4 Discussion

4.1 Response of crop yield to soil erosion

Our analysis supported the overarching hypothesis and illustrated that crop yield varied significantly with an increase in soil erosion thickness. This was in agreement with a previous integrated analysis of soil types, regions, and crops, which showed that soil erosion led to an average global productivity loss of 0.3% (den Biggelaar et al., 2004). Similarly, Wang et al. (2009) showed an average soybean yield reduction of 14.9% for every 10 cm increase in eroded soil depth. Pimentel, (2005) supported that nutrients deficient soils produce are 15 to 30% lower crop yields than non-eroded soils. In this study (Figure 5), three erosion depths (i.e., 6, 12, and 18 cm), data from Allen et al. (2009) were different than the rest. This may have been due to a reduced erosion effect as crop yield was reported 16 years after erosion simulation experiment. With these data removed, the cumulative effect size showed a consistently decreasing yield with increasing erosion depth. The early scientific evidence of the detrimental effects of erosion on crop productivity was considered sufficient (Pierce and Lal, 2017).
We found that crop yields did not change significantly when erosion affected less than 5 cm of soil (Figure 5), and this agreed with the results of previous studies (Sui et al., 2009; Rejman and Iglik, 2010). For example, slightly eroded soil (5 cm of topsoil removal) did not affect crop yields (Sui et al., 2009; Rejman and Iglik, 2010). Lack of effects may be due to the fact that similar levels of available nutrients are stored in both eroded and non-eroded topsoil (0 to 10 cm depth) (Christensen and McElyea, 1988; Larney et al., 2000; Larney et al., 2003). Many processes contributing to accumulation of nutrients occur in the surface soil regardless of erosion. For example, storage of decaying leaves and stems (David and Michael, 2013), restorative tillage, and addition of chemical fertilizers (Wang et al., 2009; Guo et al., 2010). Consequently, plant nutrients, such as alkali nitrogen, available phosphorus, and available potassium are mainly concentrated in surface soils (Huang, 2016). While water-stable soil aggregates, SOM, and particulate organic matter are found in the top 15 cm soil was not significantly different (Allen et al., 2011). In addition, bacterial biomass in the first 6 cm of soil was unaffected by the removal of 0 - 6 cm of topsoil (Allen et al., 2011).

4.2 Response of crop yield to other factors

No significant difference was observed between the cumulative effects of multiple levels of factors like Grain type and Measure in our study. This is different from other studies. For example, Sui et al. (2009) found that yield reduction in maize was greater than that in soybean. This was in agreement with the results of Lin et al. (2019), who also observed that maize yields decreased significantly with increasing erosion, but soybean yields did not. But there are obvious differences. Figure 3 is worth mentioning that soybean and maize yields are more sensitive to soil erosion than wheat yields. This finding indicates that if soil becomes less productive for one crop, it may achieve more productivity for another that is better able to exploit adverse or resource-limiting conditions.
In addition, several studies indicated that addition of manure or chemical fertilizers into eroded soils could achieve the same yield level of non-eroded soil (Schmitt et al., 2001; Hichman, 2002; Azeez, 2010; Zhou et al., 2012). Especially, a general conclusion drawn by Pierce and Lal, (2017) from 50 years of erosion and productivity research in the United States is that management inputs were sufficient to restore production to levels of undisturbed soils. Sui et al. (2009) demonstrated the value of manure to restore eroded soil productivity for corn production. For example, manure addition reduced the effect of topsoil loss, and increased corn yield by 57% in 2005 and 37% in 2006 in a 20-cm topsoil removal treatment (Sui et al., 2009). However, our results indicated that no matter what the fertilizer/manure measures used, crop yield was clearly reduced when soil was severely eroded (erosion depth > 20 cm) (Figure 4); in this case, land productivity could not be restored. Wang et al. (2009) showed that yields following fertilizer application did not reach the levels obtained in non-eroded soil. Further, the addition of fertilizer did not offset yield loss in the absence of topsoil (Allen et al., 2011). This may be due to a substantial loss of organic matter, which is mainly distributed in topsoil (0 - 20 cm) (Gu et al., 2018; Allen et al., 2011). Further, eroded soil contains about three times more nutrients per unit weight than that in the remaining soil (David and Michael, 2013). In severe erosion conditions, soils without adequate nutrients may limit plant growth, and result in yield reduction. It has been also shown that crop yield declined with an increase in topsoil removal depth (den Biggelaar et al., 2004; Wang et al., 2009; Gao et al., 2015). A decline in SOM, total nitrogen, and saturated water capacity may be the main soil factors affecting crop productivity (Sui et al., 2009; Allen et al., 2011; Gou et al., 2020). Meanwhile, significant reductions in soil clay content, significant increases in sand content, and significant decreases in water retention capacity were observed with increasing erosion levels (Duan et al., 2016; Lin et al., 2019). Decreases in nutrient levels (Wang et al., 2009), and loss of topsoil and root-restrictive layer (Gao et al., 2015) were the primary reasons for the decline in soil productivity resulting from soil erosion.

4.3 Crop yield response curve to erosion

Bakker et al. (2004) classified yield response curves to erosion into four groups: linear, convex, concave, and S-shaped. Linear curves were widely used to assess the effect of soil erosion on crop yield. However, we found significant differences in the variability in effect sizes across erosion depths of 5 cm to 15 cm. Our results indicated a larger negative response of yield in erosion depths of 5 - 15 cm than at other erosion depths. These findings support the results of Wang et al. (2009) who showed that the initial rate of decline in productivity due to early stages of soil erosion (10 cm - 20 cm) was high and significant at 15 cm of soil removal (Gorji et al., 2008). Gao et al. (2015) also showed the reduction in soybean yield was highest when the first 10 cm soil was eroded. Thus, linear relationship cannot be used to describe the relationship between erosion and crop yield. Likewise, Christensen and McElyea (1988) suggested that a linear relationship was a simplistic representation of a yield-topsoil depth relationship, as constant slope assumed that loss of each incremental depth of topsoil exerts a constant incremental decrease in soil productivity. Den Biggelaar et al. (2001, 2004) pointed out that even though the relationship between soil degradation and soil productivity was linear, linear relationships may not always represent reality; the effect of soil erosion on crop productivity depends to a large extent on intrinsic soil characteristics, especially soil loss tolerance or effective soil depth. When there is no soil to erode, crop yield is 0, or reduction in crop yield is 1. Based on various erosion depth-yield studies, our result preliminarily suggested that the crop yield response curve was convex when erosion depth was < 5 cm, and concave in other cases.