Land surface model parameter optimisation using in-situ flux data: comparison of gradient-based versus random search algorithms

Abstract

Abstract. Land surface models (LSMs), used within earth system models, rely on numerous processes for describing carbon, water and energy budgets, often associated with highly uncertain parameters. Data assimilation (DA) is a useful approach for optimising the most critical parameters in order to improve model accuracy and refine future climate predictions. In this study, we compare two different DA methods for optimising the parameters of seven plant functional types (PFTs) of the 15 ORCHIDEE land surface model using daily averaged eddy-covariance observations of net ecosystem exchange and latent heat flux at 78 sites across the globe. We perform a technical investigation of two classes of minimisation methods -local gradient-based (the L-BFGS-B algorithm) and global random search (the genetic algorithm) -by evaluating their relative performance in terms of the model-data fit and the difference in retrieved parameter values. We examine the performance of each method for two cases: when optimising parameters at each site independently (-single-site‖ approach) and when 20 simultaneously optimising the model at all sites for a given PFT using a common set of parameters (-multi-site‖ approach).We find that for the single site case the random search algorithm results in lower values of the cost function (i.e. lower model -data root mean square differences) than the gradient-based method; the difference between the two methods is smaller for the multi-site optimisation due to a smoothing of the cost function shape with a greater number of observations. The spread of the cost function, when performing the same tests with 16 random first guess parameters, is much larger with the gradient 25 based method, due to the higher likelihood of being trap in local minima. When using pseudo-observations tests the genetic algorithm results in a closer approximation of the true posterior parameter value in the L-BFGS-B algorithm. We demonstrate the advantages and challenges of different DA techniques and provide some advice on using it for the LSM parameters optimisation.Geosci. Model Dev. Discuss., https://doi

Global net biome CO2exchange predicted comparably well using parameter–environment relationships and plant functional types

“…3.4). It also includes arguably more degrees of freedom than what may be observed in a typical TBM; that is, the total number of "representative pixels" used in the aggregation includes a relatively broad sample of locations within each PFT, although the total number of PFTs considered here (n = 9) aligns reasonably well with current approaches (Bastrikov et al, 2018;Harper et al, 2018;Reick et al, 2021).…”

“…The GlobCover product, available at 300m spatial resolution, provides a discrete classification of each land surface pixel into one of 23 land cover classes, or PFTs (Arino et al, 2012). To more closely align with the level of detail in many current TBMs (Bastrikov et al, 2018;Harper et al, 2018;Reick et al, 2021), we reduced these 23 classes to 9 broad groupings (Table S2). We first determined each 4°×5° pixel's fractional PFT composition by summing the (aggregated) GlobCover classifications contained within it.…”

Towards a more detailed representation of high-latitude vegetation in the global land surface model ORCHIDEE (ORC-HL-VEGv1.0)

“…For details of the implementation see Santaren et al (2014). Note that this algorithm is more efficient to find the minimum of J than a gradient-based method as discussed in Bastrikov et al (2018).…”

“…Similarly, a first description of lichen and bryophytes was implemented in the JSBACH model (Porada et al, 2013), improved recently with a process-based implementation (Porada et al, 2016). Biogeochemical and biophysical characteristics of shrubs are already implemented in some models, such as in the Community Land Model (Oleson et al, 2013), JULES (Clark et al, 2011) and JSBACH (Baudena et al, 2015). In this study we further develop the ORCHIDEE model , the land surface component of the Institute Pierre Simon Laplace (IPSL) ESM, to represent non-vascular plants, Arctic shrubs and tundra grasses.…”