (Figure 5) implied that the properties of tracer elements were not influenced by stressors and anoxic conditions in the erosional and/or depositional process.

Variations in land use types, as the major pattern of human activity, was impacted by the combined effects of topography and socioeconomic factors [90,91]. Topographical factors, such as slope gradient, aspect, and elevation had important influences on the spatial distribution patterns of land use types in hilly and mountainous areas [92]. As shown in Figure 11, the average values of slope gradient and elevation at the sampling sites are in order: WL > AL > CL > CB. As mentioned above, CB provided 88.6% of source contributions to lakeshore sediment over the past ~160 years, although it has a smaller slope gradient and lower elevation in this study (Figure 11). One probable explanation is that the collected samples of CB were closed to the West-East Luju River, where fluvial erosion was the dominant driving force for providing the potential source [93]. For CL, the gentle slope gradient and low altitude are beneficial to the agricultural mechanization and large-scale production, and thus, the local farmers are willing to select the flat cropland for farming [94]. With the increasing slope gradient and elevation (Figure 11), land use types were transformed from CL to AL by local farmers because sloping farmland required more agricultural labor and capital inputs in comparison with flat farmland [41].

In addition, increasing slope gradient accelerated the soil erosion of AL in comparison to CL, which resulted in sloping farmland to be abandoned by local farmers [39]. To further understand the impact of topography on farmland abandonment, the diagram of wind rose in slope aspect was used to describe the distributions of slope aspects on the four land use types in this study (Figure 12). Although the aspects of SSE and SE provided stronger solar radiation for producing primary productivity [95], the farmland was still abandoned. On the other hand, the subjectivity of farmland abandonment for people was not influenced by slope aspect.

Besides this, the gap between farmer’s expenditure and incomes in CL are playing an important role in influencing farmland abandonment. For example, a household survey of local cultivating input fees on farmers with a unit of yuan per mu (y mu⁻¹, 1 mu = 666.7 m²), including seeds (80 y mu⁻¹), fertilizers (150 y mu⁻¹), membranes (30 y mu⁻¹), herbicides (20 y mu⁻¹), and ploughing (100 y mu⁻¹) indicated that it is a low income (620 y mu-1) subtracted by the total income of 1000 y mu⁻¹ in a good harvest year. Hence, many farmers wish to seek manual labor in cities rather than cultivating the farmland in countryside, which inevitably leads to the farmland abandonment [91].

4.3. Uncertainty Analysis

The uncertainty of analysis may be due to the impacts of source numbers and particle size and organic matter correction. First, the access to private farms and channel bank to take source samples constrained the number of samples that could be collected. The relatively small number of source samples limited our ability to perform more rigorous uncertainty of analysis [16]. Second, there is a lack of correction of particle size (specific surface area) and organic matter for the multivariate unmixing model (Equation (3)), which reduced the accuracy and precise of analysis results [15].

5. Conclusions

The fallout radionuclides of ¹³⁷Cs and excess ²¹⁰Pb, particle size, oxides, and minor element of surface soil samples, and of lakeshore sediment in core, from the West-East Luju River catchment of Southwest China was determined and the spatiotemporal contributor sources of land use and geology to lakeshore sediment using a Bayesian mixing model were quantitatively evaluated. Spatially, QG in geology and CB in land use were the dominant contributors to lakeshore sediment, respectively. Referring to the changes in SARs, particle sizes, precipitation, and land use policies, four zones were deduced to interpret the evolution of soil erosion after farmland abandonment over the past ~160 years. For land use, topographical factors of slope gradient and elevation, and socioeconomic factors of the gap between farmer’s expenditure and income had important impacts on the farmland abandonment. On the contrary, the impact of slope aspect was not significant. The results implied that it is urgent to restrict the farmland to be abandoned for local communities in relation to sustainability and soil erosion. Meanwhile, the spatiotemporal source apportionment of lakeshore sediment provided useful information for further managing and designing land use policies applicable to small mountainous and hilly areas elsewhere in the world.