Value of Land Change Monitoring, Assessment, Projection Collection 1

Video Transcript
Download Video
Right-click and save to download

Detailed Description

The USGS Land Change Monitoring, Assessment, and Projection (LCMAP) Initiative scientist Jennifer Rover offers examples of LCMAP Science Product uses at the 2020 American Geophysical Union (AGU) Winter Meeting.
 

Details

Date Taken:

Length: 00:12:15

Location Taken: Sioux Falls, SD, US

Video Credits

Authors: Jennifer Rover, Christopher Barber, USGS EROS
 

Transcript

The Land Change Monitoring, Assessment, and Projection (LCMAP) initiative by the U.S. Geological Survey (USGS) addresses the pressing need to monitor and understand land change. LCMAP is the foundation for an integrated USGS-wide land change science framework and supports the development of consistent data and land cover products, spanning large geographic extents over extended periods and at a higher frequency than in the past. 
LCMAP published 10 new science products. All are developed from Landsat Analysis Ready Data using a time series modeling approach. The products include annual land cover using a classification similar to Anderson level 1. LCMAP provides users with both primary and secondary land cover, and their associated model confidences. The annual land cover change product is a derivative of the primary land cover. When a change occurs between years, pixels are attributed with the from-to classes. The spectral products describe change as indicated by the Landsat time series model, which is estimated at a daily timestep. The spectral model quality product describes model details for each pixel.
All LCMAP products are currently available for the contiguous United States. The Landsat ARD are modeled through time, at a daily timestep. Aspects of the resultant models are then described in a suite of ten products. The spectral change products are meant to inform one another as they are coincident in time and space. The spectral change products include time of spectral change, this is when spectral data no longer fit the model and cause a "break". The change magnitude is a measure of the spectral difference before and after the break. Other products include a count of days since the break.  A count of days the current model was stable. Again, the spectral model quality data are also provided.
Those data and a boosted decision tree (XGBoost) were used to classify land cover. LCMAP provides an annual product for the most likely cover class on July 1 of each year. Also provided are the 2nd most likely cover class on July 1 and the associated confidences of each classification, enabling users to determine if secondary class assignments are useful for their specific applications.
LCMAP Collection 1 science products are useful for a variety of science applications. The products are appropriate for assessing land changes at local to continental scales. 
Land Surface Changes important to the land science community include single event types of change related to natural hazards such as land slides, wildfire, damage from tornadoes and hurricanes, and other gradual changes such as urban sprawl and increasing surface water in response to shifts in climate. 
In an animation of the annual land cover change product, the dark green color represents pixels classified as tree cover. In the tree covered areas, purple pops up. Purple represents class change. The tree cover is replaced by grass shrub after timber harvest. Once the trees are replanted and the stand matures, another land class change occurs, grass-shrub to tree cover. This pattern through time exemplifies the cylindrical nature of forest harvest and regrowth.
A change in the year 2014 of the animation was a devastating landslide. In the space of a minute, 19 million tons of debris traveled at 40 mile-per-hour. The Oso landslide involved a complex sequence of geological and hydrological events that ultimately resulted in a debris-avalanche flow.
Zooming into the landslide area, the area is yellow in the Time of Spectral Change product. These changes were detected during the months of March and April. The timing of the change aligns well with the March 22 event.
The magnitude product shows that the difference in the spectral properties before and after the landslide. The magnitudes were greatest for the area where the initial instability or slope failure occurred. Another area with high magnitude is where debris from the landslide destroyed the tree cover.
The Santa Fe National forest in New Mexico has a history of wildfires, many that are identified as land cover change in the LCMAP change product. On the left, mapped fires from Monitoring Trends in Burn Severity (MTBS) are shown. The red perimeter for is for the largest fire in New Mexico's history. The Las Conchas fire started on June 26, 2011 in the Jemez (Heymez) Mountains. When the fire was contained on August 3, 2011, it had burned nearly 160,000 acres of mixed conifer, pinyon/juniper, and ponderosa forest. The high-severity of the fire was associated with forest health due to drought and bark beetle stress.
In the Time of Spectral Change Product, changes were estimated to occur in the July time frame.  Most of the change occurring in 2011 correspond to the MTBS Fire perimeter, shown in red. Since the fire occurred towards the end of June and burned until August, an estimate of timing of change during the month of July is reasonable. The Magnitude of spectral change, coincident in time, shows the majority of the fire perimeter had moderate spectral change. The variance in magnitude suggests some areas may have more variation in burn severity. 
On April 27, 2011, a record number of tornadoes touched down across the southern United States, resulting in the loss of life and property. The tornadoes paths appear as long purple stripes of change in the annual Landcover change products In the Time of Spectral Change product, the estimate for the timing of change, or model break was early May 2011, just after the deadly storm. The Change Magnitude product also captures surface changes and indicates a moderate spectral response for pixels in the tornadoes' paths. 
Hurricane Andrew, it's path shown as a green line, formed over the Atlantic Ocean and made landfall in South Florida in August of 1992. The hurricane was a Category 5 storm, causing significant damage to Wetlands in Everglades National Park. Perhaps the most dramatic effect of its passage through the park was the major structural damage to trees and mangroves caused by strong winds and storm surge.
The Time of Spectral Change product indicates changes were detected along the coasts in September and October, shown in purple, while the magnitude product indicates moderate change corresponding to the September/ October time frame.
In August of 2005, Hurricane Katrina, its path shown in green, made landfall on Atlantic side of Florida as a cat 1 and weakened only slightly before moving into the Gulf of Mexico. 
Later that same year, in late October, Hurricane Wilma, its path shown in purple, made landfall in southwest Florida and tracked towards the NE. Most of the damage is on the Gulf coast of the state and once again, the spectral magnitude indicates moderate change. Unlike Hurricane Andrew, on the previous slide, these hurricanes had a more northerly path and change isn't prevalent on the Eastern, or Atlantic coast. If we look at the Primary LC product, we see that the Atlantic coast is developed land cover, unlike the south western part of the state, home to Everglades National Park, which contains 1.5 million acres of wetlands. Vegetation is likely more susceptible to spectral changes before and after storm events.
Since 1990, Meridian has experienced exponential growth, becoming the second largest city in Idaho and one of the top ten fastest growing cities in the nation coming in at #3 in 2020. The recent trends in urban development, especially zoning changes to allow for construction of higher density housing, have current residents concerned about how that affects the remaining, mostly rural community. In the 1990s to 2000, there was only one neighborhood in Meridian with more than 12 people per acre. The LCMAP primary land cover product shows an ongoing conversion from cropland to developed through time.
The late 1980s were dry in the upper plains states.  Prior to the mid 1990s, the region saw an uptick in precipitation than continued during most of the subsequent years. This change in moisture, along with higher commodity prices led to many grasslands being converted to cropland. In the animation, patches of cropland increase through time. Also, the number and size wetland and lakes increase. The expansion of central pivot irrigation are visible as round circles of cropland. These land cover changes indicate a decrease in grassland connectivity and an increase in wetland connectivity.
The spectral stability product shows the number of days that have passed since a spectral change, whether due to land cover class change or only a change in condition. As the animation shows each year through time, pixels without change will continue up the color ramp from orange at the beginning of the time series, through the yellows, and blue green. Pixels with change, due to model breaks, will be recoded to 0 days, and start the count of days of stability over.
With this long record of big data provided by the Landsat ARD, we now know that about 12% of the US has experienced a land cover change during this period, and less than 1% changes annually.  But in certain regions, the SE for example, is dominated by intense commercial timber management. This land use lends to changes in thematic land cover more often as tree cover changes to grassland/shrubland and back to tree cover through time. In contrast, the western US's change is driven by ecosystems that are severely impacted by drought conditions. The data from the 33 years of LCMAP products contribute to a better understanding of land change in the United States.
Moving forward, LCMAP will be releasing products for 2018 and 2019 in the coming months as well as Hawaii, and then Alaska. Collection 2.0 LCMAP products will be developed and available in a cloud environment.
LCMAP is establishing a Land cover Science working group. And we are looking to Expand collaborations and partnerships. To receive our latest news, please subscribe at the link provided.
LCMAP provides solutions to the science and management communities' growing need for an improved understanding of the fundamental drivers of land change, the consequences of change in human and natural systems, and feedbacks associated with land change processes.
For access to data, LCMAP Collection 1.0 science products can be downloaded by Landsat ARD tile on earth explorer OR by area of interest using the LCMAP Viewer. 
More information, as well as supporting documentation, can be found at usgs.gov/lcmap