Dryland play a major role in the global carbon cycle. The US Southwest is experiencing fewer, larger precipitation events and longer dry intervals between rainfalls. These longer droughts are likely driving physiological, phenological, morphological, and community-level responses of dryland vegetation with unknown feedbacks to atmospheric CO2. It remains unclear how seasonal drought intensity and duration affect the magnitude, duration, and direction of dryland vegetation carbon cycling and atmospheric feedbacks. To address this question, we integrated the measurements of soil hydrology, plant community, and carbon fluxes from a new rainfall manipulation experiment site (RainManSR) in the Santa Rita Experimental Range of Southeast Arizona, US into the Community Land Model (CLM5). This field experiment imposed four precipitation treatments (S1–S4), each with the same summer growing season total rainfall (205 mm) but packaged into a range of many/small to few/large rainfall events. This experiment enabled a comprehensive evaluation and parameterization of drought tolerance of semiarid grassland plant functional types (i.e. deep-rooted perennials and shallow-rooted annuals) and their effects on climate extreme-carbon cycles feedbacks. The ability of the improved CLM model to capture dryland productivity and carbon fluxes was then validated at larger scales with observed carbon fluxes from closeby AmeriFlux sites in the US Southwest, such as the semi-arid Kendall grassland site (US-WKG). Applying this model in the Arizona grassland sites indicated that high tolerances of dryland plants to relatively low soil water potential maintains the growing season length of the dryland ecosystem under drought conditions, whereas the acclimation of carbon assimilation and root dynamics to drought mitigate drought effects on vegetation productivity and interannual variability of carbon exchange.

Accurate estimation of the spatio-temporal distribution of snow water equivalent is essential given its global importance for understanding climate dynamics and climate change, and as a source of fresh water. Here, we explore the potential of using the Long Short-Term Memory (LSTM) network for continental and regional scale modeling of daily snow accumulation and melt dynamics at 4-km pixel resolution across the conterminous US (CONUS). To reduce training costs (data are available for ~0.31 million snowy pixels), we combine spatial sampling with stagewise model development, whereby the network is first pretrained across the entire CONUS and then subjected to regional fine-tuning. Accordingly, model evaluation is focused on out-of-sample predictive performance across space (analogous to the prediction in ungauged basins problem). We find that, given identical inputs (precipitation, temperature and elevation), a single CONUS-wide LSTM provides significantly better spatio-temporal generalization than a regionally calibrated version of the physical-conceptual temperature-index-based SNOW17 model. Adding more meteorological information (dew point temperature, vapor pressure deficit, longwave radiation and shortwave radiation) further improves model performance, while rendering redundant the local information provided by elevation. Overall, the LSTM exhibits better transferability than SNOW17 to locations that were not included in the training data set, reinforcing the advantages of structure learning over parameter learning. Our results suggest that an LSTM-based approach could be used to develop continental/global-scale systems for modeling snow dynamics.

Lake Powell is a major reservoir responsible for providing water resources throughout the Colorado River Basin. Here we develop a hybrid model to perform 1-4 month forecasts of the inflow to Lake Powell. First, observational data are used in early December for a binary prediction of high- or low-flow based on peak monthly inflow or April-July inflow, and our prediction is much better than the Colorado Basin River Forecast Center (CBRFC)’s forecasts for the 1982-2016 study period. Second, we calibrate the NCEP Climate Forecast System seasonal forecasts of temperature, precipitation, and snow mass, and use them to perform 1-4 month inflow prediction through a regression equation, with the coefficients determined for the high- and low-flow years separately. Our early April forecasts for the critical April-July inflow are better than those of the CBRFC from 1982-2016. Sensitivity tests have also demonstrated the value of including snow mass in our prediction.

Strong winds in Southern Ocean storms drive air-sea carbon and heat fluxes. These fluxes are integral to the global climate system and the wind speeds that drive them are increasing. The current scatterometer constellation measuring vector winds remotely undersamples these storms and the higher winds within them, leading to potentially large biases in Southern Ocean wind reanalyses and the fluxes that derive from them. This observing system design study addresses these issues in two ways. First, we describe an addition to the scatterometer constellation, called Southern Ocean Storms -- Zephyr, to increase the frequency of independent observations, better constraining high winds. Second, we show that potential reanalysis wind biases over the Southern Ocean lead to uncertainty over the sign of the net winter carbon flux. More frequent independent observations per day will capture these higher winds and reduce the uncertainty in estimates of the global carbon and heat budgets.