Agricultural intensification vs climate change: What drives long-term changes of sediment load?

Climate change and agricultural intensification are expected to increase soil erosion and sediment production from arable land in many regions. However, so far, most studies were based on short-term monitoring and/or modeling, making it difficult to assess their reliability in terms of estimating long-term changes. We 20 present the results from a unique data set consisting of measurements of sediment loads from a 60ha catchment (the Hydrological Open Air Laboratory, HOAL, in Petzenkirchen, Austria) over a time period spanning 72 years. Specifically, we compare Period I (1946-1954) and Period II (2002-2017) by fitting sediment rating curves for the growth and dormant seasons for each of the periods. The results suggest 25 a significant increase in sediment loads from Period I to Period II with an average of commonly took land use/cover change and landscape structure change as a whole when perofrming such attribution analysis, and most have been 删除[胡杨 [2]]: in 删除[胡杨 [2]]: window 删除[胡杨 [2]]:


Introduction
Soil erosion is a phenomenon of worldwide importance because of its environmental and economic consequences (García-Ruiz, 2010;Prosdocimi et al., 2016). Climate 55 change, land use/cover changes and other anthropogenic activities are commonly considered potential agents that drive variation of soil erosion rates (Nearing et al., 2004;Zhang et al., 2021). The impacts of climate change (e.g. Nearing et al., 2004;Zhang and Nearing., 2005;Mullan, 2013;Palazon and Navas, 2016) and of land use and cover change (LUCC) (e.g. Bochet et al., 2006;Korkanç et al., 2018;Nampak et 60 al., 2018;Li et al., 2019;Perović et al., 2018) on erosion have been studied in recent years. As the two agents usually exert their influence on soil erosion simultaneously, their relative contributions have also been increasingly investigated in recent years (e.g. Bellin et al., 2013;Sun et al., 2020;Zhang et al., 2021). Combining field investigations with model simulations, Zhang et al. (2021) quantitatively evaluated impact on erosion than climate change. Also, livestock grazing accelerated soil erosion was found to be more important than climate change in the Qinghai-Tibet Plateau .
The previous findings provide valuable information on understanding how LUCC and climate change affect soil erosion and sediment load. However, it seems that most of 75 the previous studies considered LUCC and landscape structure change together. The relevance of landscape structure changes alone has so far received less attention, even though land-use policies, such as land consolidation, have changed agricultural practices to a large extent since 1945, the beginning of agricultural industrialization (e.g. Moravcová et al., 2017;Devaty et al., 2019), and in particular in countries where 80 the industrialization of agriculture is relatively recent (Bouma et al., 1998;Moravcova et al., 2017;Zhang et al., 2021).
Landscape structures usually influence erosion due to the boundary effects between land uses and land units (parcels) that differ in water and sediment trapping capacity (Baudry and Merriam, 1988;Merriam, 1990;Takken et al., 1999;Phillips et al., 2011). 85 Van Oost et al. (2000) and Devaty et al. (2019) evaluated the role of landscape structure by accounting for its spatial connectivity using modelling approaches and found that landscape structure is an essential factor when assessing the risk of soil erosion affected by land use changes. Both studies emphasized the potential impacts of parcel structure changes on sediment production through altering hydrological and 90 sediment connectivity. However, both studies relied on models, making connectivity assumptions in their studies. Instead of focusing on the spatial connectivity, others (e.g. Bakker et al., 2008;Sharma et al., 2011;Chevigny et al., 2014;Wang et al., 2021;Tang et al., 2021;Madarász et al., 2021) evaluated the effect of terrain, soil properties, lithology, management practices and other processes associated with landscape and/or 95 land structure changes and highlighted their impact on sediment production. It has also been shown that the impact of landscape structure on erosion is more heterogeneous when different crops are grown, and the underlying lithology, soil properties and topography show substantial spatial variability across the catchment (David et al.,2014;Cantreul et al., 2020). 100 In our analysis, we evaluate the relative roles of climate change, LUCC and the change of land structure on sediment production, especially with a focus on understanding the respective role of LUCC and landscape structure change, based on long term observations that were not usually available in previous studies. We present the results from a unique data set consisting of measurements of sediment loads from 105 a small agricultural catchment over a time window of 72 years. The study catchment is the 66 ha Hydrological Open Air Laboratory (HOAL) Petzenkirchen (Blöschl et al., 2016), which, in addition to being exposed to climate change, has experienced a significant change in land use and land cover as well as parcel structure due to altered land management policies during the past decades. Both discharge and suspended periods of 1946-1954 and 2002-2017; ii) analyzing whether climate change or land 115 use changes (or both) were responsible for any change in the sediment regime; and iii) identifying the relevance of land structure change (i.e. land consolidation) on erosion control compared to that of a change in land use or cover.

120
The HOAL catchment is situated in Lower Austria's alpine forelands (48°9' N, 15°9' E) with elevations ranging from 268 m to 323 m a.s.l. and a size of 66 ha ( Figure 1). The climate of the catchment belongs to the temperate, continental climate zone (Dfb) according to Köppen-Geiger (Kottek et al., 2006), with a mean annual precipitation of 746 mm (1946 -2006), 62% of the rain falling between May and October. The mean 125 daily air temperature is 8. 8°C (1946-2006), and the dominant land use is arable land, accounting for, on average, 82% of the catchment over the past few years. Typical crops include winter wheat, corn and barley. Deciduous trees grow along the stream (6%), 10% of the area is grassland, and 2% is paved. The subsurface of the catchment consists of tertiary marine sediments. Soils are classified into five types: calcic measurements were stopped and started again in 1990. Therefore, data records for the period 1946-1954(Period I) and 2002 were used for this analysis. To our knowledge, for the time period 1945-1954, almost no sediment concentration data 140 are available in Austria, we therefore think that this databse from the HOAL is extremely valuable and relevant for climate impact analysis. In Period II, the stream gauge was relocated. However, the difference in catchment size is very small (around 200 m 2 ), and indicated by the different locations of the discharge gauge in Figure 1.
Due to technological advances, the measurement method of both Q and C changed 145 between the two periods. In Period I, discharge was registered at 10 min resolution by a Thompson weir and a paper chart recorder, while in Period II, it was registered at 5 min resolution by an H-Flume and a pressure transducer. Sediment concentrations were measured manually every 3-4 days in Period I, whilst an automatic method (i.e.equal-discharge-increment sampling) plus additional manual sampling was 150 applied in Period II. Daily precipitation and 5-min rainfall intensities were available for both periods, but for Period I, 5-min rainfall intensities were only available during the growing season.
We used parcel-based land use data from 1946 to 1949 and 2007 to 2012, representing Period I and Period II land use, respectively. Land use categories were agricultural 155 land, including crop type, grassland, forest, roads and settlements (i.e. paved area).
We defined a parcel as a continuous area of land with a single crop type. Parcel boundaries were specified according to the cadastral map and aerial photographs. In Period II, these boundaries were also visually inspected. Figure 1 depicts the geographic catchment location, and parcel structure and land use for a specific year in 160 each period.

Changes in rainfall erosivity and flow regime
The erosive potential of rainfall events was quantified by the R-factor of the Revised Universal Soil Loss Equation (RUSLE), which is defined as the product of kinetic energy of a rainfall event and its maximum 30-min intensity, using the rainfall erosivity tool RIST (USDA-Staff, 2019) according to where EI30 is the Annual R-factor (MJ*mm/ha*hr) calculated as the sum of single event R-factors, Ei is the total kinetic energy for a single event (kJ . m -2 ), I30 is the maximum rainfall intensity in 30 minutes within a single eventi (mm . h -1 ), and m is the number of events per year. 180 We assumed erosivity density ED (i.e. EI30 divided by event precipitation) to be a particularly relevant climatic indicator of soil erosion process and catchment sediment yield, because it is calculated as a combination of rainfall kinetic energy and maximum rainfall intensity of rain events. These are commonly considered as the relevant parameters of rain to trigger soil erosion. We, thus, tested whether the means 185 of the monthly erosivity density (ED) are significantly different between Period I and Period II by using a t-test. Due to the absence of rainfall intensity measurements, we could not directly calculate ED for the months of the dormant season (November to March) of Period I. Instead, we calculated ED from a relationship between EI30 and monthly rainfall of Period II, assuming that the relationship was sufficiently 190 temporally invariant over the investigated periods. Erosivity density is very low during the dormant season.The mean ED was 0.66±0.21 and 2.54±2.43 MJ ha -1 hr -1 respectively for the dormant season and the growing season of Period I, whilst it was 0.42±0.11 and 1.87±1.35 MJ ha -1 hr -1 respectively in Period II ( Figure 3a). Thus, the error arising from the use of this relationship is expected to be small. 195 We also compared daily flow duration curves to understand whether hydrological regime change has influenced flow transporting capacity and sediment regime change.
Following the definitions of Smakhtin (2001), we compared low flow (Q70%), high flow (Q10%) and median flow rate (Q50%) quantiles for the two periods. 200 We first estimated sediment load for the different time periods. Daily mean streamflow or daily mean sediment concentration were estimated as arithmetic average of multiple observations within a day, and then we estimated sediment load for each month according to equation (2):

Sediment regime analysis
where Y is the sediment load for a month (kg . mon -1 ), i Q and i C are mean daily discharge (l . s -1 ) and mean daily sediment concentration (mg . l -1 ), respectively, for a specific day i having data records; 1 -i Q and 1 -i C are mean daily discharge (l . s -1 ) and mean daily sediment concentration (mg . l -1 ), respectively, for the previous day with available data measurements as well ;Ti (s) is the elapsed time between day i and the 210 last previous recording day. The latter value depends on how often sediment concentration was recorded within a month. The statistical differences of sediment loads between seasons and between periods were examined by a t-test. Sediment regimes are usually analyzed using sediment rating curves (SRC).
Following a common approach (Asselman, 2000;Warrick and Rubin, 2007;Sheridan 215 et al., 2011;Vaughan et al., 2017;Khaledian et al., 2017), the SRCs were here assumed to follow a power-law function, which was fitted by least squares regression: where C is sediment concentration (mg . l -1 ), Q is discharge (l . s -1 ), and a and b are regression coefficients. The coefficient a is usually associated with easily transported 220 intensively weathered materials and may vary over seven orders of magnitude (Syvitski et al., 2000). The parameter b represents the capacity of the stream to erode and transport sediment, reflecting how sediment concentration is non-linearly related to streamflow (Sheridan et al., 2011;Fan et al., 2012). It typically varies from 0.5 to 1.5 and rarely exceeds 2. Sometimes b is also regarded as a measure of the quantity of 225 new sediment sources available (Vanmaercke et al., 2010;Guzman et al., 2013).
Considering that data records were registered with different temporal resolutions for Periods I and II (See section 2.2), for the sake of consistency, we used monthly averages, as in other studies (Syvitski and Alcott,1995;Sheridan et al., 2011;Hu et al., 2011), to construct SRC. We assumed that monthly averages could reflect a varied 230 hydrological and/or sediment response to seasonally prevailing weather characteristics such as dry periods or convective storms (Sheridan et al., 2011).
We chose arithmetic means of the observations to represent the monthly Q and C values. These monthly averages were pooled together and then grouped into growing season of Period I (Period I_G), dormant season of Period I (Period I_D), growing 235 season of Period II (Period II_G), and dormant season of Period II (Period II_D), respectively. For each of these four periods, we fitted SRC.
We analyzed the fitted SRC by two strategies to evaluate whether and how the sediment regime changed between these periods. Besides directly comparing the slopes of the four seasonal SRCs by ANCOVA analysis, we also fitted the SRC for 240 each specific year's season and plotted the regression coefficients a against their corresponding b to evaluate a possible sediment regime shift during Periods I and II.
The latter framework was adapted from Thomas (1988), and also employed by Asselman et al. (2000) and Fan et al. (2012) to examine differences in sediment regimes between spatially different sites. Also Sheridan et al. (2011) used the 245 framework to reveal post-fire temporal shifts of a sediment regime. Thomas (1988) suggested that time-based sampling methods (either random sampling or systematic sampling) preferentially use observations of relatively small discharges to fit a SRC, tending to reduce the slope and increase the intercept of a SRC (see the point C in Figure 2b); In contrast, flow-based automatic sampling methods such as 250 equal-discharge-increment sampling preferentially use observations of relatively large discharges. Thus, they tend to cause a reversed pattern of a and b (i.e. increase the slope and decrease the intercept of SRC, see the point A in Figure 2b). However, irrespective of sampling practices, the pairs of data points a against b will likely be allocated along a straight line, if sediment transport regimes are similar. The reason 255 for the a-b pairs lying nearly on a straight line is mainly due to a mathematical property, that is, the slopes could be expressed by a linear function of the intercepts with the coordinates of the common point (Thomas, 1988). Therefore, for years with similar means of log-Q and log-C, the SRCs will pass through one common point O (Thomas, 1988;Syvitski et al.,2000;Desilets et al., 2007;Sheridan et al., 2011). This  (Asselman, 2000). The coefficients a are usually inversely linearly related to b 265 (Thomas, 1988, Syvitski et al., 2000and Desilets et al., 2007, and each point is representative of a period (or a catchment). If sediment transport regimes are similar between periods (catchments), the points will be plotted on the same line (such as A, B, C in Figure 2b), with points A of Figure 2b (upper-left-side) often exhibiting steeper sediment rating curves than points C (lower-right side). As for different lines 270 in Figure 2b, the lower ones characterized by points A', B', and C' represent situations with most of the annual sediment load being transported at relatively low flow discharges, whilst the higher ones characterized by A, B, and C represent situations with suspended sediment being mainly transported at high streamflow. Compared to a direct evaluation of rating curves, relating coefficient a to exponent b is more 275 conductive to revealing temporal evolution of sediment regime (Syvitski et al., 2000;Desilets et al., 2007). The change in the relationship of coefficients a against b between the periods was also examined by ANCOVA .

290
To account for uncertainties of the fitted SRC during each period, we additionally established theoretical sediment rating curves (tSRC) i) For each period (i.e. Period I_G, Period I_D, Period II_G, and Period II_D), we carried out random sampling of log a (n=500, package "sample" in RStudio), 295 assuming that the samples of the coefficient of log a follow normal distributions, which was examined with a Kolmogorov-Smirnov test of normality (mean = 1.02, SD = 2.01, n=44, p<0.05 ); ii) Given the set of the sampled 500 values of log a, we generated a set of values b according to the previously established linear relationship between log a and b; 300 iii) Given a set of specified Q values, we derived 500 tSRC for each period, corresponding to the paired log a and b samples; iv) Using these tSRC we calculated the 50 percentile, 5 percentile, and 95 percentile for each period to to estimate the uncertainties of the sediment rating curves.
The tSRC of the periods were also used to quantify the effect of land consolidation, i.e. 305 the change of parcels structure and sizes (Parcel_effect) and the effect of land use and land cover changes (LUCC_effect). Since vegetation usually plays a minor role in the dormant season due to the absence of a dense vegetation cover on arable land and little erosive rainfall (Madsen et al., 2014;Kundzewicz, 2012;Salesa and Cerda, 2020;Hou et al., 2020), landscape structure in the dormant season was considered a 310 dominant factor for water and sediment transfer across the land surface, and thus runoff production and sediment production (Sharma et al., 2011;Devátý et al., 2019).
Therefore, we hypothesized that the total change in sediment yield (Total_effect) resulted from land cover change (LUCC_effect), land structure change (Parcel_effect) and climate change (Climate_effect). Since the area of our catchment is only 0.67 km 2 , 315 no obvious change was found in the shape of the small stream for the two periods.
Stream sediment resuspension is rather small (Eder et al., 2014), the contribution of bank erosion was not taken into account. The effects of land cover and land structure change was quantitatively separated according to the seasonal differences in tSRC after determining the climate change effect. Specifically, we assumed that the shift of 320 sediment regime from Period I_D to Period II_G was representative of the the Total_effect were estimated according to Equations (7) and (8), respectively.

335
Because climate change is often found responsible for hydrological change (e.g., Kelly et al., 2016;, we compared erosivity density (ED) and monthly precipitation (P) of the two periods to examine whether climate affected the variation of the sediment regime in the catchment (Figure 3).  Streamflow in Period I was higher than that of Period II, and the mean annual streamflow was 188 and 146 mm.yr -1 for Periods I and II, respectively. Daily flow 365 duration curves for both periods are displayed in Figure 4. An ANCOVA suggests that they are significantly different (p<0.05). The Q70% low flow of the two periods was 2.7 and 2.4 l . s -1 , the Q50% median flow was 4.0 and 3.1 l . s -1 , and the Q10% high flow was 10.7 and 7.5 l.s -1 , for the two periods, respectively. The decreased flow regime of Period II, which is probably in part due to an increased evapotranspiration over the 370 past decades (Duethmann and Blöschl, 2018), indicates that streamflow cannot account for the observed increased sediment load of Period II, otherwise an increased streamflow would be expected in Period II.  Table 1 shows how land use changed between the two periods. During Period I, 380 cropland and grassland accounted for 73% to 82% (varying between years) and 14% to 22%, respectively. However, due to agricultural intensification, cropland increased to around 82% in Period II, at the expense of a decreasing share of grassland. Forest, including sparse forest, accounted for 1.8% area during Period I but increased considerably until Period II to around 11%. The increase in cropland and forest 385 suggest higher rates of evaporation and transpiration, and consequently less streamflow, which is in line with the previously examined trend dynamics of streamflow. When further analyzing the land use classes of arable land, a substantial change was found for the crops types too, with the crops of low risk for soil erosion being replaced with crops that exhibit a high soil loss potential. This was particularly 390 true for maize. In addition, the diversity of crops decreased considerably (Table 2).

Change in land use and land organization
This shift towards agricultural uniformity likely acts as a land structure effect. A loss of heterogeneity of crop types increases the probability that different fields are managed with the same crop. Then even smaller fields may behave similar to larger fields in terms of sediment production. 395 Besides the change in land use, the parcel structure of the catchment has also changed (Table 1). This change was related to a land consolidation plan issued in 1955 in Austria (Devátý et al., 2019) and a massive trend to agricultural industrialization that evolved after 1945 (mainly referring to the massive application of advanced 400 machinery and fertilization technologies that started right in the 1950s). During Period I, arable land was fragmented across many small parcels, with a mean parcel size between 0.5 -0.6 ha and a parcel density (number of parcels per ha area) between 1.7 -2.0 ha -1 in different years. In Period II, these values increased considerably to mean parcel sizes between 1.7 -2.7 ha and parcel densities between 0.3 -0.6 ha -1 . Similarly, 405 the mean parcel size and parcel density of grassland during Period I were 0.13 -0.17 ha and 5.2 -7.2 ha -1 . It had changed to 1.06 ha and 0.9 ha -1 in Period II. As parcel structures are identified influencing sediment loads mainly due to the boundary effects (e.g. Baudry and Merriam, 1988;Takken et al., 1999;Phillips et al., 2011), the substantial decrease in parcel density and the increase in parcel size of the catchment 410 in Period II, was expected to affect sediment load as well.

Table 1 Parcel and land use statistics for Periods I and II. Land use for the years 1946 to 1949 represents Period I, land use for the years 2007 to 2012 represents Period II; N is the number of parcels for a given land use, density is the number of parcels per ha, mean size 415
represents the mean area of parcels with a particular land use.   Figure 5 shows the fitted sediment rating curves (p<0.05) for both periods. An ANCOVA suggests that the slopes of the regression lines are significantly (p<0.05) 430 different between the dormant seasons or growing seasons. Although rainfall erosivity of Period II_G was similar to that of Period I_G (Figure 3a) and streamflow of Period II was generally lower than that of Period I (Figure 4), the fitted SRC of Period II_G was steeper than that of Period I_G (Figure 5a), with the coefficients b being 0.3 and 1.6 for Period I_G and Period II_G, respectively (Table 3). The fitted SRC of Period 435 II_D demonstrated a faster response of sediment concentration to increasing flow compared to that of Period I_D (Figure 5b), the coefficients b being 0.8 and 1.7 for Period I_D and Period II_D, respectively. However, the rainfall ED in Period II_D was generally lower than that of Period I_D (Figure 3a), suggesting a lower probability of a substantial increase in sediment availability. These results indicate 440 that neither changes in rainfall erosivity nor the hydrological regime could explain the increase in sediment dynamics.

Relationship between coefficient a and b
The changing steepness of a fitted SRC does not necessarily imply a change in the sediment regime as the slopes of fitted SRC are sometimes affected by catchment size or the distribution of samples (Asselman, 2000).   the dormant seasons of Period I and Period II at flow rates lower than Q30%.. The lack of a statistically significant difference (p=0.07) of sediment loads for Period II_D (5.4 500 ±18.3ton per month) and Period I_D (1.3±3.9 ton per month) indicates the difficulty in examining the change in sediment regime for the dormant seasons, but is is much less relevant for the annual sediment budget.  The stringent requirement on flow condition to detect anobvious increase of transport regime in Unlike the situation during high flow rates, the Total_effect showed an almost zero value at flow rates less than approximately 2 l . s -1 (i.e. Q70%), suggesting no difference in sediment load between Periods I and II at low flow conditions. The increase in sediment concentration, at flow rates of 2 up to around 20 l.s -1 , seemed to be mainly 525 caused by the increase in the cropland area of Period II, as the contribution from

Parcel_effect versus LUCC_effect
LUCC_effect was consistently higher than that of the Parcel_effect.
One may note that forest cover increased considerably from Period I to Period II. It, however, did not show an influential role in erosion control. We hypothesize that even though a beneficial effect of forest increase (up to a total of 11% of the catchment)

Discussion
The industrial intensification of agriculture implemented during the last 70 years has 545 raised considerable concerns regarding erosion and sediment loading of rivers (e.g. Bakker et al., 2008;Chevigny et al., 2014). However, with global climate warming, the different contributions of LUCC, land policy adjustments such as land consolidation, and climate change affecting sediment load remain not well understood.
This paper aims at evaluating the relative roles of climate change, LUCC, and land We found that the sediment load increased almost six fold from Period I to Period II.This finding is supported by estimates of the management factor (C-Factor) and the slope and slope length factor (SL-Factor) of the RUSLE for Period I and Period II 555 (Fiener et al., 2020). While the mean C-Factor of the HOAL catchment increased from 0.16 in Period I to 0.33 in Period II, the SL-Factor increased from 0.76 to 0.96.
Added together, the changed values for these two factors increased the theoretical soil loss within the catchment by over150%. This value is smaller than the changes observed, however it should be noted, that the RUSLE has not been designed to 560 account for sediment loads of catchments but to estimate field scale soil loss within catchments. This may explain the observed differences to a certain extent.

Potential interference of different sampling methods
Due to technical advancement over the long investigation period, different sampling methods, i.e. grab sampling and automatic equal-discharge-increment sampling, were 565 used in this study for Periods I and II, which may affect both rating curve estimation and sediment load estimation (Harmel et al.,2010). Thomas (1988) found that sampling method (random or systematic versus discharge-based sampling), may bias sediment load estimates, but Groten and Johnson (2019) suggested that, for sediment with very fine textural composition, these biases may be small. In our study catchment, 570 the topsoil of the catchment is constituted of 75% silty loam, 20% silty clay loam, and 5% silt according to the USDA soil classification (Picciafuoco et al., 2019).This very fine soil texture provides confidence to attribute the change in sediment load to internal and/or external driving agents (such as climate, land surface processes), instead of a methodological discrepancy. In addition, to account for the different 575 sampling frequencies, we employed a relationship of parameter a and b of the rating curves, which has the merit of not varying with the sampling methods (see section 2.3.2). However, a further study on the effects of sampling methods in future might be shed additional light on the issue. land consolidation had no apparent adverse effect on erosion, but with increasing flow, the contribution to sediment load increased continuously, leading to a dominant role at high flow rates. This finding is partially in line with David et al. (2014) andCantreul 590 et al. (2020). They reported that landscape structure was less important for soil erosion than land use and land cover during normal flow conditions. However, they did not investigate whether the effect of landscape structure showed a dynamic behavior with increasing flow. In contrast, the LUCC_effect, i.e. the increase of crops with high erosion risk, continuously affected sediment load with gradually decreasing 595 importance at high flow conditions. Similar results were reported by Vaughan et al. (2017), who showed that sediment concentration at low and median flow conditions was considerably associated with a change in catchment characteristics, primarily land use and land cover. Although the role of land use changes was dominating for flow conditions less than Q20%, it's contribution to the total annual sediment load was 600 small. More than 75% of the total sediment load was transported during a small number of events (25 events in Period I, 8 events in Period II) and all events had flow rates above 15 l.s -1 (approximately at Q13% in Period I or Q4% in Period II, respectively), which underlines the importance of land structure for sediment loading.

Dynamic relevance of land consolidation in controlling sediment
The dynamic relevance of vegetation and land consolidation in sediment production is 605 associated with the processes and mechanisms controlling overland flow as a transporting agent for sediment (e.g. Sun et al., 2013;El Kateb et al., 2013;Nearing et al., 2017;Silasari et al., 2017;Kijowska-Strugała et al., 2018). A change in land use and land cover implies alterations of surface characteristics, such as above ground structure morphology, litter cover, organic matter components, root network (Gyssels 610 et al., 2005;Wei et al., 2007;Moghadam et al., 2015;Patin et al., 2018) and soil properties (Costa et al., 2003;Moghadam et al., 2015). These properties influence the protective role of vegetation in soil detachment, the flow capacity to transport sediment particles, and runoff flow paths to the stream channels (Van Rompaey et al., 2002;Lana-Renault et al., 2011;Sun et al., 2018). The protective effects tend not to 615 linearly increase with increasing surface runoff. Accelerated discharge and stronger scouring effects of upslope discharge might impair the protective role of vegetation (e.g. Zhang et al., 2011;Santos et al.,2017;Bagagiolo et al., 2018;Yao et al., 2018;Wang et al., 2019). Vegetation usually exhibits a smaller interception capability at high rainfall intensity, resulting in an enhanced splash erosion and availability of 620 mobile soil particles (Cayuela et al., 2018;Magliano. et al., 2019;Nytch et al., 2019).
However, the decreasing contribution of the LUCC_effect does not directly imply an absolute decrease of the magnitude of the LUCC_effect. The absolute change in SSC resulting from LUCC reveals an increasing trend as flow rates increase. Thus, the contribution of the LUCC_effect stands for the relevance of LUCC in erosion control 625 compared to the change due to land consolidation. The magnitude of the LUCC_effect probably depends mainly on where within the catchment the land cover is changed and how the proportional area of various land uses changes. We will address this topic in future analyses.
Unlike land cover and landuse change, landscape structure is usually combined with 630 other catchment properties, such as slope characteristics and soil properties (Gascuel-Odoux et al., 2011) and additional erosion and transport factors (Verstraeten et al., 2000), exerting a more complicated influence on erosion. For example, the effect of landscape structure on soil erosion may be identified on moderate slopes, while on steep slopes it may be concealed by on-site severe soil erosion (Chevigny et 635 al., 2014). However, the key process for erosion control is the fact that landscape elements and their structural position (i.e. parcel structure, field boundaries, hedges and similar) alter hydrological connectivity between land and water. This is particularly true when the land cover on both sides of boundaries is different (Van Oost et al., 2000). Reducing  connectivity significantly and results in a substantial off-site damage effect, irrespective of on-site erosion of the investigated land use (Boardman et al., 2018;Devátý et al., 2019). During low and median flow conditions, surface runoff and sediment may arrive to a lesser extent at field boundaries due to efficient interception effects of the vegetation cover. This may explain the identified dynamic relevance of 645 land structure change in sediment load found here.

Conclusions
Climate change, land use and land cover change, and other human-associated activities are widely regarded as potential agents driving hydrological change. 650 Understanding the relevance of each of these agents in the hydrological cycle is critical for implementing adaptive catchment management measures and addressing climate change. Although very significant climate change influences in the last decades have been identified for certain components of the hydrological cycle we found that climate change expressed in changes in rainfall erosivity and precipitation 655 cannot explain the increased sediment production between 1946-1954 and 2002-2017 in the investigated catchment. Instead, both land cover and land consolidation played important, dynamic roles in erosion and sediment production.
The relevance of land use and land cover change versus land consolidation change varied dynamically with changing flow conditions. The reduction in parcel density 660 undoubtedly increased sediment load, particularly at higher flows due to the decreased capacity of trapping sediment particles between parcels and increasing flow lengths herein 删除[胡杨 [2]]: Our findings are also supported by the calculation of the management factor (C-Factor) and the slope and slope length factor (SL-Factor) of the RUSLE for Period I and Period II.
While the mean C-Factor of the HOAL catchment increased from 0.16 during Period I to 0.33 for Period II, the SL factor increased from 0.76 to 0.96 from Period I to Period I. Taken together, the changed values for these two factors increase the theoretical soil loss within the catchment by over150%. This is smaller than the changes observed, however it should be noted that the RUSLE has not been designed to account for sediment loads of catchments but to estimate field scale soil loss within catchments. This may explain the observed differences to a certain extent. inside parcels. Unfavorable land use or land cover change increased sediment load at most flow conditions, although the relevance of this process decreases at high or very high flow rates. Therefore, when addressing soil conservation measures at the 665 catchment scale, the distribution of fields, land structure, and vegetation cover should be simultaneously considered. Such a strategy would be conducive to dealing with the risk of soil erosions at different flow rates. Land use policy adjustments resulting from technological development have been vital to deal with food security issues in the past.
However, now we experience the negative influence of these adjustments on the 670 hydrological cycle. Therefore, rather than focusing on climate change solely, we need to pay increased attention to anthropic management activities to counteract their negative impact on hydrological change effectively.

675
Shengping Wang has led the data analysis, drafted the manuscript, and revised the manuscript; Peter Strauss was responsible for the project design, oversaw the whole analysis, and conducted manuscript revision as the project leader; Carmen Krammer was responsible for data collection and data preparation; Elmar Schmaltzhas contributed tofigure drawing and manuscript revision; Borbala Szeleshas helped to 680 revise the manuscript, and GünterBlöschl oversaw and critically reflected on the manuscript writing as the senior scientist.

Competing interests
The authors declare that they have no conflict of interest.

Disclaimer
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.