Seasonal and interannual variability in carbon exchange along a boreal black spruce chronosequence: Tower-based observations and MODIS-driven predictions using the Vegetation Photosynthesis and Respiration Model

Boreal forests represent more than half of North American forests, and black spruce is the most common species. Forest fires are the dominant disturbance in boreal forests, resulting in a heterogeneous distribution of different stand ages across the boreal region. This mosaic of stand ages is a dominant factor regulating boreal forest carbon exchange, and regional estimates of carbon balance must take this age structure into account. We conducted two analyses to investigate this problem. First, we analyzed carbon dioxide fluxes along a chronosequence of different stand ages in north-central Manitoba, Canada. Then, we utilized the MODIS-driven Vegetation Photosynthesis and Respiration Model (VPRM) to predict ecosystem respiration (R) and gross ecosystem exchange (GEE) for the same chronosequence. We then validated the VPRM-predicted fluxes with the observations in order to evaluate its suitability for assessing regional-scale carbon balances in the boreal forest.

Tower-based observations

We studied carbon dioxide flux data from six towers consisting of forest stands 8, 17, 25, 42, 76, and 156 years in age, encompassing the successional stages along a trajectory towards a closed-canopy mature black spruce forest. We separated the observed net flux at each site into its component processes of R and GEE. We found significant seasonal and interannual variability in carbon exchange between the different-aged stands. Maximum photosynthesis rates were highest at the 25-year-old stand (up to 12 µmol m-2 s-1), and were similar at all other stands (6-8 µmol m-2 s-1). Respiration rates were highest in the mid-successional sites (25 and 42-year old), suggesting that the decomposition of coarse woody debris in boreal stands is a slow process that takes place over time scales of decades.

The onset of the growing season was up to a month earlier at the stands dominated by conifers (25, 42, 76, and 156-year-old) than at the younger, deciduous-dominated stands (8 and 17-year-old). The conifer sites exhibited more interannual variability in the onset of growing season: photosynthesis began in late April in the warm spring of 2003, but was delayed until mid-May during the cold springs of 2002 and 2004. The younger deciduous-dominated sites exhibited less variability: the onset of the growing season at the 8-year-old site varied by only four days between 2003 and 2004 (2002 data was not available), with photosynthesis beginning in late May in both years despite very different weather.

Model parameterization and validation

The VPRM is an assimilation scheme that uses the MODIS-based Enhanced Vegetation Index (EVI) and Land Surface Water Index (LSWI), and ground-based eddy flux, sunlight, and air temperature measurements to calibrate two parameters for NEE estimates across different vegetation classes. After the parameters are determined, NEE is predicted solely as a function of these biome-specific parameters, EVI, LSWI, and air temperature (locally-measured, if available, or obtained from reanalysis products). Because the driving variables for the model are available across regional scales, the VPRM has the potential to resolve carbon balances across wide regions, constrained by the ground observations used in the calibration phase.

We calibrated the VPRM for NEE observed at the Northern Old Black Spruce site, located in the same 156-year-old stand as the oldest site in the chronosequence. The parameters determined for NOBS were then used to estimate carbon exchange along the six sites of the chronosequence using EVI and LSWI determined from MODIS and air temperature measured at the sites.

The VPRM accurately predicted the start of the growing season at the 25, 42, 76, and 156-year old sites. It also captured the significant interannual variability in the onset of growing season observed in the years of 2002-2004, correctly identifying the start of the growing season within a few days of the actual observations. This suggests that the GEE representation in the VPRM is successfully identifying the critical threshold in temperature and the seasonal transitions in LSWI and EVI that signal the onset of photosynthesis in coniferous sites. However, the VPRM overpredicted early-season photosynthesis at the youngest sites (8 and 17-year old). These younger sites have a significant deciduous component, resulting in a delayed onset of growing season that the VPRM only partially captured. The VPRM was most successful in predicting overall photosynthesis rates at the 156-year-old site, which is not surprising considering its location proximate to the NOBS site that was used for model calibration. The VPRM significantly underpredicted mid-summer photosynthesis in the youngest site, as the deciduous component of the canopy was not well-captured.

The VPRM represents ecosystem respiration as a linear function of a calibration-derived parameter and air temperature. Therefore, predicted respiration peaked with the mid-July peak in air temperature. In contrast, the observed seasonal course of ecosystem respiration in the boreal forest is asymmetric, peaking in August. This lag in respiration relative to photosynthesis is driven by the slow thawing process in boreal soils, which inhibits heterotrophic respiration of organic matter in the early- and mid-summer. This mismatch in the seasonal respiration curve resulted in discrepancies between the observations and the VPRM predictions. The VPRM significantly overpredicted respiration during the first part of the growing season, but more closely matched observations later in the summer, when the soil was fully thawed.

Assessing regional carbon balances is a difficult challenge. This is especially true in boreal regions, where stand age is driven by forest fires and is the dominant factor regulating carbon exchange. While tower-based measurements provide a wealth of information about carbon exchange on multiple time scales, they are necessarily limited in scope by their fixed nature. Chronosequences, with their innovative space-for-time approach, help assess the influence of stand age on carbon exchange in a heterogeneous environment. For larger-scale carbon balances, however, a top-down approach becomes necessary. The VPRM is unique in its ability to predict hourly carbon exchange solely on the basis of MODIS products and temperature data. It was successful in predicting the timing and seasonal course of carbon exchange in the boreal forest, though its temperature-driven respiration module misses some of the subtleties present in the observations.