Sensitivity of Spruce/Moss Boreal Forest Net Ecosystem Productivity to Seasonal Anomalies in Weather

Abstract. A process-oriented, daily time step model of a spruce/moss boreal ecosystem simulated 1994 and 1995 productivity for a Boreal Ecosystem-Atmosphere Study site near Thompson, Manitoba. Simulated black spruce net primary productivity (NPP) was 139 g C m(exp -2) in 1994 and 112 in 1995; feathermoss NPP was 13.0 g C m(exp -2) in 1994 and 9.7 in 1995; decomposition was 126 g C m(exp -2) in 1994 and 130 in 1995; net ecosystem productivity (NEP) was an uptake of 26.3 g C m(exp -2)in 1994 and 2.5 in 1995. A very dry period for the first half of the 1995 summer was the major cause of that year's lower productivity. Sensitivity simulations explored the impact of 2-month long warmer, cooler, wetter, and drier spells on ecosystem productivity. Warmer summers decreased spruce NPP, moss NPP, and NEP; cooler summers had the opposite effect. Earlier snowmelt (due to either warmer spring temperatures or reduced winter precipitation) increased moss and spruce NPP; later snowmelt had the opposite effect. The largest effect on decomposition was a 5% reduction due to a drier summer. One-month droughts (April through October) were also imposed on 1975 base year weather. Early summer droughts reduced moss annual NPP by -30-40%; summer droughts reduced spruce annual NPP by 10%; late summer droughts increased moss NPP by about 20% due to reduced respiration; May to September monthly droughts reduced heterotrophic respiration by about 10%. Variability in NEP was up to roughly +/- 35%. Finally, 1975 growing season precipitation was redistributed into frequent, small rainstorms and infrequent, large rainstorms. These changes had no effect on spruce NPP. Frequent rainstorms increased decomposition by a few percent, moss NPP by 50%, and NEP by 20%. Infrequent rainstorms decreased decomposition by 5%, moss NPP by 50% and NEP by 15%. The impact of anomalous weather patterns on productivity of this ecosystem depended on their timing during the year. Multiyear data sets are necessary to understand this behavior and test these types of models.