CFSv2-based Seasonal Drought Predictability In China

With the increasing improvements at predictive skill of Climate Forecast Models, and ourunderstanding of climate predictability, this has resulted in adoption of operational climateprediction products in seasonal hydrologic and drought predictions. The seasonal hydrologic anddrought predictions will become feasible and beneficial for the decision-makers and the farmersto proactive decision making and product planning. The objective of this dissertation is toexplore the feasibility of operational forecasts from widely used climate model for seasonalhydrologic and drought prediction in China, and the skillful climate model forecasts willpotentially contribute to skillful hydrological and drought prediction. It is very important thatcorrectly evaluation of the climate models’ predictive skill and scientifically correction of themodel bias for seasonal hydrologic and drought prediction, which is the research focus in thedissertation. This research focus has significant realistic meaning to prevent the drought andreduce the loss of drought. The main contents of this dissertation are presented as follows:(1) Evaluating skill of seasonal precipitation and temperature predictions of NCEP CFSv2forecasts over17hydroclimatic regions in China. The seasonal predictive skill isquantified with skill scores including correlation coefficient, root mean square error andmean bias for spatially averaged seasonal precipitation and temperature forecasts foreach region. We find that the predictive skill of CFSv2precipitation and temperatureforecasts has a stronger dependence on seasons and regions than on lead-times. Bothtemperature and precipitation forecasts show higher skill during late summer (JAS) tolate autumn (OND) and during winter (DJF) to spring (MAM). The skill of CFSv2precipitation forecasts is low during summer (JJA) and winter (DJF) over entire China,due to low potential predictability of the East Asian summer monsoon (EASM) and theEast Asian winter monsoon (EAWM) for China. As expected, temperature predictiveskill is much higher than precipitation predictive skill in all regions.(2) Characterizing the skill of CFSv2-based seasonal drought prediction at multiple spatial scales as well as multiple lead times over China. A climate model’s predictive skill forseasonal temperature and precipitation generally varies with multiple factors, such aslocation, lead-time, season, and temporal and spatial scales. The construction of6different lead times forecasts are combined with observation to construct6monthsmean precipitation for computing SPI6, the drought predictive skills at6differentlead-times are calculated by the correlation coefficients between SPI6at6differentlead-times and observational SPI6. SPI6at6different lead-times and4multiple spatialscales are estimated, then the drought predictive skills at6different lead-times and4multiple spatial scales are acquired. Four spatial scales are point,2grids by2grids,5grids by5grids and10grids by10grids. The predictive skill is decreasing as the longerpredictive lengths, but the skill only increases in arid region with increased spatialscales. Meteorological drought that combines observation with forecast owes skill at4months lead-time over China. There is no drought predictive skill beyond4monthslead-time over China. The predictive skill exhibits the regionality and seasonality. Theeffect of the spatial scales is less than the effect of the temporal scales.(3) Bayesian Merging for improving CFSv2-based seasonal predictions in China. Statisticalpost-processing method, Bayesian merging which include model calibration andcombination, is used to improve raw model precipitation and temperature forecasts. TheBayesian merging approach has two important advantages: calibration and combination.CFSv2precipitation and temperature forecasts are improved by applying the BayesianMerging in spring, summer, and autumn over China. The improvements of springprecipitation forecasts in all regions are mainly derived from combining climatologicalinformation, due to the low predictable and low value of precipitation in spring. Theimprovement of summer precipitation forecasts are mainly in humid regions. Theimprovement of autumn precipitation forecasts are in inland regions except southernTibet. The improvements of temperature are seen in all three seasons and seventeenregions.(4) CFSv2-based seasonal drought prediction over China. This part is to explore thepotential of seasonal drought prediction from CFSv2over China. We implement twoexperiments, drought monitoring and prediction experiments, which observedprecipitation and temperature forcing and bias-corrected CFSv2precipitation andtemperature prediction were used to drive hydrologic model to get soil moisturemonitoring and prediction respectively. Monthly total column soil moisture wasconverted to the percentile as the soil moisture drought index. Three seasonal droughtpredictive events, include autumn in1997, spring in2002and summer in2003, are presented. The1997-2003droughts shown by the soil moisture monitoring droughtindex mainly exist in the region between the Yellow and Yangtze Rivers, innorthwestern and northeast of China. The spatial extent of the predicted soil moisturedrought index in autumn1997is similar to that of the simulated soil moisture droughtindex, but their magnitudes of severity are obviously different. The predicted soilmoisture drought index in spring2002cannot predict drought at2-3months lead, justshow climatological soil moisture conditions. The predicted soil moisture drought indexcan catch the severity of the simulated soil moisture drought index in Yellow River andnorthern China, but there is the weak in predicted soil moisture drought index innortheastern and southern China.