Observational Constraints on the Effect of Drought-Induced Allocation Shifts on Carbon Flux Recovery Time and Magnitude at Two Tropical Sites

Plants can respond to drought by changing how photosynthetically derived carbon is allocated to foliage, woody components, and roots. These dynamic shifts can affect future carbon fluxes through associated changes in leaf area, root water uptake ability, and indirectly, soil carbon pools. Allocation shifts have therefore been hypothesized as one of many mechanisms (pests, hydraulic failure, etc..) contributing to legacy effects of drought on carbon fluxes. However, the magnitude of both allocation shifts and their effects on carbon fluxes and pools are poorly constrained, and difficult to observe in situ. Here, we first demonstrate that drought-induced allocation shifts can be robustly inferred by assimilating observations of NBE and other data streams in a carbon cycle model-data fusion system. We then investigate how these allocation shifts change the duration and magnitude of drought's effects on NBE both during and after the events. To do so, we added a dynamic carbon allocation scheme within the CArbon DAta MOdel fraMework (CARDAMOM) data assimilation system, which robustly optimizes parameters and carbon cycle states based on multiple observational data streams of carbon fluxes and pools. We demonstrate that doing so leads to more accurate carbon cycle simulations even outside the assimilation window. We then test the impacts of dynamic allocation on carbon fluxes by constraining CARDAMOM parameters at a wet tropical (French Guiana) and a dry tropical (Cumberland Plains) flux tower site. To isolate the impacts of drought on carbon fluxes, we run simulations of CARDAMOM with drought meteorology replaced by its climatological average. Preliminary results at French Guiana find that accounting for dynamic allocation during drought leads to a 15±4% weaker simulated carbon sink when compared to static allocation. Furthermore, when isolating the impact of dynamic allocation from drought periods we show that dynamic allocations significantly increase NBE recovery time by up to 2.3 times when compared to static allocation.