TRIPLEX-Mortality model for simulating drought-induced tree mortality in boreal forests: Model development and evaluation

Highlights

• First attempt to integrate both hydraulic failure and carbon starvation mechanisms into a process-based model for boreal forests.

• The shape parameter for calculating percentage loss of conductivity is the most sensitive parameter to simulate tree mortality.

• Water limitation is a key driver for drought-induced tree mortality in the western canadian boreal forests.

Abstract

Globally, increasing drought-induced tree mortality rates under climate change are projected to have far-reaching effects on forest ecosystems. Among these forest systems, the boreal forest is considered a ‘tipping element’ of the Earth's climate system. This forest biome plays a critical role in ecosystem services, structures and functions while being highly sensitive to drought stress. Although process-based models are important tools in ecological research, very few have yet been developed that integrate advanced physiological mechanisms to simulate drought-induced mortality in boreal forests. Accordingly, based on the process-based TRIPLEX model, this study introduces the new TRIPLEX-Mortality submodule for the Canadian boreal forests at the stand level, that for the first time successfully incorporates two advanced drought-induced physiological mortality mechanisms (i.e., hydraulic failure and carbon starvation). To calibrate and validate the model, 73 permanent sample plots (PSPs) were selected across Canada's boreal forests. Results confirm a good agreement between simulated mortality and mortality observations (R²=0.79; P<0.01; IA=0.94), demonstrating good model performance in simulating drought-induced mortality in boreal forests. Sensitivity analysis indicated that parameter sensitivity increased as drought intensified, and the shape parameter ($c$) for calculating percentage loss of conductivity (PLC) was the most sensitive parameter (average SI = -3.51) to simulate tree mortality. Furthermore, the results of model input sensitivity analysis also showed that the model can capture changes in mortality under different drought scenarios. Consequently, our model is suitable for simulating drought-induced mortality in boreal forests while also providing new insight into improving model simulations for tree mortality and associated carbon dynamics in a progressively warmer and drier world.

Introduction

Global change, associated with increasing and intensifying extreme drought, has caused widespread tree mortality events across the globe biomes (Brouwers et al., 2015; Goulden and Bales, 2019; Lewis et al., 2011; Liu et al., 2014; Peng et al., 2011; Reichstein et al., 2007). Increased drought-induced tree mortality can potentially trigger an abrupt and irreversible change in forest ecosystem structure, function, biodiversity and services (Allen et al., 2015; Anderegg et al., 2019, 2013; Choat et al., 2018; Suarez and Kitzberger, 2008).

Among all forest ecosystem types, the boreal forest, which is considered to be a critical “tipping element” of the Earth's climate system (Lenton et al., 2008), represents approximately 30% of the total global forest area (Brandt et al., 2013), containing more surface freshwater than any other biome (Burton et al., 2010) and accounting for approximately 20% of all forest carbon sequestration (Pan et al., 2011). However, recent studies indicated that boreal forests are very sensitive to drought and have suffered from widespread drought-induced tree die-off events (Hogg et al., 2008; Lenton et al., 2008; Michaelian et al., 2011; Peng et al., 2011). For example, ground observations showed that drought-induced tree mortality rates of Canadian boreal forests have increased by an average of 4.7% per year from 1963 to 2008 (Peng et al., 2011). The southern edge of the Canadian boreal forest suffered massive mortality of aspen trees 4 years after a severe drought (2001–2002), and dead biomass increased by about 29Mt (Michaelian et al., 2011). Such unexpected impacts could substantially influence ecological regime shifts and ecosystem functioning of boreal forests at multiple scales (Adams et al., 2010; Wang et al., 2012a). Although most boreal forests are resilient to extreme climate changes such as drought that are currently taking place, the predicted speed and extent of climate change is expected to be unprecedented and subsequently will pose a significant threat to the tree communities and ultimately to the health of this vital forest ecosystem (Gauthier et al., 2015). Consequently, evaluating the impacts caused by drought-induced mortality in boreal forests is of significant concern.

Compared with traditional observation and experimental approaches, models can quantify the dynamics of drought-induced mortality over a long-term and at a large spatial-scale. These can thus help us better understand and predict terrestrial climate forcing and future climate change (Choat et al., 2018; Friedlingstein et al., 2006; Purves and Pacala, 2008; Sitch et al., 2003). However, drought-induced tree mortality is complex and a number of interdependent mechanisms play important roles in this process, therefore, the parameterization of the tree mortality process in models remains a considerable challenge (Choat et al., 2018; McDowell et al., 2013, 2008). Previous attempts have been made based on empirical models that were designed to incorporate observed mortality records and meteorological data to model or predict mortality (Gustafson and Sturtevant, 2013; Mitchell et al., 2016). Although these attempts have provided insight into the drivers of drought-induced mortality, their functional reliability under future climatic conditions remains uncertain given that such conditions have yet to occur (Adams et al., 2013; O'sullivan et al., 2017). Dynamic global vegetation models (DGVMs) can represent the biogeochemical and hydrological processes of vegetation dynamics. However, tree mortality algorithms in most current DGVMs are still unable to adequately simulate drought-induced mortality because, under drought stress, they represent neither the advanced understanding of the internal state of carbon and water nor the interactions between hydraulic architecture and carbon allocation in plants (McDowell, 2011a; Mencuccini et al., 2015; Moorcroft, 2006; Powell et al., 2013; Xu et al., 2013). There is thus an urgent need to integrate more mechanistic mortality algorithms into DGVMs to better predict the ecological consequences of climate change and climate–biosphere feedbacks (Wang et al., 2012a).

Process-based models that integrate physiological mechanisms are the most promising for simulating drought-induced tree mortality (Adams et al., 2013; Choat et al., 2018; McDowell et al., 2008, 2016). Recent studies on physiological mechanisms of drought-induced mortality have identified both hydraulic failure (HF) and carbon starvation (CS) as the major pathways with recent research suggesting that it is a combination of both (Adams et al., 2017; McDowell et al., 2013, 2008; McDowell et al., 2011b; McDowell and Sevanto, 2010). Death by HF is due to partial or complete loss of xylem function caused by embolisms that inhibit water transport through the vessels or tracheids. Mortality by CS is caused by partial or complete depletion of nonstructural carbon content in the plant resulting in an inability to meet growth, metabolic and defensive carbon needs, due to an imbalance between carbohydrate demand and supply from photosynthesis (Hartmann et al., 2013; McDowell et al., 2013, 2008, 2016; McDowell et al., 2011b; O'Brien et al., 2014; Sala et al., 2010; Sevanto et al., 2014). To date, the assessment of HF is generally conducted by the percentage loss of conductivity (PLC), whereas CS is mostly quantified by the dynamics of non-structural carbohydrate (NSC) concentrations (Hartmann et al., 2013; O'Brien et al., 2014; Piper and Fajardo, 2016; Quirk et al., 2013; Sevanto et al., 2014). Research suggests that HF and CS mechanisms can be influenced by tree height and DBH (diameter at breast height), and that multiple interactions between HF and CS processes may occur during mortality and may vary with tree species (McDowell, 2011a; Olson et al., 2018; Stovall et al., 2019). By integrating new evidence from the research field, McDowell et al., 2011b concluded that most of these mechanisms and their interdependencies are expected to be amplified under a warmer, drier climate. Given that recent projections (Trenberth et al., 2014) indicate longer and more intense droughts due to land surface climate warming, accurate simulations and predictions of the effects of drought on forest ecosystems are thus highly desirable.

Simulations that exploit both HF and CS have been used to establish the importance of each physiological process involved in drought-induced tree mortality. Based on the CASTANEA model and observed mortality rates, Hendrik and Maxime (2017) simulated the functional development over time of trees with different ontogenetic and phenotypic characteristics (e.g. age and leaf area index) and growing under different site conditions (e.g. altitude, soil water content). They determined the physiological carbon and hydraulic thresholds of key plant traits and indicated the significant importance of CS in drought-induced mortality. They also highlighted that the factors which predispose and induce tree death can be identified using process-based models. However, their model only uses simplified resistance and capacitance mechanisms between soil and leaves to simulate water potential. The ability to translate meteorological data (e.g., precipitation or evaporative demands) into plant water content or xylem tension is key for simulating drought induced-mortality using process-based models (Blackman et al., 2016). McDowell et al. (2013) compared model simulations from FINNSIM (Hölttä et al., 2006), Sperry (Sperry et al., 1998), TREES (Mackay et al., 2012), MuSICA (Domec et al., 2012; Ogée et al., 2003), ED(X) (Fisher et al., 2010b; Xu et al., 2012) and CLM(ED) (Bonan et al., 2012) process-based models on tree mortality events in the USA. The predictions from these models show that the mortality of all species in their study was caused by both HF and CS. They subsequently suggest that integrating the latest information on physiological mechanisms that drive drought-induced mortality into vegetation models and comparing predictions against observations could result in an improved capacity to predict drought-induced mortality. The main strengths of the study by McDowell et al. (2013) are not only the insights it provided in evaluating different tree species using both isohydric and anisohydric strategies, but also in the improved understanding it provided for the physiological processes involved in drought-induced mortality across different species. However, the results from their study do not sufficiently account for the influence of spatial variability in tree communities and soil characteristics as well as the effect of temperature on plant growth rates and the interaction between CS and HF (Mao et al., 2013). Consequently, these models may not be suitable for application in Canadian boreal forests at a regional scale. Additionally, most Canadian carbon models, such as InTEC (Chen et al., 2000; Chen et al., 2000), CLASS (Verseghy, 2000), CTEM (Arora, 2003; Arora and Boer, 2005; Melton and Arora, 2016), CLASSIC (Melton et al. 2020), CBM-CFS3 (Kurz et al., 2009, 1992) and Ecosys (Grant and Nalder, 2000), have been successfully used to simulate carbon dynamics. However, due to the lack of physiological processes involved in drought-induced tree mortality, these models are unable to explicitly represent the dynamics of drought-induced tree mortality under a changing climate. Despite recent progress, drought-induced tree mortality has not yet been well represented in current process-based models. Accordingly, a robust process-based model that provides the best possible understanding of drought-induced mortality is still highly desirable (Adams et al., 2013; Choat et al., 2018).

TRIPLEX 1.0 is a generic hybrid model that simulates the key processes of forest growth and carbon cycle. As the TRIPLEX 1.0 model integrates the advantages of both process-based and empirical models, it bridges the gap between empirical forest growth and yield, as well as process-based carbon balance models (Peng et al., 2002). In addition, TRIPLEX effectively considers the effect of spatial variability in tree stands and soil characteristics through the using of forest and soil information such as tree species, age, stocking percent, soil carbon, and texture from each simulated plot as input to the model. To date, the TRIPLEX 1.0 model has been successfully calibrated and validated for different forest ages (Peng et al., 2002; Zhou et al., 2006), tree species (Sun et al., 2008; Zhou et al., 2004), harvest disturbance and climate change (Wang et al., 2012b; W. 2011), as well as insect disturbance (Liu et al., 2019, 2018) in boreal forest ecosystems. TRIPLEX 1.0 can therefore be applied to simulate both short and long-term forest growth and carbon dynamics of boreal forests.

Given the wide availability and high reliability of the TRIPLEX model in the boreal forest ecosystems, this study intends to develop a new mortality submodule for the TRIPLEX process-based model (TRIPLEX-Mortality) that incorporates physiological CS and HF processes to better quantify drought-induced mortality in boreal forests at the stand level. The specific objectives of this study are (1) to provide the best possible understanding of the physiological processes associated with drought-induced mortality into a process-based model; (2) to calibrate and validate the model using permanent sampling plots (PSPs) from Canada's boreal forests; and (3) to conduct parameter and climate input sensitivity analyses to identify the most sensitive parameters under different drought conditions.