Thu, Aug 18, 2022: 10:15 AM-10:30 AM
516B
Background/Question/MethodsThe regional impact of wildfires on forest carbon is difficult to quantify, because pre-fire data are usually lacking and randomized experiments are impossible. Quasi-experimental methods can allow causal impact analyses from observational data where controlled experiments are unavailable. This study examines the applicability of quasi-experimental methods to estimate the impact of wildfire on aboveground forest carbon mass using national forest inventory data of Oregon and Washington, USA. Further, this study establishes guidelines for implementing distance-adjusted propensity score matching (DAPSM), a quasi-experimental method for ecological data especially useful for spatially located national forest inventory data. First, we compared three different matching methods for wildfire impact quantification: 1) propensity score matching (PSM), 2) spatial matching (SM), and 3) distance-adjusted propensity score matching (DAPSM). Then, we performed a sensitivity analysis on the impact of different data availability scenarios on the performance of DAPSM, using Monte-Carlo simulations. The results suggest a marginal sample size and key covariates for implementing matching methods for wildfire impact analyses.
Results/ConclusionsThe major findings are twofold: First, incorporating spatial information was essential in applying matching methods to forest inventory data to quantify the regional impacts of wildfires on forest carbon. Therefore, DAPSM was favored over conventional PSM and SM to account for the influence of both environmental covariates and spatial distances among the sample plots. Second, the inclusion of the spatial distance compensated for the omission of key covariates, but this compensation was not effective for small sample sizes. The marginal sample size for implementing DAPSM using national forest inventory data was ³ 100 disturbance-affected plots and ³ 1,000 disturbance-unaffected plots. Climate variables were the key covariates for assessing wildfire impacts on forest carbon.While quasi-experimental methods have been widely used in health, econometrics, and social sciences, this study shows that they can be applied in an ecological context to examine the effects of disturbance or other natural phenomena. The outcome of the study provides practical protocols for implementing quasi-experimental methods for regional wildfire impact analyses using national forest inventory data. This study highlights the need to consider the spatial components in causal inference with ecological data, and provides recommendations for data requirements based on applied examples.
Results/ConclusionsThe major findings are twofold: First, incorporating spatial information was essential in applying matching methods to forest inventory data to quantify the regional impacts of wildfires on forest carbon. Therefore, DAPSM was favored over conventional PSM and SM to account for the influence of both environmental covariates and spatial distances among the sample plots. Second, the inclusion of the spatial distance compensated for the omission of key covariates, but this compensation was not effective for small sample sizes. The marginal sample size for implementing DAPSM using national forest inventory data was ³ 100 disturbance-affected plots and ³ 1,000 disturbance-unaffected plots. Climate variables were the key covariates for assessing wildfire impacts on forest carbon.While quasi-experimental methods have been widely used in health, econometrics, and social sciences, this study shows that they can be applied in an ecological context to examine the effects of disturbance or other natural phenomena. The outcome of the study provides practical protocols for implementing quasi-experimental methods for regional wildfire impact analyses using national forest inventory data. This study highlights the need to consider the spatial components in causal inference with ecological data, and provides recommendations for data requirements based on applied examples.