Current Projects
Global Cropland and Food Security Analysis Over Four Decades Using Landsat and MODIS Satellite Data Fusion
Our lab is mapping croplands across the entire continent of North America as a part of a global cropland mapping effort. The overarching goal of the Global Food Security-Analysis Data at 30 m (GFSAD30) project is to map and monitor global croplands for ensuring sustainable water and food security in the twenty-first century. It is a NASA-funded, 5-year project in collaboration with the USGS, NASA GSFC, Northern Arizona University, University of New Hampshire, University of Wisconsin-Madison, Google Earth Engine, NASA Ames, and ICRISAT. The currently available cropland products have five major limitations:
Absence of precise spatial location of the cropped areas,
Coarse resolution of map products with significant uncertainties in areas, locations, and detail,
Uncertainties in differentiating irrigated croplands from rain-fed croplands,
Absence of crop types and cropping intensities, and
Absence of a dedicated web\data portal for the dissemination of cropland products.
At NAU, we are focusing on North American cropland mapping algorithm development, validation, and implementation, and Landsat and MODIS satellite data fusion to improve the temporal resolution of the fine spatial resolution data. We leverage NAU’s new computing cluster Monsoon in this data-intensive geospatial analysis. As we move from coarse-resolution MODIS-based classification to finer-resolution Landsat-based classification, our dataset increases exponentially in size. We estimate that the Landsat-based classification in our lab can take up 5 years when performed on an average desktop computer. However, the same analysis would take 92 hours on Monsoon. Our lab is producing the following map products as a part of this project:
Cropland area extent
Classification 8 dominant crop types that occupy 70% of the global cropland areas
Classification of irrigated versus rain fed croplands
Classification of cropping intensity: single, double, triple, and continuous cropping.
The current version of the North American cropland map is available on LPDAAC and Croplands.org. We published our work on mapping crop types in the United States at 250m in the journal Remote Sensing of Environment. The publication can be accessed here and the 250m crop-type data for the US is available here.
Predicting Snow Water Equivalence (SWE) and Soil Moisture Response to Restoration Treatments in Headwater Ponderosa Pine Forests of the Desert LCC
The U.S. Forest Service is poised to conduct forest restoration treatments through the Four Forest Restoration Initiative (4FRI) on hundreds of thousands of acres of ponderosa pine forest along the Mogollon Rim of Arizona with the goals of reducing wildfire hazard and improving forest health. One of the 4FRI’s objectives is to thin and burn to accomplish within-stand openings that “promote snowpack accumulation and retention which benefits groundwater recharge and watershed processes at the fine (1 to 10 acres) scale.” However, little is known about how the openings created by restoration treatments affect snow water equivalence and soil moisture, which are key parts of the water balance that greatly influence water availability for healthy trees and for downstream water users in the Sonoran Desert.
This project is funded by the Bureau of Reclamation for 3 years. In this project, we have 3 main objectives:
Couple remotely sensed satellite data analysis with field measurements,
Analyze spatial relationships among forest pattern, topography, SWE, and soil moisture using geospatial statistics,
Develop a predictive model of SWE and soil moisture responses to various forest restoration treatments and the consequent patterns of canopy openings.
Five existing forest restoration research sites on the Coconino National Forest and NAU’s Centennial Forest are revisited; these sites received restoration treatments within the past 5 to 20 years as part of scientific studies conducted by NAU School of Forestry, Rocky Mountain Research Station, and the Ecological Restoration Institute. At “control”, “thinned” and “thinned and burned” treatment units, we systematically measure soil texture, soil bulk density, soil moisture, and snow water equivalence through the use of in-situ soil moisture sensors, field sampling, and snow course data collection.
WorldView-2 satellite data with 2.4 m resolution and approximately monthly frequency is analyzed to evaluate surface moisture, snow extent, and forest pattern during the winter-spring seasons. This high spatial resolution satellite data is complimented with the analysis of finer resolution UAV data and coarser spatial resolution, but higher temporal resolution satellite data from the Landsat 8 (30 m resolution and 16-day interval) and MODIS (250 m resolution with 2-day interval) sensors.
We hypothesize that forest restoration treatments that create a heterogeneous pattern of tree groups and openings, with average opening sizes ranging from ¾ to 1½ times as wide as the adjacent tree height, generate the greatest SWE and May/June deep rooting zone (50 cm) soil moisture. Findings of this study will inform adaptive management and help guide silviculturalists in selecting restoration treatments that are optimal for preserving precious moisture to promote a more resilient forest. By promoting forest resilience we can help ensure healthy forests and healthy watersheds for decades to come.
We published some of the results from this work in the journal Remote Sensing of Environment and the publication can be accessed here
Remote Sensing of Tamarisk Change and Insect Herbivory in the Grand Canyon
We use a cutting-edge remote sensing technique to map and detect changes in tamarisk, an invasive tree species in the Grand Canyon. Tamarisk trees are commonly distributed in riparian habitats across the western US. Once established, it can dominate a site by drawing up belowground salts and increasing saline soils that native plant species can not tolerate.
To control tamarisk, a biological control, known as tamarisk beetle (Diorhabda carinulata), was introduced to Colorado, Utah, Wyoming, Nevada, California, and Texas. Tamarisk beetles traveled into Arizona in 2009 and have had substantial impact on tamarisk distribution along the Colorado River.
The extent of its impact and the associated tamarisk decline, however, have not been quantified. Many natural resource management concerns and scientific research questions related to tamarisk defoliation and its impact on the ecosystem (i.e., nutrient cycle, native species recovery) in the Grand Canyon hinge on quantitative estimates of tamarisk decline.
The US Geological Survey (USGS) Grand Canyon Monitoring and Research Center (GCMRC) has acquired airborne, high-resolution multispectral data of the Grand Canyon and 3-dimensional lidar images of Glen Canyon. In collaboration with the USGS scientists, we analyze the lidar data to identify tamarisk trees and classify defoliated and undefoliated tamarisk trees along a continuum of low-to-high canopy openness in Glen Canyon. We also further expand the USGS field-based observations through field trips to the Grand Canyon.
In 2016, we published this work in the journal Photogrammetric Engineering and Remote Sensing (PE&RS) and the paper can be accessed here
Long-term Vegetation Changes Associated with Climate Trends in Southwestern National Parks
The National Park Service faces tremendous management challenges in the future as climates alter the abundance, distribution, and interactions of plant species. Synthesis of climate and plant community composition data from Inventory and Monitoring (I&M) networks is essential to provide resource managers with important insights to contemporary climate responses and a sound basis to forecast likely future changes at species, community, and ecosystem scales.
We are conducting a regional cross-site analysis to build an empirically-driven model that forecasts how dryland plant communities in the southwestern US respond to climate change. To do this, we integrate Landsat satellite image analysis with past local and regional patterns in climate and long-term vegetation datasets to identify plant species and functional types that increase or decrease with climate change.
Following vegetation change detection, we compile regional-scale, gridded climate data sets including PRISM, Daymet, and WorldClim data. We plan to implement recently developed NPS tools, including the Climate Grid Analysis Toolset Measure to assist in data acquisition and assess regional climatological trends across the study region. Topographic aspect and slope are derived from digital elevation models (DEMs). We also characterize soils and topography of the study region using the NPS soil inventory, in addition to SSURGO/STATSGO databases.
These climate and site layers are used in a GIS environment to understand and explain the patterns of climate-induced vegetation change across the study region. A multiple regression technique is used to assess how climate has affected vegetation in the context of the topography and soils that modulate plant water availability.
Our recent results have been published in the journal Ecological Applications and can be accessed here
Evaluating Forest Fuel Reduction Treatment Effectiveness with UAV images: Fire Behavior Analysis
Historic land use and fire suppression in the southwestern USA have created forest conditions that are highly susceptible to catastrophic wildfires. Land managers have been designing and conducting fuels treatments to mitigate the effects of these wildfires. In November 2012, residents of Flagstaff, AZ voted on a $10 million fuel reduction treatment to help protect the city from wildfire and subsequent flooding, which could damage key areas of the city and its water supply.
As a result, the Flagstaff Watershed Protection Project (FWPP) area is currently being treated and we are examining the fuel reduction treatment effects with pre- and post-treatment measurements in the Coconino National Forest directly adjacent to the City of Flagstaff.
Unmanned aerial images offer an innovative way to measure forest fuel structure and monitor fuel reduction treatments. UAVs equipped with a multispectral sensor, and combined with structure-from-motion (SfM) software, can produce three-dimensional (3D) representations of forests that can then be used to measure forest structure.
Measurements derived from UAV data can then be used as an input into the FlamMap fire behavior software to model fire behavior. We are testing the effectiveness of the UAV-derived models to demonstrate the potential of FlamMap for fine-scale, site-specific, applications which can aid in treatment prioritization and monitoring.