Delayed Tree Dormancy Resulted in Higher Annual Savanna Gross Primary Productivity in the Northern Sonoran Desert

Drylands support over two billion people and savannas contribute about 25% to global gross primary productivity (GPP). Savannas are water limited ecosystems characterized by two dominant plant functional types: trees and an understory of grass and forbes. It is hard to predict how future changes in water across time and space will affect savanna GPP. This is due to, in part, to trees and understory species having different leaf phenology and rooting depths. By integrating three established models of the carbon-water function of drylands, we developed a general hypothesis that incorporates the phenology and root architecture of the two dominant plant functional types to explain savanna GPP controls. We test the hypothesis that productivity in savanna ecosystems is controlled by the temporal and vertical distribution of soil moisture and differences in leaf phenology growing season length of understory and tree functional types. We used phenology cameras, satellite observations from the MODIS sensors, and an in-situ tower equipped to measure carbon flux by eddy covariance to quantify tree and understory phenometrics, savanna phenometrics, and savanna GPP respectively. To quantify available water resources, we measured rainfall and soil moisture accumulation at two different depths (shallow, 30 cm). We found that the leaf phenology of trees and understory plants were distinct from each other and that each plant type was active during periods of different water availability. We found also that the effect of water on GPP varied by season, with GPP being driven most by soil moisture from the summer (R2 > 0.38, shallow & deep), then spring (R2 > 0.23, shallow and deep), and finally the winter (R2 = 0.19, only deep). While the length of the savanna growing season was not related to GPP, delayed leaf senescence tended to increase annual GPP. By comparing the response of leaf phenology in trees and understory to the phenology of the whole ecosystem estimated from satellites, we found that trees showed a similar leaf senescence response to the senescence response of the whole savanna; that the availability of deep soil moisture tended to delay leaf senescence (R2 = 0.96 for trees & R2 > 0.29 for savanna, respectively). In contrast, more shallow soil moisture tended to advance leaf senescence in the understory (R2 > 0.81). These major findings are consistent with our general hypothesis and highlight the important and sometimes compensating effects of rainfall during winter and monsoon seasons. It is likely that any future increase of heavy winter precipitation would assist tree productivity and overall carbon gain in these semi-arid savanna ecosystems