MODIS-based Cropland Classification in North America Teki Sankey and Richard Massey Northern Arizona University.

Slides:



Advertisements
Similar presentations
Applications of one-class classification
Advertisements

INSEA biophysical modelling: data pre-processing Workshop at JRC in Ispra, Italy 11 th – 12 th April, 2005 By Juraj Balkovič & Rastislav Skalský SSCRI.
A Land Cover Map of Eurasia’s Boreal Ecosystems S. BARTALEV, A. S. BELWARD Institute for Environment and Sustainability, EC Joint Research Centre, Italy.
GLC 2000 ‘Final Results’ Workshop (JRC-Ispra, 24 ~ 26 March, 2003) GLC 2000 ‘Final Results’ Workshop (JRC-Ispra, 24 ~ 26 March, 2003) LAND COVER MAP OF.
The University of New Mexico Continual Compositing of NDVI MODIS Products for Theresa R. Watson, Richard P. Watson, Louis A. Scuderi, Enrique Montaño,
Incorporating regional knowledge into global data sets: Some ideas for rice in Asia Andy Nelson Thursday11 th Sep, 2014.
Geoprocessing with GDAL and Numpy in Python Delong Zhao
Change Detection. Digital Change Detection Biophysical materials and human-made features are dynamic, changing rapidly. It is believed that land-use/land-cover.
Global Particulate Matter (PM 10 and PM 2.5 ) Emissions from Agricultural Tilling and Harvesting 1 Northeast Institute of Geography and Agroecology, Chinese.
North American Croplands Richard Massey & Dr. Teki Sankey.
ASTER image – one of the fastest changing places in the U.S. Where??
Multiple Criteria for Evaluating Land Cover Classification Algorithms Summary of a paper by R.S. DeFries and Jonathan Cheung-Wai Chan April, 2000 Remote.
Crop Yield Modeling through Spatial Simulation Model.
CS 128/ES Lecture 5a1 Working with Rasters.
CS 128/ES Lecture 5a1 Raster Formats (II). CS 128/ES Lecture 5a2 Spatial modeling in raster format  Basic entity is the cell  Region represented.
April 28-29, 2015 at Hotel Serena, Islamabad
U.S. Geological Survey U.S. Department of Interior GFSAD30m Global Cropland Extent Products of Nominal 250 m (GCE V2.0) Updates Pardhasaradhi.
U.S. Geological Survey U.S. Department of Interior GFSAD30m Global Cropland Extent Products of Nominal 250 m (GCE V2.0) Updates Pardhasaradhi.
Ten State Mid-Atlantic Cropland Data Layer Project Rick Mueller Program Manager USDA/National Agricultural Statistics Service Remote Sensing Across the.
Mapping of mountain pine beetle red-attack forest damage: discrepancies by data sources at the forest stand scale Huapeng Chen and Adrian Walton.
Ramesh Gautam, Jean Woods, Simon Eching, Mohammad Mostafavi, Scott Hayes, tom Hawkins, Jeff milliken Division of Statewide Integrated Water Management.
North American Croplands: US Updates 21 May 2015 Richard Massey; Dr. Teki Sankey.
Proposed Modification to Method for Determining Reasonable In-Season Demand for the Surface Water Coalition: Use of the USDA Crop Data Layer Presented.
Satellite Cross comparisonMorisette 1 Satellite LAI Cross Comparison Jeff Morisette, Jeff Privette – MODLAND Validation Eric Vermote – MODIS Surface Reflectance.
U.S. Department of the Interior U.S. Geological Survey Analysis of Resolution and Resampling on GIS Data Values E. Lynn Usery U.S. Geological Survey University.
Land Cover Classification Defining the pieces that make up the puzzle.
Operational Agriculture Monitoring System Using Remote Sensing Pei Zhiyuan Center for Remote Sensing Applacation, Ministry of Agriculture, China.
ASPRS Annual Conference 2005, Baltimore, March Utilizing Multi-Resolution Image data vs. Pansharpened Image data for Change Detection V. Vijayaraj,
Chapter 21 R(x) Algorithm a) Anomaly Detection b) Matched Filter.
Current Conditions in Northern Hemisphere Crop NDVI Anomaly, August 13th, Worse than normal normal Better than normal Crop NDVI Anomaly.
Crop area estimation in Geoland 2 Ispra, 14-15/05/2012 I. Ukraine region II. North China Plain.
Selection of Multi-Temporal Scenes for the Mississippi Cropland Data Layer, 2004 Rick Mueller Research and Development Division National Agricultural Statistics.
North American Croplands Teki Sankey and Richard Massey Update: 11/20/2014.
U.S. Department of the Interior U.S. Geological Survey Mega-File Data Cube (MFDC) Concept and Preparation Jun Xiong, Prasad, Pardha Nov 21th, 2013, Flagstaff.
North American Croplands Teki Sankey and Richard Massey Northern Arizona University Flagstaff, AZ.
Vegetation Condition Indices for Crop Vegetation Condition Monitoring Zhengwei Yang 1,2, Liping Di 2, Genong Yu 2, Zeqiang Chen 2 1 Research and Development.
Understanding Glacier Characteristics in Rocky Mountains Using Remote Sensing Yang Qing.
National Economic Survey of Iraq 1 The Agriculture Survey Part 2 November 21, 2004.
North American Croplands: Updates Richard Massey; Teki Sankey.
Digital Image Processing
Supplementing Existing BP & FIL Data with Crop BP &FIL Background – Current BP & FIL grids not showing BP & FIL values in areas designated NB3 by LANDFIRE.
1 U.S. Department of the Interior U.S. Geological Survey LP DAAC Stacie Doman Bennett, LP DAAC Scientist Dave Meyer, LP DAAC Project Scientist.
Croplands.org Aiming to release to public in next couple weeks. There will likely be a separation of work space from public space.
North American Croplands Teki Sankey and Richard Massey Update: 8/21/2014.
Accuracy Assessment: Building Global Cropland Reference Data Updates for March 2015 Kamini Yadav and Russ Congalton.
Cropland using Google Earth Engine
Early Detection & Monitoring North America Drought from Space
U.S. Geological Survey U.S. Department of Interior GFSAD 30 Cropland Products of Nominal 250 m Using MODIS Data and Cropland Mapping Algorithms:
Pixel Clustering and Hyperspectral Image Segmentation for Ocean Colour Remote Sensing Xuexing Zeng 1, Jinchang Ren 1, David Mckee 2 Samantha Lavender 3.
North American Croplands Teki Sankey and Richard Massey.
U.S. Department of the Interior U.S. Geological Survey Automatic Generation of Parameter Inputs and Visualization of Model Outputs for AGNPS using GIS.
1 U.S. Department of the Interior U.S. Geological Survey LP DAAC Stacie Doman Bennett, LP DAAC Scientist.
Reference data & Accuracy Assessment Dr. Russ Congalton Kamini Yadav.
Cropland mapping in South America
Monitoring land use and land cover changes in oceanic and fragmented lanscapes with reconstructed MODIS time series R. Lecerf, T. Corpetti, L. Hubert-Moy.
Cropland Extent Mapping in South America Global Food Security - Support Analysis m Chandra Giri, Ying Zhong January 19 th, 2016.
U.S. Department of the Interior U.S. Geological Survey Chandra Giri Ying Zhong 7/15/2015 Cropland Extent Mapping in South America.
Dr. Russ Congalton & Kamini Yadav GFSAD30 Meeting, Menlo Park 19 th -21 st January, 2016 Reference Data Collection & Accuracy Assessment: Some Results.
-gSSURGO- Using the Soil Data Management Toolbox Steve Peaslee USDA-NRCS National Soil Survey Center Lincoln, Nebraska March.
Limei Ran 1, Ellen Cooter 2, Verel Benson 3, Dongmei Yang 1, Robert Gilliam 2, Adel Hanna 1, William Benjey 2 1 Center for Environmental Modeling for Policy.
Stennis Space Center Phenological Parameters Estimation Tool Presented by Jerry Gasser Lockheed Martin Mission Services John C. Stennis Space Center USDA.
Website: Croplands.org. User Authentication Full user authentication system with roles. User passwords hashed (one way encrypted) with a salt using modern.
US Croplands Richard Massey Dr Teki Sankey. Objectives 1.Classify annual cropland extent, Rainfed-Irrigated, and crop types for the US at 250m resolution.
ACCURACY ASSESSMENT SOUTH AMERICA, NORTH AMERICA KAMINI YADAV DR. RUSSELL CONGALTON.
Bijay Shrestha NASA RPC Ag-Efficiency, July 10, 2007 DATA FUSION TO PRODUCE CLOUD-FREE TEMPORAL NDVI CROSS PLATFORM COMPOSITES USING TEMPORAL MAP ALGEBRA.
ASTER image – one of the fastest changing places in the U.S. Where??
Class 10 Unsupervised Classification
الدكتور: أحمد رأفت غضية صفاء عبد الجليل كامل حمادة
Corn and Soybean Differentiation Using Multi-Spectral Landsat Data
Class 10 Unsupervised Classification
Presentation transcript:

MODIS-based Cropland Classification in North America Teki Sankey and Richard Massey Northern Arizona University

Outline Datasets chosen for processing: 2000 and 2007 Preprocessing Classification Spatial and temporal extrapolation Crop type labeling

Preprocessing workflow MODIS data Re-projection Mosaicking Layer stacking NDVIBand 1 (RED)Band 2 (NIR) Min-value 16-day Composite Max-value 16-day Composite Cloud Filtering/smoothing 47 tiles, 8-day composites (Feb 2000 to Feb 2001) (Feb 2007 to Feb 2008)

Composites Cloud-covered pixels make up for most of the noise in the data stack 16-day minimum value composites for band 1 and band 2 as cloud reflectance is higher 16-day maximum value composite for NDVI as cloud NDVI is lower NDVI maximum value composite is more useful in classification

Cloud filtering NDVI Days NDVI Days Return Value Difference Direction Return Value Difference Direction NDVI Days NDVI Days Thresholds:- Return Value < 0.20 Difference > 0.15

Cloud filtering NDVI Days NDVI Days Difference NDVI Days NDVI Days Thresholds:- Difference > 0.15

Cluster computing workflow The NAU computing cluster has 32 cores each with 500 nodes, shared memory of 1.5 TB per node ENVI services engine and ENVI version 5.1 Master C Program Batch file for execution IDL code for each process Batch file for IDL process Node Allocation IDL parallel process

NDVI stack

Year 2007 No existing crop type map for classification needs labeling 2007 = 2000 in region-wise annual precipitation NASS CDL available for year 2007 Assumption: Similar spectral signatures between the two years

Region-wise annual precipitation statistics US ( , National Climatic Data Center)

NASS CDL availability for conterminous US

Spatial and Temporal Extrapolation USA  North America 2007  2000

Spatial extrapolation: GCE v1.0 Most accurate irrigated class = AOI-1 (4/4 maps) ( 63,102,129 acres) Further split: Agro-Ecological Zones

Agro-Ecological Zones based on length of growing period (GAEZ-FAO)

GCE v1.0 AOI-1 and Agro-Ecological zones

MODIS-based US Irrigation map, 2001 (Ozdogan and Gutman, 2008)

GCE v1.0 AOI-1, Agro-Ecological Zones, and Irrigated map

Irrigated map 2001 (US) GCE v1.0 Class1GCE v1.0 Class3 AEZ 1AEZ 2 AEZ 3 AEZ 14 AEZ 1AEZ 2AEZ 3 AEZ 14 ISODATA Classification Overlay …… Irrigated map 2001 (US) GCE v1.0 AOI-1GCE v1.0 AOI-3 AEZ 1AEZ 2 AEZ 3 AEZ 14 AEZ 1AEZ 2AEZ 3 AEZ 14 Class 1Class 2 Class 25 Class 1Class 2 Class 25 Group 1 Group 2 Group 3 Group 4Group 5 Group 6 Group 100 ISODATA Classification Class Grouping Overlay ……….. …… …… Spatial extrapolation: Spatial subsets

Spatial extrapolation

Spatial extrapolation: Labels Spectral correlation matrix Classes are grouped together (R 2 > 0.98)

Temporal extrapolation: 2007  AOI + Irrigated map + Agro Eco zone2007 NASS CDL 2007 AOI + Agro Eco zone Labeling of classes using NASS CDL 2007

Master-file Primary layers – Cropland extent – Crop type – Crop intensity – Irrigated/Rainfed Secondary layers – Temperature – Precipitation – Elevation AttributeNameValue Cropland ExtentNon-Cropland0 Cropland1 Irrigated/RainfedRainfed0 Irrigated1 Crop TypeNon-Cropland0 Wheat1 Rice2 Corn3 Barley4 Soybean5 Pulses6 Potatoes7 Cotton8 Others9 IntensityNo crop0 Single crop1 Double crop2 Double+ crop3 TemperatureCelsius-value- PrecipitationCentimeters-value- Elevationmeters-value- NASS CDL 2007 MIrAD US 2007 NCEP NARR 2007 SRTM DEM

Spectral database Isodata classification for each AOI Class comparison with master-file Group classes based on attributes Group member classes lie in ± 0.1 buffer of the group mean spectra for more than 80% of bands Spectral database for each attribute combination Corn: Irrigated, Single crop Wheat: Irrigated, Single crop Soybean: Irrigated, Single crop

Spectral database AttributeNameValue Cropland ExtentNon-Cropland0 Cropland1 Irrigated/RainfedIrrigated0 Rainfed1 Crop TypeNon-Cropland0 Wheat1 Rice2 Corn3 Barley4 Soybean5 Pulses6 Potatoes7 Cotton8 Others9 IntensitySingle crop0 Double crop1 Triple crop2 Triple+ crop3 TemperatureCelsius-value- PrecipitationCentimeters-value- Elevationmeters-value- Class IDBand 1Band 2………Band N Class ………0.60 Class ……… ……… ……… ……… ………0.61 Class M ………0.63 Cropland attributes Grouped classes for one set of attributes

Extrapolation rules: Correlation Spectral match between classes in 2007 and 2000 ISOdata classification 2007 result ISOdata classification 2000 result

Extrapolation rules: Buffer If the input spectra lies within ± 0.1 buffer of the database spectra for more than 80% of bands it is assigned the same label If secondary parameters indicate Drought or Abundance, buffer is adjusted accordingly Overall validation threshold: 90% Buffer

Extrapolation Generation of NDVI stack for non-US region Spatial extrapolation of labels to non-US region using updated spectral database Input spectra is assigned the same label if lies within ± 0.1 buffer of the database spectra for more than 80% of bands it Verification of extent using GCE v1.0 and secondary parameters

Class Labels Irrigated map 2000 (US) GCE v1.0 Class1GCE v1.0 Class3 AEZ 1AEZ 2 AEZ 3 AEZ 14 AEZ 1AEZ 2AEZ 3 AEZ 14 Class 1Class 2 Class 25 Class 1Class 2 Class 25 ISODATA Classification Class Grouping and Labeling: 2007 Overlay Label 1 Label 2 Label 3 Label 4Label 5 Label 6 Label 100 Temporal Extrapolation: 2007  2000 ……….. …… …… Cropland map 2000 (US) Cropland map 2007 Cropland map 2000 (North America) Spatial Extrapolation: US  non-US

Thank you!

Class labeling 2007 classes are labeled by geolocating at least 10 random points within the 2007 CDL class Classes between 2000 and 2007 are matched together via correlation (R 2 >0.98) Labeled as crop type ‘A’

Spatial and temporal extrapolation US classesNA_class1NA_class2NA_class3NA_class4NA_class5NA_class6NA_class7NA_class8NA_class9NA_class10NA_class11NA_class12NA_class13NA_class14NA_class15NA_class16NA_class17NA_class Spectral correlation matrix

Validation Preprocessing of MODIS data for validation year (2009) Generation of NDVI stack Generation of validation-file using spectral database Validation-file has same structure as master-file of normal year (2008) Comparison of NASS CDL for 2009 with the validation file