A CONTINENTAL PHENOLOGY MODEL FOR MONITORING VEGETATION RESPONSES TO INTERANNUAL CLIMATIC VARIABILITY

Regional phenology is important in ecosystem simulation models and cou pled biosphere/atmosphere models. In the continental United States, th e timing of the onset of greenness in the spring (leaf expansion, gras s green-up) and offset of greenness in the fall (leaf abscission, cess ation of height growth, grass brown-off) are strongly influenced by me teorological and climatological conditions. We developed predictive ph enology models based on traditional phenology research using commonly available meteorological and climatological data. Predictions were com pared with satellite phenology observations at numerous 20 km x 20 km contiguous landcover sites. Onset mean absolute error was 7.2 days in the deciduous broadleaf forest (DBF) biome and 6.1 days in the grassla nd biome. Offset mean absolute error was 5.3 days in the DBF biome and 6.3 days in the grassland biome. Maximum expected errors at a 95% pro bability level ranged from 10 to 14 days. Onset was strongly associate d with temperature summations in both grassland and DBF biomes; DBF of fset was best predicted with a photoperiod function, while grassland o ffset required a combination of precipitation and temperature controls . A long-term regional test of the DBF onset model captured field-meas ured interannual variability trends in lilac phenology. Continental ap plication of the phenology models for 1990-1992 revealed extensive int erannual variability in onset and offset. Median continental growing s eason length ranged from a low of 129 days in 1991 to a high of 146 da ys in 1992. Potential uses of the models include regulation of the tim ing and length of the growing season in large-scale biogeochemical mod els and monitoring vegetation response to interannual climatic variabi lity.