How plant production in the Mongolian grasslands is 1 affected by wind-eroded coarse-textured topsoil 2

21 While it is known that soil erosion by wind in drylands results in soil loss and redistribution 22 and changes the texture of topsoil, there is little information about how these changes in the 23 topsoil might affect the productivity of vegetation and if they result in degradation of the 24 grasslands in wind-eroded regions such as Mongolian grasslands. In this study, we compared 25 two different scenarios of vegetation growth, namely a wind-eroded scenario and an actual 26 field condition, on two different grasslands in Mongolia (steppe and desert steppe) using an 27 ecosystem model. The simulations of the wind-eroded scenario were based on a topsoil (0–0.1 28 m depth) with 1% clay and 99% sand, designed to represent an extremely wind-eroded soil 29 surface that had permanently lost the fine clay particles and had gained sand particles. The 30 effects of temperature, nutrient and water stresses on plant production were quantitively 31 estimated. The model gave reasonably good simulations of the vegetation and soil water 32 dynamics during the growing seasons (April – September) from 2002 – 2011. The simulation 33 results showed that water had more effect on plant production than nitrogen and temperature 34 at the two sites, and stresses because of a lack of water and nutrients generally affected plant 35 production in the wind-eroded coarse-textured topsoil. Plant production was 20.2% lower in 36 the wind-eroded scenario than in the actual field condition in the desert steppe under water- 37 stressed conditions but plant production was slightly higher (5.0%) in the wind-eroded scenario on the steppe that received more rainfall, because of a reverse texture effect, where 39 water continues to infiltrate from the coarse topsoil (0–0.1 m depth) to the deeper root-zone 40 (0.1–0.3 m depth) because of lower evapotranspiration from soil, and facilitates growth. 41 When this happens, there is enough soil moisture in the root-zone, and plant growth is mostly 42 affected by the nitrogen supply. 43


49
The area of degraded land in dryland areas is increasing at an alarming pace, threatening food 50 security and environmental quality (UNCCD, 1994). Soil erosion, mainly by wind and water, 51 is the main driver of land degradation. Wind erosion of soil is a global phenomenon that 52 occurs in arid and semi-arid regions worldwide (Shao, 2008;Shinoda et al., 2011). Soil that is 53 eroded by strong wind causes aeolian dust events that threaten human and livestock health, 54 present risks to life, and cause environmental problems, such as land degradation and air 55 pollution, and economic losses in both the source and downwind areas. particles that were supplemented through saltation (Shao, 2008) from the windward side. This 156 type of texture may be common in topsoils in severely degraded grasslands, particularly these 157 Mongolian grasslands, that eroded by wind. The simulations of the actual field condition were 158 based on the observed data of the soil texture at the steppe and desert steppe field sites (Table  159 1). 160 161

Model description 180
The DAYCENT model is a process based terrestrial ecosystem model that simulates how 181 fluxes of carbon, nutrients (e.g., nitrogen, phosphorus and sulfur), and trace gases in the 182 atmosphere, soil, and plants, change in response to human activities, such as fire and grazing 183 (Del Grosso et al., 2001;Parton et al., 1998). This model is the daily time-step version of the 184 CENTURY biogeochemical model (Parton et al., 1994), and includes routines for simulating 185 the movement of nutrients and water through soil layers, plant growth, and many other 186 ecosystem components. The model input variables include (1) climate variables (daily 187 maximum and minimum air temperature, and daily precipitation), (2) site-specific variables 188 such as soil properties (texture, depth, pH, bulk density, and field capacity) (Table 1) intercepted is a function of the plant biomass and the amount of rainfall (Parton, 1978). When 207 the daily air temperature is below freezing, precipitation is assumed to fall as snow and is 208 accumulated in the snowpack. 209

11
In addition, the soil organic matter and nutrient sub-models represent the flow of carbon (C), 210 nitrogen (N), phosphorus (P), and sulfur (S) in plant litter, and different organic and inorganic 211 soil pools . The N sub-model, which has the same structure as the soil C 212 sub-model (Parton et al., 1994), was the focus of this study. The N flows are equal to the 213 product of the C flows and the C/N ratio of the state variable that receives the C. The inputs of 214 N can be calculated using equations for atmospheric deposition, and soil and plant N fixation. 215 The N losses due to leaching are related to the soil texture and the amount of water moving 216 through the soil profile (Parton et al., 1994(Parton et al., , 1996(Parton et al., , 2001Del Grosso et al., 2000). 217 218 Simulating water, temperature, and nitrogen stresses with the DAYCENT model 219 In the DAYCENT grassland sub-model, plant production is controlled by initially having soil 220 moisture and temperature at a maximum, and then decreased if the soil nutrient supply is 221 insufficient. The grassland model also includes the effect of shading from dead vegetation, 222 while the forest model includes the effect of live leaf area on plant production (Parton et al., 223 1993). During the simulation processes, the maximum potential (or genetic maximum) above-224 ground plant production (AGPmax), not limited by temperature, water, or nutrient stresses, is 225 primarily determined by the level of photosynthetically active radiation, the maximum net 226 assimilation rate of photosynthesis, the efficiency at which carbohydrates are converted into 227 plant constituents, and the rate at which respiration is maintained (van Heemst, 1986). Thus, 12 the parameter for AGPmax has both genetic and environmental components. The potential 229 production (AGPpot) is a function of the AGPmax for grassland and 0-1 environmental scalars 230 depending on soil temperature, soil water status, shading from dead vegetation, and seedling 231 growth (Parton et al., 1993). Here, seedling growth and shading from dead vegetation will 232 have a negligible effect on AGPpot, because the seedling growth for grass is not limited 233 (Parton et al., 1992). Also, the shading effect on AGPpot is a response surface that depends on 234 the amount of live and dead vegetation. We found that amounts of observed above-ground 235 live and dead vegetation in the Mongolian grasslands ( Fig. 2) were lower than the threshold 236 values at which shading occurs and shoot senescence increases (150 and 60 g m -2 , 237 respectively) (Parton et al., 1992). Therefore, we assumed that the soil water status and 238 temperature were the main controls on the AGPpot in the Mongolian grasslands (Eq. 1). The 239 effects of soil water availability and temperature on plant production (Water stress, Wstress; 240 Temperature stress, Tstress) were calculated as shown in Eq. (2) and Eq. (3). 241 (1) 242 Where ST is an environmental scalar of soil temperature, which is calculated as a function of 245 air temperature and the optimum plant temperature. Sw is an environmental scalar of soil 246 moisture statue, which is identified by the soil-water sub-model (Parton et al., 1993). The 247 13 values of Tstress and Wstress are both range from 0 to 1, and the values close to 1 indicate the 248 maximum stress on plant production. 249 The plant production also decreases if there is insufficient mineral nutrient for uptake and to 250 satisfy the C/N ratio for producing plants. The actual production (AGPact) is limited to what 251 can be achieved with the nutrient supply available at the time with plant nutrient 252 concentrations Eq. (4). We assumed how the AGPact was affected by the lack of nitrogen 253 (nitrogen stress, Nstress) as shown in Eq. (5). 254 Where SN is an environmental scalar of nitrogen insufficiency, which is identified by the soil 257 organic matter and nutrient sub-models. 258 In this study, we mainly focused on the changes in the effects of Wstress and Nstress on AGPact 259 for the actual condition and the wind-eroded scenario during the critical growing season 260 (June-August). Eq. (6), Eq. (7) and Eq. (8) were therefore proposed to examine coarse-261 textured topsoil impacts on AGPact: 262 Where ΔAGPact, ΔWstress, and ΔNstress are the differences in the actual plant production, water 266 14 stress, and nitrogen stress between the actual condition and the wind-eroded scenario. The 267 subscripts "actual" and "eroded" denote the actual condition and the wind-eroded scenario, 268 respectively. 269 The changes in the plant production between the actual condition and wind-eroded scenario 270 (ΔAGPact) were examined. For example, when ΔAGPact < 0, the AGPact in the actual condition 271 was higher than in the wind-eroded scenario. The reasons for the changes in ΔAGPact from the 272 ΔWstress and the ΔNstress were also analyzed. When ΔWstress > 0 and ΔNstress ≤ 0, plant production 273 is mainly limited by water, while when ΔWstress < 0 and ΔNstress > 0, plant growth is limited by 274 nitrogen. 275

Site-level model parameterization 277
Previous studies have shown that global grassland ecosystems can be simulated using 278 relatively few data of site-specific parameters that change as the circumstances change 279 (Parton et al., 1995;1998). The DAYCENT model was parameterized and calibrated with the 280 field experiment data (soil physical and chemical properties, and vegetation) at BU 281 (Nandintsetseg and Shinoda, 2015) and TsO (Table. 1). AGPmax, and the optimum and 282 maximum temperatures for production were parameterized using information from the 283 agrometeorological database for the Mongolian grasslands (IMH, 1996). 284 The soil and vegetation were in equilibrium for the actual condition and wind-eroded scenario 285 at BU and TsO, historical simulations were processed in DAYCENT for 1980 years by 286 repeating the long-term climate averages over 32 years   community (e.g., Parton et al., 1993;1995), these results suggest that the DAYCENT model 331 can give reasonable simulations of seasonal and inter-annual changes in soil moisture and 332 AGM in the steppe and desert steppe ecosystems. Generally, the rainfall was higher (soil 333 moisture), and the evapotranspiration was lower, at BU (steppe) than at TsO (desert steppe). 334 The modelled and observed data therefore showed that the conditions were more favorable for 335 higher plant production at BU than at TsO. periods, respectively. Our results are consistent with a previous study on the Mongolian 361 grasslands by Nandintsetseg and Shinoda (2011), who reported that the emergence coincided 362 as the trends in soil moisture changed from decreasing to increasing as temperature increased. 363 To assess how plant production was influenced by textural changes in the wind-eroded 364 topsoil, we compared the simulations of the actual condition and the wind-eroded scenario at 365 both sites from June to August. Figure 4 shows Table 2 shows that the ΔAGPact was less than 0 for 10 months at BU and 28 months at TsO 370 while the ΔAGPact was greater than 0 for 20 months at BU and 2 months in 2006 summer at 371 TsO. At TsO (Fig. 4-d1), plant production was mainly controlled by water (ΔWstress > 0 and 372 ΔNstress ≤ 0) for 28 months when ΔAGPact was less than 0. Nitrogen (ΔWstress < 0 and ΔNstress > 373 0) was the main control on plant production for only 2 months during the summer of 2006, 374 which was wetter than normal, when ΔAGPact was greater than 0. This shows that plant 375 production decreased in the wind-eroded scenario at TsO generally, and this decrease was 376 caused by an increase in water stress. At BU (Fig. 4-d2), when ΔAGPact was less than 0, plant 377 production was mainly controlled by water for 7 months, and by nitrogen for 2 months in 378 June 2002 and June 2003 because of the higher precipitation in previous month. When 379 ΔAGPact was greater than 0, plant production was primarily controlled by water for 5 months 380 20 and by nitrogen for 14 months. The results show that this higher plant production at BU 381 (+5.0%) was mainly because the water stress decreased, and plant growth was thereafter 382 controlled by nitrogen in the critical growing season. 383 384

385
When we compared the simulations of the AGPact between the actual condition and the wind-386 eroded scenario, we found that plant production in the wind-eroded topsoil decreased by 387 20.2% in the desert steppe area (TsO) and slightly increased by 5.0% in the wetter steppe 388 (BU) during the critical growing season from 2002 to 2011. These results indicate that the 389 effects of the coarse-textured topsoil on the plant production may change with variations in 390 the environmental conditions in the Mongolian grasslands. In the desert steppe area, the 391 annual precipitation was lower (Fig. 5-a1) (between 38.2 and 164.5 mm from 2002 to 2011) 392 than at BU. In both scenarios, this precipitation did not penetrate into the deeper soil but was 393 stored in the topsoil layer (Figs. 5-a2 and 5-a3). The coarse topsoil meant that the water 394 holding capacity in the wind-eroded scenario was low, and the evapotranspiration was greater 395 than in the actual condition (Fig. 5-a4). Therefore, there was less soil moisture in the root-396 zone (0.1-0.3 m depth) in the wind-eroded scenario than in the actual condition (Fig. 5-a5). 397 Previous studies have stated that biological processes in arid ecosystems with extremely low 398 precipitation were mainly controlled by water (Noy-Meir, 1973). Field studies have also 399 shown that, when the available precipitation was less than 200 mm, the productivity of natural 400 grass was low and did not differ significantly under different N fertilizer application rates 401 (Smike et al., 1965). Also, Nandintsetseg and Shinoda (2011)  In contrast, in the wetter steppe (BU), the main reason for the slight increase in plant 408 production was a decrease in water stress. As shown in Fig. 5-b1 5-b3). The evapotranspiration from the coarse topsoil in the wind-eroded scenario was 413 therefore lower than that from the loamy topsoil in the actual condition (Fig. 5-b4). Together, 414 these factors mean that there was more water in the root-zone to facilitate plant growth (Fig.  415 5-b5) in the wind-eroded scenario. Previous researchers have called this the inverse texture 416 hypothesis (e.g., Noy-Meir, 1973, Sala et al., 1988), and it has been observed in sandy regions 417 with relatively high precipitation (< 300 mm). After adding equal amounts of water in 418 22 experiments in a warm greenhouse, Alizai and Hulbert (1970) showed that evaporation was 419 often greater from a loam bare soil than from sand. Field studies (Sala et al., 1988;Yang et al., 420 2009) have shown that the plant production was higher in sandy soils with lower water 421 holding capacity in wetter grasslands than in loamy soils with higher water holding capacity 422 on the Tibetan Plateau and the Central Grassland region of the US. Although the plant growth 423 is limited by the soil available water in semi-arid regions where inverse texture effects are 424 obvious, it is more sensitive to nitrogen stress when there is enough soil water during the 425 summer, as also reported elsewhere. From their field study, Hooper and Johnson (1999)  426 reported that plant production was limited by both water availability and nitrogen in a semi-427 arid region, but that production responded positively to N additions as the water availability 428 increased. Also, Kinugasa et al. (2012) reported that, when N was added to the wetter steppe 429 soils at BU, had more effect on plant production in wetter years (when the water stress was 430 lower) than in drier years (when the water stress was higher). 431 Relatively few studies have discussed changes in vegetation because of changes in soil texture 432 driven by wind erosion as they occur slowly (Lyles, 1975(Lyles, , 1977Larney et al., 1998), and may 433 only become noticeable after several years or decades. Previous studies found that the sand 434 content in surface soil increased by 6.   Table 1. Model parameterizations of the vegetation, soil, and meteorological characteristics 768