lcmap workshop welcome - amazon web services · v.3.0 (5/18/2016) national base sample of 25,000...

32
Jesslyn F. Brown USGS/EROS Nov 6, 2018 LCMAP Workshop Welcome

Upload: others

Post on 08-Sep-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

Jesslyn F. Brown

USGS/EROS

Nov 6, 2018

LCMAP Workshop Welcome

Page 2: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

LCMAP: Workshop Goals

▪ Land Change Monitoring, Analysis, and Projection (LCMAP)

▪ Provide an update on current status (Version 1)

▪ Algorithm approaches

▪ Analysis Ready Data used in LCMAP

▪ Data Products

▪ New analysis tool: TapTool

▪ Validation

▪ Assessments

▪ Example Application

▪ Get the word out to potential users

▪ Allow interaction with prototype data

▪ Collect feedback

Page 3: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

Nov 6-8, 2018 workshop schedule

Tues, Nov. 6

Morning (Webinar)

Welcome

What is LCMAP?

Lessons Learned using ARD in LCMAP

Break

Data and Product Examples

Q&A

Afternoon (On-site)

LCMAP Visualization

Hands-on Data Exploration

Wed, Nov. 7

Morning (Webinar)

Using LCMAP Data to Inform Chesapeake BayRestoration

LCMAP Accuracy Assessment

Break

Land Change in the Piedmont Ecoregion

Q&A

Afternoon (On-site)

Hands-on Exploration

Wrap-up Session

Collect Feedback

Page 4: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

LCMAP: The team and a little history

EROS TeamLoveland Tollerud Jenkerson Barber Taylor

Brown Auch Briden Zhou Austin

Gallant Sohl Reker Smith Hill

Vogelmann Rover Pengra Zelenak

Dwyer Sayler Horton Davis

Labahn Xian Bouchard Docktor

University PartnersBoston University

Texas Tech University

State University of New York

Forest Service PartnersCohen

Healy

Schroeder

Yang

Page 5: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

What is LCMAP?

Jesslyn F. Brown

USGS/EROS

Nov 6, 2018

Page 6: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

LCMAP Vision▪ Provide documentation and understanding of historical land change and contemporary land change

as it occurs. Provide ongoing answers to questions on where, how, and why the landscape is

changing.

▪ Explain how past, present, and future land change affects society, natural systems, and the

functioning of the planet. Assess the impacts of land change locally, regionally, and

nationally. Topical emphases include land-change impacts on weather and climate, the carbon

cycle, water resources, and ecosystem functioning.

▪ Alert relevant stakeholders to important or emerging patterns of land change in their jurisdictions.

▪ Support users in the application of land-change data, information, and science results. This includes

a state-of-the-art applications support capability, aggressive communications and outreach, and

web-based capabilities for accessing all products. Provide “webinars” to explain and share land-

change products and information.

Page 7: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

LCMAP: The next generation

▪ Land Change Monitoring, Assessment, and Projection (LCMAP)

▪ Builds on rich USGS EROS history of LC R&D

▪ At higher resolutions

▪ Integrated land cover and change

▪ Targets demand for land cover products

▪ Over larger geographic extents

▪ Over longer time periods

▪ With faster delivery

▪ Foundational elements include:▪ Landsat Analysis Ready Data (ARD)

▪ Land change and land cover derived from all available Landsat data

▪ Independent reference data (accuracy assessment)

Page 8: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

Basic

vision

and

structure

of LCMAP

Page 9: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

ARD: Analysis Ready Data

▪ Reconditioning of theLandsat archive

▪ Based on Collection1, Tier 1 TM-OLIimagery

▪ Radiometric processing to TOA, surfacereflectance, & brightness temperature

▪ All available high-quality observationspackaged in 150 x 150 km tiles

▪ USGS/EROS began distribution of CONUS ARDin October 2017

(adapted from Dwyer et al., Pecora20)

Page 10: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

Inputs matter: ARD per-pixel data richness

Estimated from Level-1 metadata

▪ “Per-pixel” datarichness 1982-2017

Doesn’t reflect impact of

ETM+ SLC-off data gaps

Page 11: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

LCMAP: Mapping & Monitoring

▪ Continuous Change Detection & Classification (CCDC)

▪ Zhu & Woodcock (2014)

▪ Based on analysis of the time series of available cloud/shadow free Landsat observations L4-L8 (1982-present)

▪ Change detection capability (CCD):

▪ Land cover conversion

▪ ‘Conditional’ changes (e.g. forest thinning, insect outbreaks)

▪ Potential for near real-time change detection capability

Zhu, Z.; Woodcock, Curtis E. (2014). Continuous change

detection and classification of land cover using all available

Landsat data. Remote Sensing of Environment, 144, 152-171.

Page 12: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

LCMAP Monitoring using CCDCMethods that allow tracking of gradual and abrupt change over time

Crop f

Date

Landsat Mid-infrared Band

1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

Year fire occurred

Post-fire recovery

Longer-term recovery

Pre-fire

Page 13: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

CCDC Methodology▪ Change monitoring and land cover mapping is based on the use of

the Continuous Change Detection and Classification methods developed by Boston University researchers (Zhu and Woodcock, 2014). CCDC development was funded through a USGS Landsat Science Team award.

▪ All cloud- and shadow-free observations Landsat 4-8 data (1982-present) are used. NLCD data are used for training the land cover classification.

Calculate time segments for each pixel and determine breaks

(changes)

Extract spectral change and classify the time

segment into products

Using xgboost, derive the thematic info and

generate maps

Page 14: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

LCMAP Products: Ten Land Cover and Spectral Change and Products

Page 15: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

LCMAP: Annual output productsName Description Note

Land Cover ProductsPrimary Land Cover Land cover classification consisting of nine general land cover Classes: Developed, Cropland, Tree Cover, Grass/shrub, Wetland,

types Water, Ice/Snow, Barren

Secondary Land Cover Land cover classification consisting of nine general land cover Alternative cover class with the 2nd highest algorithmic probability,

types same as primary LC classes

Primary Land Cover Confidence A measure of confidence in the primary land cover class Measure from 0 to 100; higher values imply higher levels of

designation confidence

Secondary Land Cover Confidence A measure of confidence in the secondary land cover class Measure from 0 to 100; higher values imply higher levels of

designation confidence

Annual Land Cover Change Indictor of thematic land cover change that has occurred from the Categories that indicate two land cover states, the prior year land

prior year to the current year (in other words, From:To land cover and the current year land cover

cover)

Spectral Change ProductsTime of Spectral Change The day that a spectral change was detected for a given year Provides data indicating location and timing of spectral breaks

Change Magnitude A measure of the spectral magnitude of the change found within Higher values may be indicative of different types of changes

a given year

Time Since Last Change The cumulative number of days since the last spectral change Provides an indicator of how long a location has been in its current

1stoccurred as of July of the given year state

Spectral Stability Period A measure of the amount of time in days that a pixel has been Provides information of the frequency of land surface change

spectrally stable

Model Quality Characterization of time series model quality as it relates to Provides a spatial measure for interpreting LCMAP product results

model input data and model fit

Page 16: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

LCMAP: Version 1 prototype results

DevelopedAgriculture

Grassland & Shrubland

Tree CoverWater

Wetlands

Snow & Ice

Barren

▪ Cascades / Columbia Plateau

Primary Cover, 2014

300 km

Page 17: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

Version 1 Prototype Tiles

▪ Processing in prototype mode to provide data for initial QA/QC

▪ Eval sites

▪ Eastern region

Page 18: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

Visualization of CCDC2010

2016

Page 19: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected
Page 20: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

Basic

structure

of LCMAP

Page 21: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

USGS/USFS Validation Collaboration

Joint USGS/USFS Response Design

V.3.0 (5/18/2016)

National Base Sample of

25,000 pixels

Time Sync V3

5 variables x 30 years x 25,000 plots = 3,750,00 manually collected data points

The full set of variables are then cross-walked and aggregated to create output comparable to the LCMAP

legend at each plot for each year of the time series (1985-2016). 750,000 data points – LCMAP Legend

Page 22: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

Independent Reference Data

▪ Joint effort with US Forest Service Landscape Change Monitoring System (LCMS)

▪ LCMAP validation & error estimation

▪ Initial sample set: 25,000 points

▪ Spatial extent: CONUS

▪ Characterized across 1985-present

▪ Full TM/ETM+/OLI record

▪ Evaluated for land cover & change

▪ Full Landsat record, high res, and ancillary data

▪ Can be updated ongoing, and/or intensified

(adapted from Pengra et al., Pecora20)

Page 23: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

Assessments and Projections

LCMAP ultimately is a capability to continuously track and characterize

changes in land cover, use, and condition and parlay such information into

assessments of current and historical processes of change as a science

foundation to support evaluations and decisions relevant to environmental

management and policy.

Channel dynamics

Energy development

Water quality

Drought monitoringPollinator

landscapes

Wet / dry cycles

Page 24: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

Prototype Assessment of Land Change in the Piedmont

Page 25: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

Modeling Landscapes Beyond the Satellite Era

HistoricalContemporary

(Satellite Era)

Future

Scenarios

Consistent USGS Land-cover Database

Modeled

(Backcast)Modeled

NLCD

CCDC

Trends

Page 26: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

Projections▪ Use of CCDC output to inform FORE-SCE:

▪ Model scenario construction

▪ Short-term “business-as-usual” scenarios, based on simple extrapolation from most recent LCMAP output

▪ Transition probabilities to inform alternative scenarios

▪ Model parameterization▪ “Elasticities” (based on transition probabilities)

▪ Patch characteristics

▪ Spatial distribution

▪ Model Validation / Assessment

Page 27: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

LCMAP: Applications Assistance

Communicate and engage with LCMAP user community to:

▪ Grow existing and initiate new partnerships

▪ Align applications with partner requirements

▪ Design cooperative and innovative applications

▪ Develop examples/scenarios of ARD and LCMAP applications

▪ Provide feedback related to future R&D directions

Page 28: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

LCMAP: CONUS engineering and production schedule

▪ Version 1 Production is our next big goal: 422 tiles, 10 products for 33 years ~140,000 map products

▪ Have locked down our methodology to produce annual products

▪ Transferred science code to engineering team

▪ Code is being tested on operational system

▪ Starting production November, 2018

▪ Production estimate is approximately 5 months

▪ ~2 months are needed to perform the accuracy assessment

Page 29: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

Notional Schedule and Key Milestones

▪ IOC will ‘operationalize’ mapped products derived from CCDC models and US Landsat ARD

▪ Released products verified with LCMS reference data

▪ Released products will be available on Earth Explorer

▪ IOC includes local Science R&D access to IWDS & SEE

▪ Broader/remote access methods under study

US ARD Released CU Ingested into IWDS US Ingest

CCDC SW Beta

L-C/C Product Gen SW Beta

US CCDC Runs Done

First Products Released

CCDC IOC Complete

Ongoing Studies/Plans/Dev

Sentinel-2 ARD

MultiSpectral Scanner ARD

Landsat Collection-2 Global ARD

Science R&D Access

CY17 CY18 CY19

CYQ4CYQ1 CYQ2 CYQ3 CYQ4 CYQ1 CYQ2 CYQ3 CYQ4 CYQ1 CYQ2 CYQ3

Page 30: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

Sample Data

▪Prototype results

▪All 10 annual products

▪ 1986-2017 (33 years)

▪H10V03 (Southern CA)

▪Expected to be available by end of 2018 with draft metadata and a product guide

Page 31: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

Research and Development

▪ Improve and further systematize change detection

▪ Near-real-time change detection

▪ Research approaches for utilizing different wavelengths and instruments

▪ Refine land cover legend in classification

▪ Research approaches for defining the effect of weather on change detection

▪ Further improve classification methods

▪ Develop plan for updates in subsequent years (annual forward processing)

▪ Research approach to change attribution

Page 32: LCMAP Workshop Welcome - Amazon Web Services · V.3.0 (5/18/2016) National Base Sample of 25,000 pixels Time Sync V3 5 variables x 30 years x 25,000 plots = 3,750,00 manually collected

LCMAP: For more information

▪ Existing web page is being reworked:

https://eros.usgs.gov/science/land-change-monitoring-assessment-and-projection-lcmap

▪ USGS EROS Center Customer Service:

▪ Email: [email protected]

▪ Phone: 1-800-252-4547

▪ Social media: USGS/EROS twitter and facebook