Nonlinear carbon cycling responses to precipitation variability in a semiarid grassland

Highlights

• A field study showed nonlinear C cycling-precipitation variability relationships.

• Soil water and nitrogen availability controlled the nonlinear relationships.

• The nonlinear relationships cannot be captured by DNDC model.

• Precipitation variability-C cycling relationships should be incorporated in models.

Abstract

Changes in precipitation amount and variability would profoundly affect carbon (C) cycling in arid and semiarid grasslands. However, compared to the effects of precipitation amounts, little is understood about the impacts of precipitation temporal variability on terrestrial C cycling. To explore relationships between precipitation variability and C cycling processes and the underlying mechanisms, we conducted a 3-year field experiment and a 12-year model simulation, in which the constant seasonal precipitation amount was temporally manipulated with four and six levels of precipitation variability, respectively, in a semiarid grassland. Based on the manipulative experiment, we found various nonlinear relationships between C cycling processes and the coefficient of precipitation variability (P_cv), including a trinomial relationship for soil respiration (tipping point: 3 and 4.3), convex relationships for gross and net ecosystem production (peak at 3.5) as well as belowground biomass (peak at 4) and a nonlinear negative relationship for ecosystem respiration (peak at 2.5). Such relationships were regulated by seasonal averaged soil water content (SWC), early-growing season precipitation amount, soil inorganic nitrogen availability (SIN), and both SWC and SIN, respectively. However, these results from the manipulative experiment did not match those from the model simulation, in which ecosystem C cycling processes, dominated only by SWC responses, showed positive linear responses to P_cv. Our results mirror that the nonlinear responses of grassland C cycling to precipitation variability as regulate by SWC and SIN should be incorporated into models to forecast future ecosystem shifts under climate change.

Introduction

Arid and semiarid regions cover ca. 40% of global land area (Safriel and Adeel, 2005), 87% and 54% of which are occupied by grasslands, respectively (UN, 2011). Overall, annual mean precipitation will likely decrease but precipitation variability will likely intensify in many arid and semiarid regions (Stocker, 2014). This combination of changing precipitation mean and temporal variability has great and complex impacts on carbon (C) cycling processes in arid and semiarid grasslands (Gherardi and Sala, 2015b; Knapp et al., 2002). Thus, quantifying the magnitude of arid and semiarid grassland responses to changing precipitation is crucial for accurately evaluating future global C budgets.

Impacts of precipitation variability on terrestrial C cycling have received much less attention than effects of changes in precipitation amount. Nonetheless, in some cases, increased precipitation variability characterized by larger event size and longer dry intervals may have more implications on terrestrial C cycling than that of decreased precipitation amounts (Fay et al., 2000). However, the majority of such studies manipulated only two levels of precipitation variability (e.g. ambient control and prolonged drought treatment; Thomey et al., 2011; Wilcox et al., 2015) rather than multiple levels (see Heisler-White et al., 2008; Gherardi and Sala, 2015a). Thus, our understanding of relationships between C cycling and precipitation variability was largely limited, especially with respect to nonlinear responses to precipitation variation (Knapp et al., 2017; Y. Luo et al., 2017).

Ecosystem attributes are controlled by resources availability such as water and nutrients, which are also the vital regulating factors for ecological responses to climate change. According to Liebig's Law of the Minimum, the least available resource is the limiting factor at a particular time (von Liebig and Gregory, 1842). A growing body of evidence suggested that many ecological processes and functions are co-limited by multiple resources (Harpole et al., 2011). Furthermore, the primary resource limitation could shift from one toward another, or from single-resource limitation to multiple-resources co-limitation or conversely, for a specific ecosystem by climate change (Falkenberg et al., 2013; Ren et al., 2017). Determining causality between observable ecological responses (e.g. biomass production, CO₂ fluxes) and the dominant resource drivers (e.g. water availability or nutrient content) under climate change is of great value for understanding and forecasting ecosystem dynamic and feedback in the future.

It is a common view that better soil water status would enhance C cycling, as water is the primary limited resource for arid or semiarid ecosystems. Changed precipitation variability would be expected to directly alter soil water availability and subsequently, ecosystem C cycling processes. Past two-level precipitation variability studies suggested that the direction and magnitude of alteration in water availability caused by high precipitation variability varied depending on ecosystem type (e.g. xeric vs. mesic system; Heisler-White et al., 2009; Knapp et al., 2008) and the annual precipitation amounts (e.g. dry vs. wet year; Thomey et al., 2011). Additionally, another potential explanation for the inconsistent results from the past two-level experiments may relate to nonlinear responses of soil moisture to precipitation variability. Soil water availability may reach its peak at moderate precipitation variability, because small precipitation events are just able to wet the soil surface and are quickly evaporated, while large precipitation events increase potential water losses from the ecosystem through runoff, deep soil water percolation or relative high evaporation (Huxman et al., 2004; Yin et al., 2018). But these potential mechanisms have not been well-examined due to a lack of experiments with multiple-level precipitation variability gradient. Furthermore, in contrast to the common view, altered water availability may not always result in expected C cycling responses under changed precipitation variability. For example, increased precipitation variability did not change seasonal average soil moisture, but largely promoted productivity in a shortgrass prairie; in contrast, increased precipitation variability increased soil moisture but did not affect the productivity in a tallgrass prairie (Wilcox et al., 2015). Taken together, above-mentioned studies suggested that predicting changes in C cycling, mediated by soil water availability, is not straightforward.

Nitrogen (N) availability, another important resource regulating ecosystem C cycling (Wieder et al., 2015), may also be profoundly impacted by precipitation variability in complex and counteracting ways. First, if altered precipitation variability increases (decreases) seasonal average soil moisture, we may expect increased (decreased) mineralization of organic N to plant-available inorganic N (Guntiñas et al., 2012; Knoepp and Swank, 2002; Larsen et al., 2011). Second, higher seasonal average soil moisture would increase soil N loss through N₂O emission (Li et al., 2019b). Third, mineralization of microbial biomass killed by water stress may promote the accumulation of inorganic N during the prolonged drought intervals of an extreme precipitation regime (Liu et al., 2014). Finally, the subsequent, large precipitation pulse may stimulate abnormally high N₂O emissions and soil inorganic N loss (Li et al., 2016). Collectively, changes in dry-wet cycling and average soil moisture may result in diverse and even opposite responses of N availability. Ultimately, the direct effects of precipitation variability on C cycling may even be overwhelmed by the impacts of altered N availability (Ren et al., 2017) on processes including photosynthesis (Reich et al., 2009), plant biomass production (Wang et al., 2019) and ecosystem CO₂ flux (Zong et al., 2013). However, the responses of N availability and their effects on ecosystem C cycling are often overlooked in precipitation variability studies.

To quantify the relationships between C cycling and precipitation variability, here we experimentally imposed four levels of precipitation frequency and magnitude without changing the total growing season precipitation amount over a semiarid grassland from 2014 to 2016. Furthermore, to examine the above relationships at more precipitation alterations and a larger timescale, a process-oriented biogeochemistry model, DNDC, which is calibrated and validated at this site (Kang et al., 2011), was employed to model biomass carbon and CO₂ fluxes under six precipitation variability simulation scenarios during growing seasons from 2000 to 2011. Specifically, we hypothesized that 1) seasonal mean soil water content (SWC) will have convex relationships with increasing variability of precipitation. 2) Accordingly, CO₂ fluxes as well as seasonal biomass production will peak at moderate precipitation variability, controlled primarily by the water availability. 3) Changed precipitation variability induced various in N availability may exert measurable control on CO₂ fluxes and biomass accumulation for conditions when water availability is not limiting or similar across treatments.