Bayesian optimization of a light use efficiency model for the estimation of daily gross primary productivity in a range of Italian forest ecosystems

Abstract In this study we applied a modified version of Prelued, a simple semi-empirical light use efficiency (LUE) model, to eight eddy-covariance Italian sites. Since this model has been successfully applied mainly to coniferous forests located at northern latitudes, in our study we aimed to test its generality, by comparing Prelued's outputs in coniferous, broadleaf forests and in a Mediterranean macchia, at different climatic and environmental conditions. The model was calibrated for daily gross primary production (GPP) observed over one year in each flux site and validated for another year. The model uncertainties on both GPP and model parameters were estimated, applying a Bayesian calibration based on a multiple chains Markov Chain Monte Carlo sampling. The accuracy of the model estimates of daily GPP over the entire period of simulation differed widely depending on the site considered, with generally good model performance when applied to evergreen and broadleaf forests and poor performances in the Mediterranean macchia. The values of the modifiers accounting for the response to climatic variables suggested the soil water content to be non-limiting in temperate mountain evergreen but limiting in Mediterranean forests. Model uncertainties were always smaller than data uncertainties, with variable magnitude depending on the site considered. Both modeled GPP and uncertainties were largely dependent also on uncertainties on the data, which made their calculation a key process in this modelling exercise. In conclusion, this semi-empirical model appears to be suitable for estimating daily and annual forest GPP in most of the considered sites, with the exception of Mediterranean macchias, and for supporting its application to a large range of ecosystems provided a site-specific calibration. The Bayesian calibration did not confer a clear advantage in terms of model performances in respect to other methods used in previous studies, but allowed us to estimate uncertainties on both parameter values and model estimates, which were useful to analyse more in detail the ecosystem response to environmental drivers of GPP.

[1]  Steffen Fritz,et al.  Improved light and temperature responses for light-use-efficiency-based GPP models , 2013 .

[2]  Damian Barrett,et al.  Conversion of canopy intercepted radiation to photosynthate: review of modelling approaches for regional scales. , 2003, Functional plant biology : FPB.

[3]  Atul K. Jain,et al.  A model-data comparison of gross primary productivity: Results from the North American Carbon Program site synthesis , 2012 .

[4]  Peter E. Thornton,et al.  A diagnostic carbon flux model to monitor the effects of disturbance and interannual variation in climate on regional NEP , 2006 .

[5]  S. T. Gower,et al.  A cross‐biome comparison of daily light use efficiency for gross primary production , 2003 .

[6]  Andrew P. Robinson,et al.  Model validation using equivalence tests , 2004 .

[7]  S. T. Gower,et al.  Direct and Indirect Estimation of Leaf Area Index, fAPAR, and Net Primary Production of Terrestrial Ecosystems , 1999 .

[8]  J. Monteith.  Climate and the efficiency of crop production in Britain , 1977 .

[9]  Henry L. Gholz,et al.  Climatic factors controlling the productivity of pine stands: a model-based analysis , 1994 .

[10]  Juan A. Blanco,et al.  Testing the performance of a forest ecosystem model (FORECAST) against 29 years of field data in a Pseudotsuga menziesii plantation , 2007 .

[11]  I. C. Prentice,et al.  An integrated biosphere model of land surface processes , 1996 .

[12]  A. Mäkelä,et al.  Acclimation of photosynthetic capacity in Scots pine to the annual cycle of temperature. , 2004, Tree physiology.

[13]  T. A. Black,et al.  Optimization of ecosystem model parameters through assimilating eddy covariance flux data with an ensemble Kalman filter , 2008 .

[14]  Ron Smith,et al.  Bayesian calibration of process-based forest models: bridging the gap between models and data. , 2005, Tree physiology.

[15]  Juan A. Blanco,et al.  Complexity in modelling forest ecosystems: How much is enough?☆ , 2008 .

[16]  Igor Linkov,et al.  Untangling drivers of species distributions: Global sensitivity and uncertainty analyses of MaxEnt , 2014, Environ. Model. Softw..

[17]  Paola Annoni,et al.  Sixth International Conference on Sensitivity Analysis of Model Output How to avoid a perfunctory sensitivity analysis , 2010 .

[18]  P. Hari,et al.  Uncertainties in measurement and modelling of net ecosystem exchange of a forest , 2006 .

[19]  Shaun Quegan,et al.  Model–data synthesis in terrestrial carbon observation: methods, data requirements and data uncertainty specifications , 2005 .

[20]  L. S. Pereira,et al.  Crop evapotranspiration : guidelines for computing crop water requirements , 1998 .

[21]  K. Schulz,et al.  Identification of a general light use efficiency model for gross primary production , 2010 .

[22]  D. Hollinger,et al.  MODELING GROSS PRIMARY PRODUCTION OF AN EVERGREEN NEEDLELEAF FOREST USING MODIS AND CLIMATE DATA , 2005 .

[23]  Damiano Gianelle,et al.  A user-friendly forest model with a multiplicative mathematical structure: a Bayesian approach to calibration , 2014 .

[24]  Joe Landsberg,et al.  Modelling forest ecosystems: state of the art, challenges, and future directions , 2003 .

[25]  Christopher J. Still,et al.  Large‐scale plant light‐use efficiency inferred from the seasonal cycle of atmospheric CO2 , 2004 .

[26]  R. Waring,et al.  A generalised model of forest productivity using simplified concepts of radiation-use efficiency, carbon balance and partitioning , 1997 .

[27]  Pierre Hiernaux,et al.  Analysis of the in situ and MODIS albedo variability at multiple timescales in the Sahel , 2008 .

[28]  David Gustafsson,et al.  Simulated climate change impacts on fluxes of carbon in Norway spruce ecosystems along a climatic transect in Sweden , 2008 .

[29]  R. D'Arrigo,et al.  Monsoon drought over Java, Indonesia, during the past two centuries , 2006 .

[30]  Markus Reichstein,et al.  On the assignment of prior errors in Bayesian inversions of CO2 surface fluxes , 2006 .

[31]  Steven W. Running,et al.  Testing scale dependent assumptions in regional ecosystem simulations , 1994 .

[32]  D. Hollinger,et al.  Statistical modeling of ecosystem respiration using eddy covariance data: Maximum likelihood parameter estimation, and Monte Carlo simulation of model and parameter uncertainty, applied to three simple models , 2005 .

[33]  P. Hari,et al.  Effects of dormancy and environmental factors on timing of bud burst in Betula pendula. , 1998, Tree physiology.

[34]  E. Davidson,et al.  A distinct seasonal pattern of the ratio of soil respiration to total ecosystem respiration in a spruce‐dominated forest , 2006 .

[35]  W. Oechel,et al.  Parallel adjustments in vegetation greenness and ecosystem CO2 exchange in response to drought in a Southern California chaparral ecosystem , 2006 .

[36]  Alessandro Cescatti,et al.  Importance of advection in the atmospheric CO2 exchanges of an alpine forest , 2005 .

[37]  Eero Nikinmaa,et al.  Developing an empirical model of stand GPP with the LUE approach: analysis of eddy covariance data at five contrasting conifer sites in Europe , 2007 .

[38]  Mark A. Friedl,et al.  Using FLUXNET data to improve models of springtime vegetation activity onset in forest ecosystems , 2013 .

[39]  T. Vesala,et al.  Deriving a light use efficiency model from eddy covariance flux data for predicting daily gross primary production across biomes , 2007 .

[40]  T Vesala,et al.  Contributions of climate, leaf area index and leaf physiology to variation in gross primary production of six coniferous forests across Europe: a model-based analysis. , 2009, Tree physiology.

[41]  J. Yeluripati,et al.  A Bayesian framework for model calibration, comparison and analysis: Application to four models for the biogeochemistry of a Norway spruce forest , 2011 .

[42]  J. P. Kimmins,et al.  Modelling forest ecosystem net primary production : the hybrid simulation approach used in FORECAST , 1999 .

[43]  A. Mäkelä,et al.  Modelling five years of weather-driven variation of GPP in a boreal forest , 2006 .

[44]  A. Arneth,et al.  Nitrogen controls plant canopy light-use efficiency in temperate and boreal ecosystems , 2008 .

[45]  Christian Feigenwinter,et al.  A new mass conservation approach to the study of CO2 advection in an alpine forest , 2009 .

[46]  John Skilling,et al.  Data analysis : a Bayesian tutorial , 1996 .

[47]  Cajo J. F. ter Braak,et al.  A Markov Chain Monte Carlo version of the genetic algorithm Differential Evolution: easy Bayesian computing for real parameter spaces , 2006, Stat. Comput..

[48]  A. Mäkelä,et al.  Does canopy mean nitrogen concentration explain variation in canopy light use efficiency across 14 contrasting forest sites? , 2012, Tree physiology.

[49]  Oscar García,et al.  Evaluating forest Growth Models , 1997 .

[50]  D. Baldocchi.  ‘Breathing’ of the terrestrial biosphere: lessons learned from a global network of carbon dioxide flux measurement systems , 2008 .

[51]  T. Vesala,et al.  On the separation of net ecosystem exchange into assimilation and ecosystem respiration: review and improved algorithm , 2005 .

[52]  S. Wellek.  Testing Statistical Hypotheses of Equivalence , 2002 .

[53]  W. Rawls,et al.  Estimating generalized soil-water characteristics from texture , 1986 .

[54]  Georgios Xenakis,et al.  Sensitivity and uncertainty analysis from a coupled 3-PG and soil organic matter decomposition model , 2008 .