Long-term interactions of climate, vegetation, humans, and fire in eastern Washington

Abstract

Modern ecological studies are unable to examine long-term processes operating on the order of hundreds of years. Because of the limited length of modern and historic records, questions about long-term interactions between people and the environment can only be answered using paleoecological and archaeological information. This dissertation uses prehistoric records spanning the past millennium to examine issues of human paleoecology on the Columbia Plateau of Washington State. Unlike many previous studies, this study (1) quantifies past human population; (2) compares relative inputs of humans, climate, fire, and vegetation using multivariate statistics; (3) examines relationships between variables when leads and lags of different lengths are introduced; (4) identifies multicolinearity, allowing variables of no explanatory value to be eliminated.For this analysis, lake sediments provided pollen, charcoal, and oxygen-isotopes that served as proxies for past vegetation, fire, and climate, respectively. These data were compared to a previously published paleodemographic reconstruction. Results showed that simple bivariate analyses were misleading. For example, human population was significantly correlated with charcoal when charcoal was lagged by 0, 50, 100, or 150 years, indicating that human action led to future increases in fire. When past all variables were all entered simultaneously in a multivariate analysis, however, population was shown to have no effect on fire after 50 years. Vegetation (pine woodlands) was shown to have a significant short-term influence on fire, and climate was shown to be the only long-run predictor of fire. Similar analyses were run to determine which factors (if any) were patterning future values of vegetation, human population, and climate. Results showed that the history of the system had an important influence on ecological outcomes. The analysis showed that the different factors in the ecosystem were intricately interrelated to one another, with feedbacks occurring between the variables over several different time scales. This study indicates that research on human impacts that focuses on bivariate patterns, such as simple comparisons of coeval human population and fire, can suffer from the problem of equifinality. The protocol employed in this work can avoid this problem, and has the potential to be used in any area.