For orest est Mon Monitor itorin ing g For or Action Action - - PowerPoint PPT Presentation

▶
for orest est mon monitor itorin ing g for or action
SMART_READER_LITE
LIVE PREVIEW

For orest est Mon Monitor itorin ing g For or Action Action - - PowerPoint PPT Presentation

Jun 28, 2023 236 likes •1.48k views

For orest est Mon Monitor itorin ing g For or Action Action David Wheeler, Center for Global Development Robin Kraft and Dan Hammer, University of Maryland Scipy 2010 rkraft@umd.edu cgdev.org/forest Forest Monitoring for Action (FORMA)

slide-1
SLIDE 1

For

  • rest

est Mon Monitor itorin ing g For

  • r Action

Action

David Wheeler, Center for Global Development Robin Kraft and Dan Hammer, University of Maryland Scipy 2010 rkraft@umd.edu cgdev.org/forest

slide-2
SLIDE 2

Forest Monitoring for Action (FORMA)

  • Set of algorithms to identify deforestation

using satellite imagery quickly and cheaply Outline

  • Why deforestation?
  • What is FORMA?
  • Python and FORMA
slide-3
SLIDE 3

Caveats

  • Economist/geographer, not a programmer
  • Parallel computing broadly speaking – not

necessarily in robust compsci terms

slide-4
SLIDE 4

Why do we care about deforestation?

  • Biodiversity
  • Local and regional environment
  • Climate change
slide-5
SLIDE 5

How do we deal with deforestation?

  • Address economic drivers
  • Monitor outcomes
slide-6
SLIDE 6

Monitoring: Issues with traditional methods

  • Slow turnaround
  • Relatively opaque methodologies
  • Data black box
  • Local accuracy vs. wide coverage
  • Needed software tools expensive and hard to use
  • Research aimed at scientists more than people on

the ground.

  • BUT – it’s still really cool stuff!
slide-7
SLIDE 7

Forest Monitoring for Action

  • Set of algorithms to identify deforestation using rapidly

updated, free satellite imagery.

  • Prototype: monthly maps at 1km resolution
  • Early-warning system to complement hi-res approaches
  • Special features

– Rapid (monthly) updates: 2006-present – Potential pan-tropical application – Indonesia-wide prototype is unique

slide-8
SLIDE 8

The (very) basic intuition

  • Forested areas look green
  • Forest + fires + browning = deforestation?
  • FORMA algorithms search for telltale patterns
  • f fires and “browning”
slide-9
SLIDE 9

FORMA Methodology

  • Identify relevant patterns and time trends in

“greenness” and fires

  • Train algorithms using historical data (2000-5)
  • Apply parameters from training to subsequent

data

  • Output: probability of forest clearing, by pixel
slide-10
SLIDE 10

Original workflow for Indonesia: Doesn’t scale

  • Two dual core Windows desktops
  • One 1tb hard drive that got quite full
  • One ArcGIS license ($3000) for GIS data and

imagery

  • One Stata license ($1500) for statistical modeling
  • Python 2.5 as glue
  • Start to finish: 4 weeks
  • Start to finish on 4 cores: ~1+ week
slide-11
SLIDE 11

Pan-tropical application @ 250m resolution

  • ~130x more raw data
  • Data comes split into pieces – thanks NASA!
  • “Easy” to parallelize if you can get rid of pricey

software and manage all the data …

slide-12
SLIDE 12

New workflow: Homebrew “Map/Reduce” in the cloud

  • Amazon Web Services (AWS):

– EC2: small Linux spot instances – approx. $0.06/hr – SQS: fault tolerant job management – S3: persistent storage for intermediate and output data

  • Boto: interact with AWS
  • GDAL/OGR: load images as Numpy arrays; re-project images; GIS

data management

  • Numpy:

– Slice images into pieces, stack slices as time series in tabular format – Statistical modeling by pixel, pixel neighborhood or ecoregion means plenty of room for parallelization

slide-13
SLIDE 13

The cloud/FOSS difference for Indonesia prototype

  • 20 Linux instances:

– 10x the compute power for $1.20/hour, with no software costs

  • Easily scalable with SQS – just re-run queue processing script on

new instance

  • Per-image processing significantly faster – Numpy arrays are great!
  • Parallel pre-processing in the cloud

– RUNNING FOR THE FIRST TIME THIS MORNING! – Preprocessing time drops from from 3 days to <5 hours

  • Allows us to experiment with different algorithms and data much

more easily

slide-14
SLIDE 14

Future plans

  • Matplotlib + basemap for visualization
  • Numpy to mask out water and non-forest

pixels

  • PiCloud? StarCluster?
slide-15
SLIDE 15

Maps of forest clearing: monthly time series for Riau, Indonesia

slide-16
SLIDE 16
slide-17
SLIDE 17
slide-18
SLIDE 18
slide-19
SLIDE 19
slide-20
SLIDE 20
slide-21
SLIDE 21
slide-22
SLIDE 22

270 km 167 mi

Forest in 2000

Forested in 2000

slide-23
SLIDE 23

270 km 167 mi

Cleared by 2005

Forest in 2000 Cleared 2000 - 2005

slide-24
SLIDE 24

What happened after 2005?

slide-25
SLIDE 25

Cleared 2000 - 2005 Forest in 2000 60 – 70% 50 – 60% 70 – 80% 80 – 90% > 90% Probability

270 km 167 mi

Cleared by 10/2009

slide-26
SLIDE 26

Forest in 2000 Cleared 2000 - 2005

11/2005

slide-27
SLIDE 27

12/2005

slide-28
SLIDE 28

1/2006

slide-29
SLIDE 29

2/2006

slide-30
SLIDE 30

3/2006

slide-31
SLIDE 31

4/2006

slide-32
SLIDE 32

5/2006

slide-33
SLIDE 33

6/2006

slide-34
SLIDE 34

7/2006

slide-35
SLIDE 35

8/2006

slide-36
SLIDE 36

9/2006

slide-37
SLIDE 37

10/2006

slide-38
SLIDE 38

11/2006

slide-39
SLIDE 39

12/2006

slide-40
SLIDE 40

1/2007

slide-41
SLIDE 41

2/2007

slide-42
SLIDE 42

3/2007

slide-43
SLIDE 43

4/2007

slide-44
SLIDE 44

5/2007

slide-45
SLIDE 45

6/2007

slide-46
SLIDE 46

7/2007

slide-47
SLIDE 47

8/2007

slide-48
SLIDE 48

9/2007

slide-49
SLIDE 49

10/2007

slide-50
SLIDE 50

11/2007

slide-51
SLIDE 51

12/2007

slide-52
SLIDE 52

1/2008

slide-53
SLIDE 53

2/2008

slide-54
SLIDE 54

3/2008

slide-55
SLIDE 55

4/2008

slide-56
SLIDE 56

5/2008

slide-57
SLIDE 57

6/2008

slide-58
SLIDE 58

7/2008

slide-59
SLIDE 59

8/2008

slide-60
SLIDE 60

9/2008

slide-61
SLIDE 61

10/2008

slide-62
SLIDE 62

11/2008

slide-63
SLIDE 63

12/2008

slide-64
SLIDE 64

1/2009

slide-65
SLIDE 65

2/2009

slide-66
SLIDE 66

3/2009

slide-67
SLIDE 67

4/2009

slide-68
SLIDE 68

5/2009

slide-69
SLIDE 69

6/2009

slide-70
SLIDE 70

7/2009

slide-71
SLIDE 71

8/2009

slide-72
SLIDE 72

9/2009

slide-73
SLIDE 73

10/2009

slide-74
SLIDE 74

Cleared 2000 - 2005 Forest in 2000 60 – 70% 50 – 60% 70 – 80% 80 – 90% > 90% Probability

270 km 167 mi

Cleared by 10/2009

slide-75
SLIDE 75

12/2005

slide-76
SLIDE 76

1/2006

slide-77
SLIDE 77

2/2006

slide-78
SLIDE 78

3/2006

slide-79
SLIDE 79

4/2006

slide-80
SLIDE 80

5/2006

slide-81
SLIDE 81

6/2006

slide-82
SLIDE 82

7/2006

slide-83
SLIDE 83

8/2006

slide-84
SLIDE 84

9/2006

slide-85
SLIDE 85

10/2006

slide-86
SLIDE 86

11/2006

slide-87
SLIDE 87

12/2006

slide-88
SLIDE 88

1/2007

slide-89
SLIDE 89

2/2007

slide-90
SLIDE 90

3/2007

slide-91
SLIDE 91

4/2007

slide-92
SLIDE 92

5/2007

slide-93
SLIDE 93

6/2007

slide-94
SLIDE 94

7/2007

slide-95
SLIDE 95

8/2007

slide-96
SLIDE 96

9/2007

slide-97
SLIDE 97

10/2007

slide-98
SLIDE 98

11/2007

slide-99
SLIDE 99

12/2007

slide-100
SLIDE 100

1/2008

slide-101
SLIDE 101

2/2008

slide-102
SLIDE 102

3/2008

slide-103
SLIDE 103

4/2008

slide-104
SLIDE 104

5/2008

slide-105
SLIDE 105

6/2008

slide-106
SLIDE 106

7/2008

slide-107
SLIDE 107

8/2008

slide-108
SLIDE 108

9/2008

slide-109
SLIDE 109

10/2008

slide-110
SLIDE 110

11/2008

slide-111
SLIDE 111

12/2008

slide-112
SLIDE 112

1/2009

slide-113
SLIDE 113

2/2009

slide-114
SLIDE 114

3/2009

slide-115
SLIDE 115

4/2009

slide-116
SLIDE 116

5/2009

slide-117
SLIDE 117

6/2009

slide-118
SLIDE 118

7/2009

slide-119
SLIDE 119

8/2009

slide-120
SLIDE 120

9/2009

slide-121
SLIDE 121

10/2009

slide-122
SLIDE 122

Cleared 2000 - 2005 Forest in 2000 60 – 70% 50 – 60% 70 – 80% 80 – 90% > 90% Probability

270 km 167 mi

Cleared by 10/2009

slide-123
SLIDE 123

Forest Monitoring for Action

Coming soon to a tropical forest near you

Maps and data available at cgdev.org/forest FORMA@cgdev.org rkraft@umd.edu

A tool from the Center for Global Development