Recipient Organization
UNIVERSITY OF ARKANSAS
(N/A)
FAYETTEVILLE,AR 72703
Performing Department
Forestry And Natural Resources
Non Technical Summary
The natural resources in the state of Arkansas is facing critical challenges resulted from warming climate and intensified human activities. It is important for us to document such changes, such as how many lands have been converted from wetlands to croplands, and from forests to human settlements. Earth observation techniques offer us this opportunity to determine the amount and distribution of natural resources at high accuracy and fine resolution, not only for the contemporary time, but also for the old decades. The main purpose of this project is to quantify the long-term records of natural resource dynamics, as well as the healthy conditions of forest ecosystems disturbed by beetles via an integration of geospatial techniques, ecological niche modeling and landscape ecology analysis. This type of information can be used by decision makers to evaluate the effectiveness of their programs, such as the Conservation Reserve Program, and can be applied in other scientific domains, such as the estimation of total carbon emissions caused by the loss of forests, and biodiversity decline due to the degradation of wetlands. In the meantime, we realize that changes in natural resources could bring catastrophic consequences to human health, which can get intensified with the climate change. But questions like how will the changes happen, to which degree will the changes affects the human health system, will adverse and beneficial effects co-exist, and which effect will play a dominant role, remain uncertain. Thus, another goal of this project is to elucidate the relationship between climate change, natural environment and human health, which will be beneficial to plan adaptation and mitigation strategies ahead.
Animal Health Component
70%
Research Effort Categories
Basic
20%
Applied
70%
Developmental
10%
Goals / Objectives
Arkansas is an area rich in natural resources, but is also a vulnerable region under the intensified pressure from human activities and climate change. For instance, forest resources, which occupy a large portion of this natural configuration, experiencing dramatic changes over the past decades. During the 1980s, 5.2 million acres of timberlands in the Lower Mississippi Alluvia Valley were lost to agriculture and development due to large-scale land clearing and farming. Since the 1980s, conservation incentive programs, such as the Conservation Reserve Program, have been offered to farmers to convert or revert environmentally sensitive land from agricultural production to forests, and forest area has been observed gradual and noticeable increases. Besides human impacts, insect disturbances is another factor that is significantly altering forest ecosystem's composition, structure and function. For example, the southern pine beetle (SPB) is the most destructive insect pest of pine forests in the South, which had killed $493 million total value of trees from 1991 to 1996. The effects of insect infestation range from altered surface fuel and wildfire hazards, changed vegetative composition, converted live carbon sinks to dead and slowly decaying carbon sources, impacted nutrient cycling and water quality, modified local surface energy balance and changes in the regional climate. By infesting and killing stressed trees, insects could result in complex dynamic patterns of dead trees, and the subsequent recovery will also be highly spatially and temporally variable.Changes in geographic ranges and conditions of natural resources are of critical importance to biodiversity conservation, environment protection, bioenergy development, water and food security, and carbon sequestration. Thus, accurate, timely and spatially explicit retrospective review on the landscape patterns, and understanding the biophysical and anthropogenic drivers of the changes are critically important to natural resource managers, decision makers and stakeholders. However, given its importance, there is a substantial lack of information with regard to natural resource distribution and condition changes in Arkansas. The value of geospatial techniques in the studies of natural resources has long been recognized, and it is a powerful tool to produce long-term, fine-resolution, and frequently updated data that can depict the landscape dynamics over time.Changes in natural resources and environment, such as deforestation and biodiversity loss, will not only bring catastrophic consequences to the material and energy cycle of ecosystems, but will also cause detrimental health impacts on humans. For instance, the decreasing tree cover and biomass could convert many forest ecosystems from a carbon sink to a carbon source, and in turn cause heavier pollution, toxins and more weather extremes. The accompanied urbanization and agricultural intensification with deforestation could reduce the suitable habitat for forest wildlife, expel them into human crowed zones, raise the chances of disease transmission from wild pathogen hosts to humans, and cause genetic mutation. Meanwhile, climate change is accelerating the process of environment deterioration, and pose huge threats to human health. For instance, the rising number of large forest fires as witnessed over the past decades, are indicated to be affected by the variation in climate factors. Emissions from wildfires, such as particulate matter and carbon monoxide, could have significant impacts on local air quality, and consequently affect human's respiratory system. Thus, a better understanding towards the relationships among natural environment, climate change, and human health is necessary to protect public and environmental health, such as the design of pro-active mitigation and reliable projection of vulnerability, yet it remains unsolved due to the complexity in the multiple component and system interaction in this socioenvironmental issue.The overall goal of this project is to develop theoretical models and novel datasets in GIScience, to monitor the geographic distribution and condition dynamics of natural resources, and to solve the natural resource related health issues. Three specific objectives are listed as following:Long term monitoring and spatial-temporal simulation of forest insect disturbance dynamics.Monitoring landscape dynamics and conditions of natural resources in the Mississippi Alluvial Valley of Arkansas.Evaluate relationships among climate change, natural environment and human health.
Project Methods
To achieve the first objective of this project, I plan to use a forest-growth trend analysis methodology that integrates temporal trajectories in remotely sensed images and decision tree techniques to generate annual forest disturbance maps for mountain pine beetle, southern pine beetle, and red oak borer over a period of one decade. The classification accuracy will be evaluated by field survey, FIA plots, and visual interpreted samples from high-resolution images. Future infestation scenarios will be simulated using a cellular automata (CA) model. The CA model will be developed upon the time-series Landsat imagery derived forest disturbance map. To test whether the CA model is not confined to specific sites and can be widely applicable, three similar-sized study sites will be chosen across the three different forest regions. Training samples for CA modeling will be extracted through a stratified random sampling scheme, and be used to train the key parameters in CA model including transition rules, neighborhood, constraints, and stochastic perturbation. A stepwise comparison will be employed to conduct the sensitivity analysis of the operational components in the proposed Insect-CA model. To assess the performance of the CA models, I will compare the simulation results with the observed maps via two ways: pixel-based accuracy, and pattern-based spatial metrics. Overall accuracy will be used as an indicator for pixel-based assessment.The second objective of this project will be completed via a multi-stage classification workflow. First, 1-m resolution NAIP images will be conducted with object-based classification. In the segmentation algorithm, three parameters are required: threshold, shape and compactness. Multiple combinations of parameters and certain quantitative measurements will be utilized to select the optimal combination. The purpose of segmentation is to partition the landscape into a group of objects, with each object representing a homogeneous patch of cropland field. Shape metrics, such as perimeter-area ratio, fractal dimension index, will be calculated with FRAGSTATS. At the second stage, Landsat time-series stack will be employed to extract vegetation phenology information, such as the beginning of green-up or dormancy. Those phenological variables can be treated as proxies for the seasonal progress of crops. Since various crop types should possess different growing patterns, a set of optimally selected and calibrated phenological metrics is thus capable of classifying them into distinct categories. Finally, all shape metrics and phonological variables will be used as inputs in the Random Forest classifier to generate multi-year cropland classifications. Random Forest is a machine-learning technique that combines the results of thousands of weak classifiers, such as classification and regression trees, and is often praised for its good overall performance, as well as robustness to noise.The major classification and accuracy assessment tasks will be conducted in the data mining software WEKA. The maps will be evaluated by comparing the results with FIA forest resources assessment reports in 1986 and 2013 at both the county and state level.The third objective of this project will be conducted through an integrated meta-studies analysis and geospatial technique. I will conduct the original research literature search in the mainstream bibliographic databases. Basic literature information and abstract information will be retained for the next step of literature filtering. Firstly, duplicated papers from different bibliographic databases will be removed. Secondly, the literatures will be ranked based on their quality. Peer-reviewed papers with specific findings or conclusions supported by data collection and solid experimental or computational design will be ranked as the highest quality, while literatures in other forms such as review, editorial, commentary, will be ranked lower. Finally, only articles with the focus on human related health issues can be retained. The climate change-natural environment-human health relationship will first be understood using exploratory spatial data analysis. I will further use regression to quantitatively describe the relationship between three terms of interest. In order to show connections among various thematic terms, such as disease types, climate variables, environmental factors, methodology, and spatial-temporal scales, a "climate-environment-disease-method-scale" thematic network that is made up with a set of nodes and edges will be designed. In this web-like illustration network, each node is one thematic term with its size representing how well a node is connected, and the edge thickness indicates the frequency of occurrence of a specific relationship. Based on this network, the connectivity degree between various thematic terms can be visualized and quantified.