Developing analytical approaches and models to address forest health issues that affect the sustainability of forest ecosystems is greatly needed. The proposed work will adopt a multi-scale/level (individual tree, stand, landscape, entire region) modeling approach evaluate forest health condition under a changing climate and to quantify the effect of multi-scale factors on forest health indicators. A hierarchical, spatial modeling framework will be developed to quantify the change of forest health indicators with climatic, geographical and site factors by using Forest Inventory and Analysis (FIA)/FHM data and downscaled climate data. This research will help the Gulf Research Program and other agencies to fill the technical and information gap in climate change and coastal longleaf pine forest ecosystem health.
Data acquire and methods
Publicly available data will be used in this project. Plot level data will be collected from the FIA DataMart. The coastal longleaf pine forest ecosystem along the Gulf of Mexico will be modeled. It covers the five southern states; Texas, Louisiana, Mississippi, Alabama, and Florida. Longleaf pine plots with in the selected forest type will be identified from the FIA database which is publicly available. FHM indicators along with the spatial locations and other plots level data will be collected from the selected FIA plots. FIA/FHM data from 2001 to 2015 will be classified into three time interval groups: 2001-2005, 2006-2010, and 2010-2015. Those groups are the representation of the entire range and provide three levels of temporal data.
Monthly climatic data (1970-2070) will be used in this analysis which will be acquired from the Western Kentucky University. The data are at 10km resolution including the station based observational data from 1970 to 2016 and projected data with the Weather Research and Forecasting (WRF) model.
Analytical approaches
FIA/FHM plot data/health indicators will be transformed and aggregated at three levels: Ecoregion, landscape level (watershed and/or forest type), and FIA/FHM plot cluster based on Bailey’s ecological classification system hierarchy. Considering the non-stationary characteristics of a forest health indicator, we will first spatially classified the study domain (the southeast US) into a set of disjoint regions based on the spatial variation of a selected health indicator by using nonparametric kernel smoothing method. The disjoint regions will be further classified into a set of smaller of FIA/FHM clusters with a cluster only belonging to the same landscape and ecoregion. A FIA/FHM cluster is a basic unit reflecting the interaction of ecoregion (climate) factors, forest type (bio-geographical) factors and local disturbance and site factors. Because of the weak correlation between plot data and climate and vegetation data, probability models instead of regression (mean-based) models will developed to predict the probabilistic distributions of a health indicator within a cluster.
The regression models will be developed based on a set of embedded FIA/FHM clusters and associated factors by using statistical resampling methods. The regression models are to predict the quantity/abundance of a health indicator. Within an ecoregion (section), the distribution parameters (e.g., mean, variance) of a health indicator and its change with climate will be estimated by using stochastic spatial-temporal models. The three-level hierarchical modeling will be implemented in the Bayesian framework through the Markov chain Monte Carlo (MCMC) methods.
Impact/Significance of the Synthesis
Overall, this synthesis research will be a very useful decision making tools to the Gulf Research Program and other agencies to take decision regarding the climate change and coastal longleaf pine forest ecosystem health. Compared to other studies, the hierarchical modeling framework is unique in the following three aspects. 1) It takes a bottom-up approach starting with FHM/FIA plots and uses downscaled fine resolution climate data as model input for other site/plot variables. The downscaled fine-resolution (10 km) climate data will overcome the scale mismatch problem in data analysis and modeling. Moreover, the downscaled climate data include observed or projected climate data for the past, current and future time. The data thus make possible model validation and comparison, which were poorly addressed in other studies. 2) We fully consider the uncertainty of forest health indicators and employ a stochastic aggregation process to quantify health indicators dynamics by ECS levels (e.g., ecoregion, forest type/group, land type association). The ESC hierarchy matches well at the resource planning and decision making level. 3) We use ground truth (FIA/FHM plots) to quantify climate change impact on FHM health indicators at multiple levels.
Author: Sunil Nepal Topics: Forest health
Summary of synthesis
Developing analytical approaches and models to address forest health issues that affect the sustainability of forest ecosystems is greatly needed. The proposed work will adopt a multi-scale/level (individual tree, stand, landscape, entire region) modeling approach evaluate forest health condition under a changing climate and to quantify the effect of multi-scale factors on forest health indicators. A hierarchical, spatial modeling framework will be developed to quantify the change of forest health indicators with climatic, geographical and site factors by using Forest Inventory and Analysis (FIA)/FHM data and downscaled climate data. This research will help the Gulf Research Program and other agencies to fill the technical and information gap in climate change and coastal longleaf pine forest ecosystem health.
Data acquire and methods
Publicly available data will be used in this project. Plot level data will be collected from the FIA DataMart. The coastal longleaf pine forest ecosystem along the Gulf of Mexico will be modeled. It covers the five southern states; Texas, Louisiana, Mississippi, Alabama, and Florida. Longleaf pine plots with in the selected forest type will be identified from the FIA database which is publicly available. FHM indicators along with the spatial locations and other plots level data will be collected from the selected FIA plots. FIA/FHM data from 2001 to 2015 will be classified into three time interval groups: 2001-2005, 2006-2010, and 2010-2015. Those groups are the representation of the entire range and provide three levels of temporal data. Monthly climatic data (1970-2070) will be used in this analysis which will be acquired from the Western Kentucky University. The data are at 10km resolution including the station based observational data from 1970 to 2016 and projected data with the Weather Research and Forecasting (WRF) model.
Analytical approaches
FIA/FHM plot data/health indicators will be transformed and aggregated at three levels: Ecoregion, landscape level (watershed and/or forest type), and FIA/FHM plot cluster based on Bailey’s ecological classification system hierarchy. Considering the non-stationary characteristics of a forest health indicator, we will first spatially classified the study domain (the southeast US) into a set of disjoint regions based on the spatial variation of a selected health indicator by using nonparametric kernel smoothing method. The disjoint regions will be further classified into a set of smaller of FIA/FHM clusters with a cluster only belonging to the same landscape and ecoregion. A FIA/FHM cluster is a basic unit reflecting the interaction of ecoregion (climate) factors, forest type (bio-geographical) factors and local disturbance and site factors. Because of the weak correlation between plot data and climate and vegetation data, probability models instead of regression (mean-based) models will developed to predict the probabilistic distributions of a health indicator within a cluster.
The regression models will be developed based on a set of embedded FIA/FHM clusters and associated factors by using statistical resampling methods. The regression models are to predict the quantity/abundance of a health indicator. Within an ecoregion (section), the distribution parameters (e.g., mean, variance) of a health indicator and its change with climate will be estimated by using stochastic spatial-temporal models. The three-level hierarchical modeling will be implemented in the Bayesian framework through the Markov chain Monte Carlo (MCMC) methods.
Impact/Significance of the Synthesis
Overall, this synthesis research will be a very useful decision making tools to the Gulf Research Program and other agencies to take decision regarding the climate change and coastal longleaf pine forest ecosystem health. Compared to other studies, the hierarchical modeling framework is unique in the following three aspects. 1) It takes a bottom-up approach starting with FHM/FIA plots and uses downscaled fine resolution climate data as model input for other site/plot variables. The downscaled fine-resolution (10 km) climate data will overcome the scale mismatch problem in data analysis and modeling. Moreover, the downscaled climate data include observed or projected climate data for the past, current and future time. The data thus make possible model validation and comparison, which were poorly addressed in other studies. 2) We fully consider the uncertainty of forest health indicators and employ a stochastic aggregation process to quantify health indicators dynamics by ECS levels (e.g., ecoregion, forest type/group, land type association). The ESC hierarchy matches well at the resource planning and decision making level. 3) We use ground truth (FIA/FHM plots) to quantify climate change impact on FHM health indicators at multiple levels.