Understanding the sources of US drought predictability using seasonal reforecasts of sixty years (1958-2017) initialized with multiple land analyses

Based on a set of recently completed ensemble seasonal reforecasts covering 1958-2014 (to be extended to 2017) using the NCEP Climate Forecast System, version 2 (CFSv2) and initialized from observation-based ocean, land and atmospheric states, we propose to evaluate the predictive skill of US droughts over 60 years and to identify the sources of prediction skill in the ocean, land and global climate trends. As far as we know, this will be the first time that a record of this length of seasonal reforecasts would be used for the study of drought, comparable to the Atmospheric Model Intercomparison Project (AMIP) simulations that are widely used in drought mechanism studies. Taking advantage of this long reforecast dataset, together with AMIP simulations by the CFSv2 atmospheric component with large ensemble sizes, we plan to conduct the following studies of US drought predictions: (1) We will examine whether there are differences in US drought predictioskill in the 1960s-1970s, 1980s-1990s and 2000s-present, associated with different phases of the Pacific decadal oscillation. For this purpose, we will use an advanced statistical method to extract the predictable patterns of the US precipitation from ensemble predictions and examine the connection of these patterns to the ocean, land and atmospheric forcing factors during each of these three periods. We will also examine the mechanisms and model prediction skills of selected major drought events during each of these three periods. Sensitivity experiments will be conducted to identify the potential contributions of specific SST anomalies on the particular events. (2) We will examine the contributions of the observation-based land initial states to the seasonal predictions of the US precipitation and drought events. For this purpose, we will conduct land initialization experiments in which climatological soil moisture will be used to initialize a selected group of drought events. The comparison of these experiments with the reforecasts will isolate the effects of the initial land signals. Furthermore, we will conduct a set of reforecasts using an alternative set of model-based land initial states and specified persisted anomalies to assess the effects of the uncertainty of the land initial states. (3) Using the 60-year reforecasts and the AMIP runs, we will examine the global change trends in evapotranspiration and its potential influence on US droughts by comparing the variability of the temperature, evapotranspiration and precipitation between earlier and later periods (e.g., 1960s-1970s vs. 2000s-present) associated with observed CO2 forcing and initialized or specified boundary conditions. This proposal addresses two of the principal objectives in the MAPP drought project, i.e., developing a better understanding of sources of predictability toward improving predictions of drought onset, evolution, and termination on subseasonal to interannual timescales, and understanding the role of the temperature and evapotranspiration in affecting droughts. The Center for Ocean-Land-Atmosphere Studies and the US Climate Prediction Center will work closely to conduct this collaborative research.

Climate Risk Area: Water Resources