Seasonal patterns in forest-grassland transition zones predicted by the Community Land Model
Evaluations of plant patterns in global ecosystem models have traditionally focused on large carbon stocks, like closed-canopy forests, or areas of large albedo feedbacks, like the arctic. However, forest-grassland transition zones (savannas, woodlands, wooded grasslands, shrublands) are highly sensitive to climate, and may already be changing due to warming, changes in precipitation patterns, and/or CO2 fertilization. Shifts between closed canopy forest and open grassland, as well as shifts in phenology, could have large impacts on the global carbon cycle, water balance, albedo, and on the humans and other animals that depend on these regions.
Here we compare 23 years of monthly leaf area index (LAI) outputs from several offline versions of the Community Land Model (CLM) to the normalized difference vegetation index (NDVI) as measured by the AVHRR satellite from 1982 to 2004. Specifically, we focus on seasonal patterns in regions defined as ‘woodland’, ‘wooded grassland’, ‘closed shrubland’, ‘open shrubland’, and ‘grassland’ by the University of Maryland’s Land Cover Classification separated by continent. We consider and compare three versions of CLM (v. 4, v. 4.5, and CLM-ED) in several different configurations (e.g. with and without fire). We note that while NDVI and LAI do not scale linearly across their ranges, they have been shown to have a linear relationship at low (< 4) LAI values, which are being considered here, and so patterns can be compared directly between these two measures.
We found that the CLM4CN was able to capture phenological patterns well in all woodland regions except for Australia. Seasonal patterns of LAI corresponded to NDVI patterns with R2 values greater than 0.5 in North America, South America, and Europe. Phenology was also well represented in temperate northern hemisphere grasslands (R2 > 0.6), and reasonably well represented in African wooded grasslands (R2 = 0.45). Phenology was very poorly captured in open and closed shrublands across all continents, and wooded grassland patterns were not well represented anywhere except for Africa. Overall we are not surprised that CLM performs best in the most highly studied, highly instrumented parts of the globe, but we emphasize that in order to capture global vegetation patterns more focus needs to be placed on understanding processes in the perhaps less charismatic vegetation classes, like shrublands and savanna-type systems.