Land carbon models underestimate the severity and duration of drought’s impact on plant productivity

H.R. Kolus, D.N. Huntzinger, C.R. Schwalm, J.B. Fisher, N. McKay, Y. Fang, A.M. MichalakK. Schaefer, Y. Wei, B. Poulter, J. Mao, N.C. Parazoo and X. Shi


Terrestrial ecosystems' ability to sequester carbon in plants and soils helps offset the increase of carbon dioxide in the atmosphere, but with increased drought events questions about how these ecosystems respond remain. We assessed whether models accurately estimate how plant biomass production, or net primary productivity (NPP), responds to drought events. While models correctly showed a decrease in NPP, the initial response to drought is overestimated while legacy effects of drought are underestimated.  Misrepresentation of how plants respond to drought could lead to an overestimation of an ecosystem's ability to sequester carbon, which could lead to an underestimate of the forecasted amount of carbon dioxide in the atmosphere.


Figure: Schematic depicting how severe drought can impact NPP. This example ecosystem begins at typical vegetation growth rates (the height of the green surface). Drought (the orange shaded period) can decrease productivity even after the return to nominal climate conditions. Drought response is the initial depression in growth following the onset of a drought (depth of the trough in NPP). Lag time is the time between the onset of drought and the drought response. Recovery is the amount of time required to recover normal growth rates. Drought legacy effects correspond to the magnitude of NPP depression during drought recovery. The cut-out wedge illustrates the total loss of NPP, relative to typical productivity. Vegetation may be unable to recover from drought and, instead, follow the path to mortality. Note that the relative magnitudes of these variables depend on factors such as the drought’s characteristics and the type of vegetation; this illustration provides one example of drought’s impact on productivity. Illustration by Victor O. Leshyk. Printed with permission by Victor O. Leshyk under a CC BY open access license (


The ability to accurately predict ecosystem drought response and recovery is necessary to produce reliable forecasts of land carbon uptake and future climate. Using a suite of models from the Multi-scale Synthesis and Terrestrial Model Intercomparison Project (MsTMIP), we assessed modeled net primary productivity (NPP) response to, and recovery from, drought events against a benchmark derived from tree ring observations between 1948 and 2008 across forested regions of the US and Europe. We find short lag times (0–6 months) between climate anomalies and modeled NPP response. Although models accurately simulate the direction of drought legacy effects (i.e. NPP decreases), projected effects are approximately four times shorter and four times weaker than observations suggest. This discrepancy between observed and simulated vegetation recovery from drought reveals a potential critical model deficiency. Since productivity is a crucial component of the land carbon balance, models that underestimate drought recovery time could overestimate predictions of future land carbon sink strength and, consequently, underestimate forecasts of atmospheric CO2.

Kolus, H.R., D.N. Huntzinger, C.R. Schwalm, J.B. Fisher, N. McKay, Y. Fang, A.M. Michalak, K. Schaefer, Y. Wei, B. Poulter, J. Mao, N.C. Parazoo, X. Shi (2019) "Land carbon models underestimate the severity and duration of drought’s impact on plant productivity," Scientific Reports, 9 (2758), doi:10.1038/s41598-019-39373-1.