The Terrestrial Biosphere Model Farm

Abstract

Model Intercomparison Projects (MIPs) are fundamental to our understanding of how the land surface responds to changes in climate. However, MIPs are challenging to conduct, requiring the organization of multiple, decentralized modeling teams throughout the world running common protocols. We explored centralizing these models on a single supercomputing system. We ran nine offline terrestrial biosphere models through the Terrestrial Biosphere Model Farm: CABLE, CENTURY, HyLand, ISAM, JULES, LPJ-GUESS, ORCHIDEE, SiB-3, and SiB-CASA. All models were wrapped in a software framework driven with common forcing data, spin-up, and run protocols specified by the Multi-scale Synthesis and Terrestrial Model Intercomparison Project (MsTMIP) for years 1901-2100. We ran more than a dozen model experiments. We identify three major benefits and three major challenges. The benefits include: (a) processing multiple models through a MIP is relatively straightforward, (b) MIP protocols are run consistently across models, which may reduce some model output variability, and (c) unique multimodel experiments can provide novel output for analysis. The challenges are: (a) technological demand is large, particularly for data and output storage and transfer; (b) model versions lag those from the core model development teams; and (c) there is still a need for intellectual input from the core model development teams for insight into model results. A merger with the open-source, cloud-based Predictive Ecosystem Analyzer (PEcAn) ecoinformatics system may be a path forward to overcoming these challenges.

Keywords: Earth System Model; PEcAn; ecoinformatic; ecosystem model; land surface model; model intercomparison project; terrestrial biosphere model; vegetation model.

Figure 1

Most global terrestrial biosphere models (TBMs) are developed primarily in the US, Europe, Australia, and Japan. This cartogram inflates/deflates the size of the country based on the number of TBMs developed, n = 58. From Fisher et al. (2014).

Figure 2

Terrestrial Biosphere Model Farm software flowchart consists of five stages. Stage 1: forcing data (e.g., from MsTMIP) are partitioned and distributed among supercomputer nodes. Stage 2: forcing data are converted to individual model requirements (format, temporal resolution, and units). Stage 3: models are spun up and run. Stage 4: model output is converted again (format, temporal resolution, and units) for standardization across models. Stage 5: model output is gathered off supercomputing nodes and recombined into global (or otherwise) grids.

Figure 3

An automated diagnostic visualization PDF is generated from the Terrestrial Biosphere Model Farm. (top‐left) Map of average variable value over simulation time span; (top‐middle) comparison map of average variable value; (top‐right) difference map between 1 and 2; (bottom‐left) annual time series along with comparison models; (bottom‐middle) average monthly values with comparison models; and (bottom‐right) average zonal/latitudinal values with comparison models. Here, we show Net Primary Productivity (NPP) with CENTURY as the highlighted model of interest for the MsTMIP Phase II simulation (2011–2100) with CESM1‐CAM5 climate and RCP4.5 CO₂ scenario. Comparison models are from MsTMIP Phase I SG3 simulation. Four additional models from the Farm are shown for 1901–2010 (CABLE, CENTURY, HyLand, and SiB‐3) with the latter three also shown for the Phase II simulation.

Figure 4

Inter‐simulation visual diagnostics for MsTMIP Phase I experiments show how a single model changes output among different experiments. Here, we show output (e.g., Net Primary Productivity, NPP, for HyLand) from four experiments: (1) spin‐up equilibrium conditions (RG1); (2) varying climate (SG1); (3) varying climate and land use/land cover change (LCLUC) (SG2); and (4) varying climate, LCLUC, and CO₂ (SG3). The top set of lines (primary y‐axis) is the annual average from 1901 to 2010. The bottom set of lines (secondary y‐axis) is the latitudinal/zonal sum from 55°S to 85°N.

Figure 5

Inter‐simulation visual diagnostics for MsTMIP Phase II experiments show how a single model changes output among five different climate projections and two Representative Concentration Pathways (RCPs). The climate projections are from five Earth System Models: (1) CESM1‐CAM5, (2) GFDL‐CM3, (3) HadGEM2, (4) IPSL‐CM5A‐MR, and (5) MPI‐ESM‐MR. The two RCPs are: (1) business‐as‐usual (RCP8.5) and (2) medium mitigation (RCP4.5). Output is for Net Primary Productivity (NPP) for the HyLand model. The top set of lines (primary y‐axis) is the annual average from 1901 to 2100 (1901–2010 is from the retrospective SG3 scenario, black line). The bottom set of lines (secondary y‐axis) is the latitudinal/zonal sum from 55°S to 85°N.

Figure 6

The terrestrial Earth as represented by the top‐performing terrestrial biosphere model in MsTMIP based on a suite of benchmarks. From Schwalm et al. (2015).