Material and methods
Characterization of drought and risk
According to the IPCC (2002), drought is defined as “a phenomenon that occurs when precipitation is significantly below normal recorded levels and that causes significant hydrological imbalances that are detrimental to systems of land resources production”. More precisely, from the ecophysiological point of view, drought is a reduction of the soil water reserve sufficiently severe to prevent the optimal functioning of trees due to insufficient precipitation, high temperature and large water uptake by trees. The definitions of drought vary greatly from country to country, but the literature identifies four different types of drought, including the edaphic (or agronomic) drought that is particularly of interest to us since it refers to the soil and to the impacts on living beings.
The precipitation regime is the first determinant in the development of a state of drought. It results from a pluviometric drought, which is a prolonged rainfall deficit compared to the mean or median (that is the normal state). However, drought also depends on the evapotranspiration level that is closely related to the temperature and atmospheric drought. The estimation of the water balance makes it possible to define the conditions under which precipitation distribution, soil water reserves and losses by evapotranspiration or drainage induce a negative effect on trees, referred to as water stress. According to Lebourgeois et al. (2005), water stress is the most important concept for the forest manager since water is the determinant of good stand health. We use the available water content (AWC) to illustrate this water stress.
According to Crichton (1999), drought risk can be described in terms of three components: the hazard, the stand exposure to the hazard and the stand’s vulnerability. The hazard is characterized by its intensity (i.e., the magnitude of the phenomenon), its severity (linked to the duration of the phenomenon), and its frequency (i.e., the probability of damage). Exposure is the level or the conditions at which the stand may be in contact with the hazard. It is a function of the geographical location and the physical context, which can limit or accentuate the hazard (e.g., compact and shallow soils). Vulnerability refers to the internal characteristics of the stand, influenced by species ecology, soil characteristics or stand density. It shows the extent to which the stand is likely to suffer from damage related to the hazard. Consequently, it takes the sensitivity of individuals to the effects of a hazard into account, as well as their ability to resist, adapt to them, and to return to the baseline situation (i.e., resilience) (UNEO, 2007). A hazard (which is only a natural process) becomes a natural risk only when there is an interaction between the hazard and the population, goods and activities affected (Veyret et al., 2013). The natural risk therefore implies the perception of this hazard by the population and, subsequently, its management (for cohabitation with the danger) (Veyret et al., 2013). Adaptation strategies will consequently play a role on vulnerability through the implementation of a water-saving silviculture.
The impacts of drought may be classified as biological or socioeconomic. Four categories of biological impacts can be distinguished: accommodation through changes in physiological functioning (Bréda and Badeau, 2008; Matesanz and Valladares, 2014), in phenology or in tree growth (Solberg, 2004; Matesanz and Valladares, 2014); genetic adaptation (de Miguel et al., 2012); and migration and tree mortality (Spiecker et al., 2004; Galiano et al., 2011; Galiano et al.,
2012). The biological impacts begin at the tree level, which result in impacts at the stand level, which, in turn, result in impacts at the ecosystem level. Thus, at the stand level, loss of growth proportional to drought intensity induces loss of productivity, whereas at the ecosystem level, drought reduces most of the biological cycles that affect the functions of the forest and that lead to a loss of ecosystem services, mainly wood production and carbon sequestration (Maroschek et al., 2009). In terms of socio-economic impacts, drought generates financial losses linked to the current value of felled timber resulting from the loss of marketability, a decrease in future stand value, the additional cost of forest restoration, and the loss of hunting and other regular income (Birot and Gollier, 2001). Additionally, drought is also linked to the loss of carbon sequestration, which generates financial and social losses, as well as the loss of other amenities such as recreation (Thürig et al., 2005).
These impacts are likely to be intensified in the near future due to climate change. Indeed, climate change is a global phenomenon due to an anthropogenic cause: the increase in the atmospheric concentration of greenhouse gases, the most important of which is carbon dioxide (CO2) (IPCC, 2013). Climate will thus evolve towards an increase in average temperature, an escalation in the differences between wet and dry regions, a decrease in water availability, and an increase in the frequency and the intensity of extreme events such as severe drought (Spiecker, 2003). However, increasing CO2 can also limit the drought effect by increasing the water use efficiency of plants (Davi et al., 2006; Keenan et al., 2013).
In order to try to limit the increasing impacts of drought, several adaptation strategies can be identified. We chose to test two main adaptation strategies according to their importance in the literature and according to the classification of soft and hard adaptation strategies4 given by the World Bank (2010): (i) the reduction of rotation length (soft adaptation); and (ii) species substitution from beech to Douglas-fir (hard adaptation). These two strategies are analysed separately as well as jointly, and in combination with a third strategy, density reduction (soft adaptation).
First, the reduction of rotation length reduces the time of exposure to a drought event and the vulnerability of trees due to aging (Spiecker, 2003; Bréda and Peiffer, 2014). Young and old trees are the most vulnerable to drought (Archaux and Wolters, 2006): Special attention must therefore be paid to the establishment of young trees and to avoiding long rotations.
Second, the introduction of drought-tolerant species and provenances reduces the aerial carbon balance, while using the same forest area (FAO, 2011; Keskitalo and Carina, 2011). Moreover, it would be preferable to introduce so-called transitional species or varieties, i.e., species able to thrive in both the current and projected future climate (e.g., pine, Douglas-fir, Robinia).
Third, the reduction of the leaf area and, therefore, of the stand density, improves the resistance of forest stands to the lack of water (Archaux and Wolters, 2006; Bréda and Badeau, 2008), reduces the intensity and duration of water deficits, and increases water availability (Spiecker, 2003). This results in an increase in initial planting space (Spiecker, 2003) and more intensive and earlier thinning (Spiecker, 2003; Keskitalo and Carina, 2011) in order to stabilize and thus protect stands (i.e., to have a continuous forest cover and to protect it from all hazards) (Spiecker, 2003; Bernier and Schoene, 2009), to take advantage of CO2 fertilization to maximize and accelerate growth (Bernier and Schoene, 2009), to increase resistance and resilience to future damage (Kerhoulas et al., 2013), and to stimulate the growth of trees remaining after a drought (Kerhoulas et al., 2013).
Case study area: Burgundy region
Burgundy is a rural region and one of the major forest regions in France in terms of afforestation (30% afforestation rate), which has increased over the last 30 years. It has a great geographic (from valleys to mountains) and geological diversity. Its contrasted climate is of the Atlantic type with rainfall spread out throughout the year, ranging from 600 mm (Loire valley) to 1500–1800 mm (Morvan peaks), average temperatures between 9.5 and 11.5°C, events of snow and frost, as well as frequent late frosts in May. However, biotic (pests and pathogens such as canker and bark beetle) and abiotic factors (e.g., late frosts, repeated water deficits, soil compaction due to forest mechanization) threaten the health of forests. Burgundy forests are characterized by private property (68% according to IGN, the French National Forest Inventory), a primary function of production, and a dominance of deciduous trees except in the Morvan. Indeed, beech and oak represent 90% of the forest areas. However, these two species are sensitive to summer water deficit and many beech diebacks can be observed, which may be amplified by a weakly dynamic silviculture. This is why, during the turnover of Burgundy stands, deciduous forests gradually shift to forests with more productive and valuable species such as Douglas-fir in order to anticipate future climate changes and to thus avoid financial losses, and to respond to the growing demand for wood, with a more dynamic silviculture. Beech and Douglas-fir also produce commercially highly-valued wood in Burgundy, i.e., their annual production is 221,000 m3 and 898,000 m3, respectively, in private forests.
Species of interest
Beech (Fagus sylvatica L.) is a natural species representing 15% of the forest production area in France. It is a typical shade-tolerant species, requiring a certain degree of atmospheric humidity and sufficient soil moisture (Latte et al., 2015), which can barely tolerate extreme conditions, as well as spring frosts (Godreau, 1992). More precisely, it is the climate criteria (annual distribution of precipitation and temperature) that determine the presence or the state of health of beech, rather than soil conditions (Godreau, 1992). However, due to climate change, this species could decline or even disappear (Charru et al., 2010). Indeed, the increase in the frequency and intensity of spring droughts and heat waves has already negatively affected the annual growth of beech trees (Latte et al., 2015). Damage can lead to the death of beech when the proportion of dead aerial biomass exceeds a threshold of 58% (i.e., percentage of foliar deficit reached) (Chakraborty et al., 2017). This mortality is directly related to the availability of water and light resources, as well as to the increase in neighboring interactions and in the diversity of tree species (Chakraborty et al., 2017). Overall, distribution in France is limited by temperature for Mediterranean species and by water supply for northern species as well as deciduous species (beech, oak) and conifers (Douglas-fir, spruce, fir). This is why the hydric constraints in the northern half of France cast doubts on the existence and the production of these latter species, particularly beech that has had many diebacks on superficial soils with low water reserves. Substitution with a species that is more productive under a dry climate and more valuable, such as Douglas-fir, seems to be a better economic solution, as suggested by Latte et al. (2015) for the regeneration of old beech stands. In addition to this, with the interest of the French public authorities (e.g., the National Forest Fund in France during the period 1946–2000) and some professionals (builders, wood producers, furniture industries) in the rapid growth, the lower cost of production and maintenance, and the standardized sawing techniques of conifers (pine, fir), the demand would be based on an accelerated national production of conifers. Since two-thirds of the French forest is composed of deciduous trees, the transition could be backed by a less water-consuming silvicultural system, which is linked to the subject of our study.
A native of western North America, Douglas-fir (Pseudotsuga menziesii Mirb.) is an introduced species valued by forest managers for its rapid growth and the quality of its wood (Da Ronch et al., 2016). It appears to be able to provide significant wood production under a relatively dry climate (Eilman and Rigling, 2012; Da Ronch et al., 2016). However, despite all these qualities, Douglas-fir is more sensitive to high temperatures due to its high leaf area (i.e., strong transpiration) than to droughts. This explains the damage reported in France after the drought in 2003 (because of its combination with a heat wave), in particular in the Burgundy region (Sergent et al., 2014). Moreover, although Douglas-fir is described by some authors as a drought-resistant species (Eilman and Rigling, 2012), it does not seem to be well-adapted to the range and accumulation of intense and recurrent episodes of drought after a severe one, which could be explained by a lack of resilience, e.g., after the drought in 2003 (Sergent et al., 2014).
Beech and Douglas-fir are both mesophilous species, i.e., species that grow in habitats that are neither extremely dry nor extremely humid (ONF, 1999). They prefer mountainous areas due to their high requirement for atmospheric moisture, although they are present in the plains. They are therefore sensitive to heat. Douglas-fir and beech have the same skewed and moderately deep rooting, but with different transpiration control during drought (ONF, 1999). Indeed, beech has a higher midday soil water potential and, consequently, a higher sensitivity to drought compared to Douglas-fir (ONF, 1999; Pierangelo and Dumas, 2012). Additionally, deciduous trees have a higher demand for available water content than conifers (ONF, 1999): Beech therefore consumes more water reserves than Douglas-fir in summer. However, edaphic drought can be aggravated by the existence of a high evaporation demand. Finally, Bréda and Badeau (2008) confirmed that the development of beech is dependent on water balance and drought, whereas for species such as Douglas-fir, its development is mainly related to temperature, supporting our suggestion to substitute beech with Douglas-fir.
For this study, we chose to test two levels of drought risk defined according to the level of soil available water capacity (AWC). Three levels of AWC were considered: 150, 100 and 50 mm. These levels were chosen according to the range of AWC of current beech stands in Burgundy. The level of 150 mm represents optimal water conditions for beech growth, 100 mm is the initial risky scenario with one-third less of the baseline level of water availability for trees, and 50 mm is the second risky scenario in which the water availability is below 40% of the baseline. This threshold of 40% of the maximum AWC represents the conditions under which beech starts to regulate water consumption and thus has difficulties to grow and survive (Lebourgeois et al., 2005).
With respect to the uncertainty of future climate, the consequences of the two extreme climate scenarios from the IPCC were analysed: RCP 4.5 and RCP 8.5 (IPCC, 2013). RCP 4.5 represents the most optimistic scenario, and RCP 8.5 represents the most pessimistic one (higher temperature, higher CO2 concentration, etc.). All of these elements result in [(2 baselines + 7 scenarios × 2 drought risks) × 2 climates], which is equal to 32 scenarios. The two baselines and the seven scenarios are summarized in Table I.1. The scenario is indicated by the following code for the benchmark (AWC of 150 mm): Baseline_Species (B for beech or D for Douglas-fir).
The scenario is indicated by the following code for both levels of drought risk (AWC of 100 mm and 50 mm): Species (B for beech or D for Douglas-fir)_Silviculture (NA for no adaptation, DR for density/rotation reduction and S for substitution). Scenarios for beech were composed of a classical path (Baseline_B and B_NA) and three dynamic ones (B_DR1, B_DR2 and B_DR3) representing the silviculture of the density/rotation reduction strategy. Simulations for Douglas-fir were composed of a classical path (Baseline_D and D_S) representing the silviculture of the substitution strategy plus two dynamic ones (D_S+DR1 and D_S+DR2) in order to test the combination of the two strategies.
To compare the adaptation options to deal with the drought-induced risk of forest dieback, we first simulated forest growth with different silvicultural treatments according to these different adaptation strategies, the three different levels of water content and the two climate scenarios. The simulations were run with the CASTANEA model. The economic approach was then applied to the outcome of the simulations.
Simulation of forest growth and silvicultural treatments
CASTANEA requires three different files as inputs: the inventory file, the species file and the weather file. First, the inventory file contains all the trees with their characteristics related to the simulated stand. R software makes it possible to generate the list of all the trees according to soil characteristics. The soil characteristics (height, stone content, etc.) are directly linked to the AWC and parameters of the managed stand (tree diameter, LAI, etc.). Second, the species file contains all the species-specific parameters that control the energy budget, growth (photosynthesis, respiration), carbon allocation and water consumption (see Table I.B1 in Supplementary Material). Third, the weather file contains the climatic characteristics of the studied site (global radiation, air temperature, relative air humidity, wind speed, precipitation). These georeferenced data for current and future climates (RCP 4.5 and RCP 8.5) came from the Météo France network for four different SAFRAN points of 8×8 km (3202, 3710, 4303, 5121), chosen to represent the variety of climates in Burgundy. All of the results for each scenario are then taken from the average of the four SAFRAN points (see Figure I.B1 in Supplementary Material).
CASTANEA simulates photosynthesis and respiration to estimate net primary production. Carbon is then allocated to six compartments following the allocation rules described in Davi et al. (2009) and Davi and Cailleret (2017): large roots, fine roots, reserves, leaves, branches and trunks. Biomass growth in the trunk is converted into volume growth from the density of the wood at the end of the year. This makes it possible to estimate growth in ring width and volume on an annual basis.
The annual output data were the volume of wood, the mortality rate, and the carbon sequestrated into the forest stand. Risk of mortality by carbon starvation and hydraulic failure was assessed according to Davi and Cailleret (2017). For this purpose, we simulated non-structural carbohydrates ([NSC]) and midday leaf water potential. Hydraulic failure is computed when the midday leaf water potential drops below the P50 of the species (leaf water potentials below which 50% of conductivity loss occurs). The threshold of mortality on [NSC] is estimated by fitting the threshold to minimize the difference among simulated and measured annual mortality rates between 2000 and 2015 once the hydraulics failure was computed. The mortality measurements were taken from the French National Inventory on Burgundy.
The CASTANEA model simulated the forest growth of a stand of 1 ha through different silvicultural paths starting from a 125-year-old beech forest in Burgundy from 2000 to 2100. The silvicultural paths arise from the CRPF (Regional Center for Privately-Owned Forests) of Burgundy for both species. Table I.2 presents the different characteristics of each silvicultural path.
The seven silvicultural paths were simulated through three different AWC (50, 100 and 150 mm) that characterized the drought effect and two different IPCC scenarios (RCP 4.5 and RCP 8.5) that characterized the climate effect.
Figure I.1 illustrates, for one given IPCC scenario, the structure of the applied methodology from the simulation of forest growth to economic results. The resulting volume of wood for each scenario (outputs of the CASTANEA model) was the input of the economic approach.
Our objective was to compare the 32 LEVs among scenarios. All the comparisons of LEV are detailed according to Figure I.1 as follows (taking only one IPCC scenario into account):
• (LEV 1 with LEV 3) and (LEV 1 with LEV 7): effect of drought.
• (LEV 3 with LEV 4) and (LEV 7 with LEV 8): effect of density/rotation reduction strategy.
• (LEV 1 with LEV 2) and (LEV 3 with LEV 5) and (LEV 7 with LEV 9): effect of species substitution strategy.
• (LEV 3 with LEV 6) and (LEV 7 with LEV 10): effect of species substitution strategy combined with that of density/rotation reduction.
First, the sum of an infinite number of rotations made it possible to calculate the land expectation value, commonly referred to as the Faustmann criterion in forest economics (Faustmann, 1849), as follows:
∞ −1 − (1)
( ) = ∑ ∑
(1 + )( . + )
where B is the benefits, C the costs, r the discount rate, n the stand age, N the rotation length and i the rotation number.
It is assumed here that the forest owner has a single objective: to maximize LEV. The infinite horizon used by this criterion makes it possible to compare management options associated with different temporal horizons, assuming that the silvicultural path was identical for each subsequent rotation after the first one. In other words, each silvicultural operation (thinning, maintenance, harvest) was implemented at the same age and for the same cost or benefit an infinite number of times. This may be seen as a limit of this criterion. However, other existing ones present greater limitations and are rarely adopted (Fraysse et al., 1990; Morel and Terreaux, 1995). Faustmann’s LEV takes the costs and the benefits from wood harvesting into account. After discussion with forestry experts, a discount rate r of 3% was chosen. A sensitivity analysis on this parameter was performed and is presented in Section 4.3.
We also asked ourselves if the consideration of forest ecosystem services may impact the economic results. In the context of mitigation of climate change, we chose to consider carbon sequestration in particular. In fact, carbon loss is rarely considered in the literature in addition to economic loss (see Yousefpour and Hanewinkel (2014) for an exception).
For that purpose, we also calculated Hartman’s LEV, which makes it possible to consider the benefits from wood harvesting and from amenities simultaneously (Hartman, 1976), in our case, carbon sequestration5.
The discount rate r was also 3% for beech and Douglas-fir in order to be able to compare LEVs. To compute the benefits from carbon sequestration, we considered the additional sequestration of the standing wood and we chose the social cost of carbon of 44 EUR/T (Watkiss and Downing, 2008). The social cost of carbon is “an estimate of the total cost of damages generated by each ton of CO2 that is spewed into the air” (Howard and Sterner, 2014). It therefore gives the total value of avoided damage caused by the flow of carbon to the atmosphere in the case of potential total deforestation.
Table of contents :
Context and motivation
Research questions and objectives of the thesis
Description of the chapters
Chapter I: An economic comparison of adaptation strategies towards a drought-induced risk of forest dieback
2. Material and methods
2.1. Some definitions
2.1.1. Characterization of drought and risk
2.1.2. Adaptation strategies
2.2. Case study
2.2.1. Case study area: Burgundy region
2.2.2. Species of interest
2.2.3. Study scenarios
2.3.1. Simulation of forest growth and silvicultural treatments
2.3.2. Economic approach
3.1. Forest growth and mortality
3.2. Economic comparison
3.3. Carbon sequestration
4.1. Adaptation from an economic perspective
4.2. Carbon consideration
4.3. Sensitivity analysis
4.4. Limits and perspectives
A. Silvicultural operations with associated net benefits from wood production and carbon sequestration for each scenario
B. Supplementary data
Chapter II: Is diversification a good option to reduce drought-induced risk of forest dieback? An economic approach focused on carbon accounting
2. Material and methods
2.1. Study area: Grand-Est region and species of interest
2.2.1. Scenarios tested
2.2.2. Forest growth simulation
2.2.3. Economic analysis
18.104.22.168. Double-weighted land expectation value
22.214.171.124. Carbon price scenarios
3.1. Effect of drought recurrence on optimal rotation length, tree mortality, carbon sequestration, and LEV
3.2. Effect of diversification and combined diversification on optimal rotation length, tree mortality, carbon sequestration and LEV
3.3. Effect of carbon price on optimal rotation length and LEV
4.1. Diversification is a good adaptation option to reduce drought-induced risk of forest dieback from an economic perspective
4.2. Diversification and combining both diversification strategies lead to synergies
4.3. Financial balance vs. carbon balance
4.4. Valorising carbon decreases the optimal rotation length and increases LEV
4.5. Limits and perspectives of the study
A. Drought recurrences definition
B. Creation of fictive stands
C. Simulation of forest management
D. MATHILDE and CAT
E. Land expectation value and sensitivity analysis of discount rate
F. Synergy analysis of adaptation strategies
Chapter III: Composition diversification vs. structure diversification: How to conciliate timber production and carbon sequestration objectives under drought and windstorm
2. Material and methods
2.1. Study area: Drought and windstorm in the Grand-Est region and species of interest
2.2.1. Scenarios tested
2.2.2. Forest growth simulation and economic analysis
3.1. Effect of drought and/or windstorm recurrence on timber volume, carbon sequestration, tree mortality, and LEV
3.2. Effect of adaptation strategies on timber volume, carbon sequestration, tree mortality, and LEV
3.3. Effect of carbon price and discount rate on LEV
4.1. Diversification can be an economically effective adaptation strategy to reduce drought- and windstorm-induced risks
4.2. Considering both risks impacts the results and recommendations compared to investigating each risk separately
4.3. Diversifying the stand as well as combining both strategies lead to synergies
4.4. Valorising carbon increases forest value
A. Windstorm frequencies computation
B. MATHILDE and CAT
C. Sensitivity analysis of the discount rate on LEV
D. Synergy analysis of adaptation strategies
Chapter IV: Index insurance for coping with drought-induced risk of production losses in French forests
2. Literature review
3. Material and methods
3.2. Insurance policy design
3.2.1. Indemnity schedule
3.2.2. Tested indices
3.2.3. Optimisation of insurance contract
5. Discussion and perspectives
5.1. Optimal insurance contracts generate low gain, high compensation and a high basis risk
5.2. Including a regional differentiation on the species-specific insurance contract can improve the results
5.3. Other perspectives of the study
A. Optimal insurance contract and effectiveness criteria of the insurance contract (relative risk aversion coefficient of 1)
B. Optimal insurance contract and effectiveness criteria of the insurance contract (relative risk aversion coefficient of 0.5)
C. Optimal insurance contract and effectiveness criteria of the insurance contract (relative risk aversion coefficient of 2)
Summary of the main results
Public policies issues
French summary of the thesis
Contexte et motivation
Description des chapitres
Principaux résultats et conclusion