Data Issues in the Miombo Network

1
Data Issues in the Miombo Network

• Paul V. Desanker (desanker_at_virginia.edu)
• IGBP/START/LUCC Miombo Network

2
Outline

• Who we are
• Sample activities and use of IT (RS, GIS, spatial
  modeling)
• Data bundling
• Lcluc mapping, modeling
• Monitoring and assessment WoodCAS
• Lessons
• Discussion of challenges from a developing
  country perspective

3
The Miombo Region, based on Whites 1983
Vegetation Map of Africa
Broad vs fine leaf savana Poor soil vs rich
4
Slash and Burn Agriculture
5
Woodland is Slashed and Burnt for Agriculture,
Fields Coppice Profusely
6
Different States of Miombo from intact woodland
to tree-grass mixtures, in response to fire
7

The Miombo Network is a regional alliance of
researchers working on land use and land-use
change, under the auspices of the IGBP/IHDP Land
Use and Cover Change (LUCC) Project and the
IHDP/IGBP/WCRP Global Change System for Analysis,
research and Training (START) Members in Africa
include government, university and research
institutions in DRC, Malawi, Mozambique,
Tanzania, Zambia and Zimbabwe. Elsewhere, members
include Universities, research institutions and
NGOs in the U.S., Europe and Australia. Main
sponsors include NASA, WWF, START, and NSF.
8
History

• 1995 Initial planning meeting, and
• Development of a joint science agenda under the
  central theme of land use and land-use changes
  and their consequences, available as IGBP Report
  41 (also on the web at http//miombo.gecp.virginia
  .edu).
• Several projects developed under NASA LCLUC,
  NOAA, WWF and series of modeling and integrated
  analysis discussions under NSF funding
• Annual meetings and training workshops (modeling,
  RS/GIS, assessment).

9
edited by Desanker, Frost, Scholes, Justice
(1997), Available from IGBP in Stockholm, START,
or from the authors, can request via the Miombo
Web Page at http//miombo.gecp.virginia.edu
10
Process study of Land Use and major drivers of
change including fire, a major component of the
Miombo Network science portfolio. Also impacts
of land use change, and global change
implications (trace gas and aerosol emissions,
ecosystem dynamics, goods and services) Capacity
Building
11
Ecological Transitions in Miombo
fire
More fire
Slash burn
abandon
degradation
12
Examples of 15 day fire counts
for Africa determined from AVHRR
13
Data Bundles Improving Access to Data
14
The Miombo CD Project
The first Miombo Network meeting in Zomba, Malawi
December 1995 identified data availability as a
major constraint for development of a strong
scientific agenda, and in general, for global
environmental change research in Central and
Southern Africa. The Miombo CD project was then
designed to make data widely available to Miombo
Network Projects and more generally, to regional
scientists. Data were collated from various
international data archives and from individuals,
and processed to be accessible using a web
browser on a CD-ROM (this was 1995-1996).
15
(No Transcript)
16
CD Design Data Integration
Specific Tools
DEM
Boundaries
Internet
Climate
M i r a M o n
Vegetation
GIS
LU-LC
NDVI
Local Sources

Population
Country Wide
01011110
11100101
17
Tools to Access Data on the CD
Web Browser
MiraMon GIS
Adobe Acrobat
MiraDades
Display Zoom Query by loc. attribute Print
Digitization Import/Export Analysis
18
Viewing the CD from a single starting point


Index.htm
Web Browser
Overview and Userguide
Graphic Materials
Helping tours
Documents
Tools
Databases
Refs
IGBP41
VU
Photos
MM
GIS
auto
Database descriptions
MiraMon GIS

Graphical and tabular data
( MiraDades)
Country Wide
DEM
Climate
Vegetation
NDVI
LU-LC
Population
Boundaries
19
Postprocessing Data from the CD
DEM
Boundaries
Export
Climate
Raster IMG
Idrisi
IMGIdr
Vegetation
GIS
LULC
NDVI
Original Data
Arc/Info ArcView
Vector E00
Population
Country Wide
20
New Spatial Data Bundle for Miombo Region, to
be available online and CD-ROM in progress
21
The Miombo Global Observation of Forest Cover
GOFC Projectfunded by NASA LCLUC
22
Objectives

• 1) Mapping the miombo region using Landsat 5 and
  7 data by working in conjunction with Southern
  African national mapping agencies harmonize
  existing maps and fill gaps
• 2) Measurement of carbon densities in
  representative land cover/forest cover types
• 3) Outline a carbon accounting model for 1990 and
  2000 and links to dynamic land use change
• 4) Development of a regional spatial database for
  site characterization and
• 5) Development of an information management
  system that will distribute satellite data for
  the miombo region

23
Data Distribution to Stakeholders

• Distribution of Landsat data on CD to national
  reps each country has copies of ALL data
  available in the Miombo Network archive for their
  region(total of gt300 scenes, mostly older data)
• Online access to data still not possible for most
  countries in region
• Acquisition of Landsat 5 and 7 (see map) ongoing.

24
Miombo Landsat 5 and 7 Data Archive
(Also have 1990 Landsat 5 data for area boxed
from Earthsat)
25
Regional Mapping

• Availability of Landsat TM in 1990s made
  possible national land cover mapping
• Each country used own classification schemes,
  mostly based on visual interpretation of imagery
• Final maps in map units, problems of joining into
  national products (hardware limitations then), or
  regional products (different legends)

26
Data Sources for Regional Mapping
27
Miombo Regional Classification System for
Vegetated Land Covers (after Zimbabwe System)
Height gt15m 15-5m 5-1m lt1m Canopy
Cover 100-70 Nat Forest 100-70 Forest Forest
Forest Plantation Plantation Plantation
70-20 Woodland Bushland 20 -
2 Wooded Wooded Bushland Grassland Grassland Sa
vanna lt2 Grassland Grassland Grasland
Extensive knowledge of local landscapes greatly
improve ability to Identify plantation versus
woodland versus bush/shrub
28
Joined from 64 map units. Map based on1995
Landsat TM Working with different sources of
maps a major hassle
29
Mozambique - Raw ETM 1999
Final Map
Detail lost through mapping to correspond to
paper map applications
30
(No Transcript)
31
Product for the 1990-5 period showing
Forest/Non-Forest Classes based on harmonized
legend
32
(No Transcript)
33
New Estimates of Land Cover basedon Regional Map
for 1992
34
UMD (DeFries et al.) and Miombo Regional Legends
Miombo Classes 1 Natural Forests 2.
Plantations 3. Woodlands 4. Bushlands/thickets
5. Wooded grassland 6. Grassland/Dambos 7.
Barren areas 8, Water 9. Swamps and
Marshes 10. Cultivated Land 11 Builtup area
UMD Classes 0 WATER (and Goode's
interrupted space) 1 EVERGREEN
NEEDLELEAF FOREST 2 EVERGREEN
BROADLEAF FOREST 3 DECIDUOUS
NEEDLELEAF FOREST 4 DECIDUOUS
BROADLEAF FOREST 5 MIXED FOREST
6 WOODLAND 7 WOODED
GRASSLAND 8 CLOSED SHRUBLAND
9 OPEN SHRUBLAND 10
GRASSLAND 11 CROPLAND 12
BARE GROUND 14 URBAN AND BUILT-UP
35
Mapping with SAfMA
36
Miombo Contribution to Millennium Assessment in
S/Africa
37
WCMC Wetlands Map over Landsat Mosaic (right)
Kafue Flats clear need to generate improved
wetlands extent
38
Dambos and Wetlands of Central Malawi
Kasungu/Lilongwe area wetland map too coarse and
misses out elaborate dambo system, also lacks
temporal changes
39
Dambos and Wetlands of Central Malawi
Kasungu/Lilongwe area
40
(No Transcript)
41
(No Transcript)
42
1990 Landsat Mosaic - Raw
43
Modeling Land Use Change
44
MELTmiombo ecosys land transformation model

• Partitions landscape into admin boundaries called
  Traditional Authorities TA (several villages)
  as land use decisions constrained within these
  (for public land use) (Communities in previous
  diag)
• At national level, land distributed to major land
  uses to reflect broad population needs for food,
  fuel and for conservation efforts (including
  protection)
• Processes at different levels operate at their
  own timeframe (10-year for national planning,
  annual for landscape level)
• Demand for food major driver at local level for
  land needs, constrained by available land within
  TA, moderated by land production potential
  (function of soil, climate as well as inputs

45
MELT Miombo Ecosystem Land Transformation Mode
l Sample run from Bare in Malawi
using Traditional Authority admin boundaries
46
Landsat TM mosaic for 1990 showing agric land use
in purple
47
(No Transcript)
48
WoodCASWoodland Carbon Accounting System(in
planning)
49
Goal

• Outcome of Miombo GOFC pilot
• The overall aim is to provide a complete and
  systematic monitoring of land cover changes and
  an accounting of ecosystem carbon in support of
  assessments, global environmental conventions
  (Climate Change, Biodiversity and
  Desertification) and natural resources
  management of the vast area of dry forest and
  woodland of Southern Africa.

50
Objectives

• Establish a regional observatory of land
  cover/land use change (LCLUC) that includes a set
  of Landsat MSS, TM and ETM, to provide a basis
  for collaborative assessment of land cover change
  at the regional level (we are calling this system
  WoodCAS woodland carbon accounting system).
• Develop baseline land cover products for major
  gaps identified for the Southern Africa region,
  as well as regional products that highlight hot
  spots of land use change and major drivers of
  change.
• Collate forest inventory data from forest
  plantations in the region as well as soil carbon
  information as basis for a carbon accounting
  model and system.
• Implement carbon models based on CENTURY for
  belowground carbon and biomass-based approaches
  for aboveground carbon (VYTL) for use in
  predicting spatially explicit carbon accounts and
  potential for carbon improvement, for the
  southern Africa region.
• Develop a decision support tool that would be
  accessible via the web, and would enable
  exploration of carbon projects in landscapes of
  the region.

51
Information system design for the dry forest
LCLUC and carbon accounting system
52
Design of the Decision Support System for Carbon
Accounting via the Web
53
Decision support Models for Carbon to be
accessible via Internet

• Use state of the art inputs and models, easily
  update and customize to new policy constraints
  and guidelines, e.g. issues related to carbon
  trading needs
• Force standards in data classification, archival,
  availability etc
• Overcome capacity constraints in individual
  countries with fairly similar forests/woodland
  ecosystems
• Processing to be distributed across network to
  overcome computer limitations locally

54
Figure 2. Carbon stock in one hectare of Miombo
woodland and agriculture field (Walker and
Desanker , in press Sarah Walker, Masters
Thesis, University of Virginia, November 2002 ) -
Part of ongoing studies to calibrate soil carbon
under different land use histories
55
Outline of a carbon model showing pools and major
drivers of fluxes. This model will be
parameterized for major land cover types (forest,
woodland, savanna, agricultural land).
56
Lessons/Discussion
57
Lessons

• Data mining answer to information explosion and
  lack of (Internet) access Miombo CD model
• Data bundles for specific user groups targeted
  products
• Online maps (limited use to national stakeholders
  in developing countries in Africa since cant
  download data beyond composite maps, ideal for
  new comers)
• Need to connect legacy data with latest
  technology major obstacle to adapting new and
  latest GIS/RS products
• Most use of GIS is obsessed with digital
  maps/atlases. Lots of effort in digitizing maps
  and derived maps into digital layers, often from
  paper atlases. Need to convert to digital
  datasets and modeling, e.g. reconstructing
  derived layers using GIS, such as AEZs, etc. Do
  not digitize what can be derived.

58
Lessons

• Need to develop new methods/models that use
  RS/GIS directly versus computerizing
  manual/legacy methods
• Costs although Landsat 7 cheap, still costly o
  get multiple images per year needed to fully
  capture land cover/land use characterization
• A lot of redundancies in producing map products
  by different agencies and communities UN system,
  national govt depts, science networks, etc. No
  easy mechanism to coordinate
• Access to data that is potentially available
  still an issue such as large acquisitions of
  Landsat

59
Challenges and Possibilities for Developing
Countries Miombo Experience

• What are major constraints
• Internet facilities
• Computing power
• How can they be resolved?
• Online maps
• Converting analog systems to digital in forestry,
  impacts, need for modeling, standards, etc
• Build Data/information base versus direct feeds
  into decision processes
• Core needs
• Monitoring systems
• Accounting systems (monitoring processmodels)
• Training
• Implementation of standards
• Need for Synergies National versus local and
  international monitoring and assessment needs
  (scale, content)
• Mainstreaming Need plans to transition into new
  digital systems

60
end
61
Processing steps for Landsat in Miombo Woodlands
using a hybrid of supervised and unsupervised
steps (Jaiteh, 2000).