Predicting the response of the Amazon rainforest to persistent drought conditions under current and future climates: a major challenge for global land surface models

Émilie Joetzjer, Christine Delire, Hervé Douville, Philippe Ciais, Bertrand Decharme, Rosie A. Fisher, Bradley Christoffersen, Jean‐Christophe Calvet, Antonio Costa, Leandro Valle Ferreira, Patrick Meir2014 · DOI

TL;DR

This paper investigates how well a land surface model (ISBACC) simulates the Amazon rainforest's response to drought, finding that while it captures soil moisture reasonably well, it struggles with vegetation responses like gas exchange and biomass loss. The study highlights the critical need for better parameterizations of water stress and the inclusion of drought-induced mortality for accurate climate change projections.

What Problem Does This Paper Address?

Global climate models consistently project drier and longer dry seasons for the Amazon rainforest, raising concerns about its future stability and role in the global climate system. However, significant uncertainties remain regarding how this vital ecosystem will respond to persistent drought, especially under future climate conditions with elevated CO2 and temperatures. Accurately modeling these responses is crucial for predicting carbon cycle feedbacks and understanding potential tipping points, such as forest dieback. This paper addresses the challenge of evaluating and improving the ability of a specific land surface model (ISBACC) to simulate these drought impacts, focusing on both water and carbon budgets.

Methods

The study employs the ISBACC (ISBA Carbon Cycle) land surface model, which simulates exchanges of water, energy, and carbon between the land surface and atmosphere. A key aspect of the methodology involves evaluating different "Water Stress Functions" (WSFs) within the model. These WSFs are empirical formulations that describe how soil moisture deficit affects plant physiological processes like stomatal conductance. The researchers tested four WSFs: two physiologically-based "drought-avoiding" and "drought-tolerant" strategies, and two simpler "linear" and "SiB3" formulations. They also conducted simulations with idealized future climate conditions, including increased CO2 concentrations and temperatures, to assess the model's sensitivity.

Datasets

The study utilizes data from two Throughfall Exclusion (TFE) experiments conducted in the eastern Amazon (Tapajós and Caxiuanã). These experiments involved artificially reducing rainfall to simulate drought conditions. The researchers compared the model's outputs to observed data on:

  • Soil Water Content (SWC)
  • Stomatal Conductance ($g_s$)
  • Sap Flow (SF)
  • Evapotranspiration (ET)
  • Gross Primary Production (GPP)
  • Ecosystem Respiration ($R_e$)
  • Autotrophic Respiration ($R_a$)
  • Leaf Respiration ($R_l$)
  • Wood Respiration ($R_w$)
  • Root Respiration ($R_r$)
  • Soil Respiration ($R_s$)
  • Leaf Area Index (LAI)
  • Above Ground Biomass (AGB)

The study also references data from other sources for model validation and comparison, including fluxtower measurements and outputs from other land surface models.

Key Insights and Ideas

  • WSF Choice Matters Dramatically: The choice of Water Stress Function (WSF) significantly impacts how ISBACC simulates the rainforest's response to drought. Some WSFs lead to a much stronger simulated response than others, highlighting the importance of accurate parameterization.
  • Model Struggles with Vegetation Dynamics: While ISBACC can reasonably simulate soil moisture, it fails to accurately capture key vegetation responses to drought, such as the timing and amplitude of gas exchange reductions and significant biomass loss.
  • Drought-Induced Mortality is Crucial: The model's inability to simulate tree mortality, a key observed impact of drought, is a major limitation. This omission means the model likely underestimates the long-term consequences of severe drought events.
  • CO2 Fertilization Mitigates Drought Sensitivity: Increased CO2 concentrations, through enhanced Water Use Efficiency (WUE), can partially offset the negative impacts of drought in the model, making the vegetation less sensitive to water stress. This suggests a complex interplay between rising CO2 and drought.
  • Generic Drought Response is Challenging: Defining a single, generic drought response for the diverse Amazonian forest is inherently difficult due to the functional diversity of its species.

Key Results

  • The "linear" WSF provided the best overall simulation of vegetation response among the tested functions, though still imperfect.
  • ISBACC tended to simulate soil moisture that was too wet during the dry season, with correlations around 0.65 at Tapajós.
  • The model underestimated observed losses of Leaf Area Index (LAI) and Above Ground Biomass (AGB) during drought experiments, particularly for AGB where it often failed to capture any loss.
  • Simulations with elevated CO2 (+4°C and 4xCO2) showed a significant dampening of drought sensitivity, with SWI remaining high even under simulated TFE.
  • While ISBACC reasonably simulated annual ecosystem respiration fluxes, it overestimated woody tissue respiration and underestimated leaf and root respiration.

Limitations and Future Work

The authors acknowledge several limitations:

  • WSF Parameterization: The WSFs were originally derived from measurements on temperate species and not specifically calibrated for mature tropical rainforest trees, leading to potential inaccuracies.
  • Missing Mortality: The model lacks a deterministic relationship between climate and tree mortality, a critical process for understanding drought impacts.
  • Limited Observational Data: While TFE experiments provide valuable data, they are plot-scale and may not fully represent basin-wide responses. There's a need for more comprehensive data, especially on mortality and physiological processes under drought.
  • Model Complexity: Implementing more mechanistic models of drought stress requires significantly higher model complexity and more detailed data for specification and validation.

Future work should focus on:

  • Developing more accurate and site-specific WSF parameterizations for tropical forests.
  • Incorporating explicit drought-induced tree mortality processes into land surface models.
  • Improving the mechanistic understanding of forest responses to drought under warmer climates and higher CO2.
  • Combining data from TFE experiments with other datasets (e.g., long-term forest inventories, trait-based mortality models) for more robust model evaluation and development.

Summary updated

Abstract

While a majority of Global Climate Models project dryer and longer dry seasons over the Amazon under higher CO2 levels, large uncertainties surround the response of vegetation to persistent droughts in both present-day and future climates. We propose a detailed evaluation of the ability of the ISBACC Land Surface Model to capture drought effects on both water and carbon budgets, comparing fluxes and stocks at two recent ThroughFall Exclusion (TFE) experiments performed in the Amazon. We also explore the model sensitivity to different Water Stress Function (WSF) and to an idealized increase in CO2 concentration and/or temperature. In spite of a reasonable soil moisture simulation, ISBACC struggles to correctly simulate the vegetation response to TFE whose amplitude and timing is highly sensitive to the WSF. Under higher CO2 concentration, the increased Water Use Efficiency (WUE) mitigates the ISBACC's sensitivity to drought. While one of the proposed WSF formulation improves the response of most ISBACC fluxes, except respiration, a parameterization of drought-induced tree mortality is missing for an accurate estimate of the vegetation response. Also, a better mechanistic understanding of the forest responses to drought under a warmer climate and higher CO2 concentration is clearly needed.
BibTeX
@misc{joetzjer2014predicting,
  author = {Émilie Joetzjer and Christine Delire and Hervé Douville and Philippe Ciais and Bertrand Decharme and Rosie A. Fisher and Bradley Christoffersen and Jean‐Christophe Calvet and Antonio Costa and Leandro Valle Ferreira and Patrick Meir},
  title = {{Predicting the response of the Amazon rainforest to persistent drought conditions under current and future climates: a major challenge for global land surface models}},
  year = {2014},
  doi = {10.5194/gmdd-7-5295-2014},
}
Figure