Dense forest, closed canopyForest Unit IFO

Logged by SCB in 1994

The dense humid forest of Central Africa is one of the less studied area of the world. As a result, national forest services, park managers, and the research community lack the most basic information on forest distributions, biomass, and rates of land use change. In collaboration with Central African partners, including government forest agencies, conservation groups, and logging companies, we are developing a system for monitoring forest resources that emphasizes on the integration of remote sensing with forestry and biodiversity inventories.

Selective logging is the most extensive land use in the region, with millions of hectares of forested land under concession (i.e. allocated for logging). In order to better manage these forests for timber extraction and biodiversity conservation, we are using satellite imagery (30m Landsat and 1-4m IKONOS) and field surveys to monitor and analyze logging activities at several study sites in the region. Here we focus on the Tri-National Park area on the border between Cameroon, Central African Republic, and the Republic of Congo.

Our Integrated Forest Monitoring System (INFORMS) for logging includes: (A) maps of the distribution of logging activities in the region, (B) development of new integrated monitoring and validation techniques, (C) vegetation/ habitat maps for forestry, wildlife conservation, and carbon modeling activities, (D) quantification of deforestation rates around logging towns, and (E) estimation of logging intensity and timber extraction rates using satellite imagery. All of these results are conveyed to a newly created Central Africa Global Observations of Forest Cover (OSFAC) network. More details on the results of this project are available through our project web site.

Nadine Laporte, Tiffany S. LinThe Woods Hole Research Center,

P. O. Box 296, Woods Hole, MA 02543-0296, U.S.A. (http://www.whrc.org/africa)

More than 40% of the dense humid forest in central Africa has been allocated for timber exploitation. Logging in the region is very selective in comparison to that in the Amazonian forests, with an average extraction rate of ±1 tree/ha, or roughly 10 m3 of wood/ha. With the exception of Gabon, most of the trees harvested are species of African mahogany (Entandrophragma spp.). Until recently, okoume (Aucoumea klaineana) had been the main species harvested in Gabon. It is now replaced by secondary species Ceiba pentadra, which is likely an indication of the impoverishment of the Gabonese forest. Together with adequate forest management policies and planning, accurate mapping of forest types & distribution and monitoring of land cover changes are essential to ensure the sustainability of timber resources.

Baltimore

• National forest services

• Logging companies

•Conservation programs

•Research community

LOGGING MONITORING APPROACH INFORMS Project

Results/ Outputs: Forest monitoring methods Forest type maps Biomass distribution Logging intensity estimates

LOGGING IS THE MOST EXTENSIVELAND USE IN CENTRAL AFRICA

A simple logging index, derived from Landsat satellite imagery (E1), allows the estimation of the number of trees logged by harvesting unit (E2). This relation was established for the CIB forest unit of Mokobo in the Kabo UFA. Logging began in 1998 and completed in 2000. Only 2 recent cloud-free Landsat images are available for this forest unit (1999 and 2001). An index estimating logging intensity was developed using a classification derived from the 2001 Landsat image. This index needs to be validated in other production units before it can be operationalized.

Logging roads create a network of corridors for hunters into the most remote forests of Central Africa. In northern Congo, southeastern Cameroon and C.A.R, more than 4,669 km of roads have been built around the Tri-National Park region. If not managed properly, the opening of roads can seriously undermine wildlife conservation, threatening the survival of endangered species like lowland gorillas, chimpanzees, leopards, bongos, and forest elephants.

TRI-NATIONAL PARK (TNP) AREA“An Island of conservation in a sea of timber exploitation”

Laporte N.T., S. Lin, J. LeMoigne, D. Devers, and M. Honzak (2004), Toward an Integrated Forest Monitoring System for Central Africa. In: Land Change Science: Observation, Monitoring, and Understanding Trajectories of Change on the Earth Surface, Remote Sensing and Digital Image Processing, Vol (6), Ed. G. Gutman. ISBN: 1-4020-2562-9, p 97-110. Chan J. C.W., Laporte N., Defries R., (2002), Texture classification of logging in tropical Africa using machine learning algorithms, International Journal of Remote Sensing, in press.

Laporte N., Goetz S.J., Justice C.O., Heinicke M., (1998), A new land cover map of central Africa derived from multi-resolution, multi-temporal AVHRR Data, International Journal Remote Sensing, 19(18): 3537-3550.

Moigne J., Laporte N., and Netanyahu N.S., (2001), Enhancement of tropical land cover mapping with wavelet-based fusion and unsupervised clustering of SAR and Landsat Image Data, International Society for Optical Engineering (SPIE), Toulouse, 17-21 September 2001, Toulouse, France, 6p.

Wilkie D., Laporte N., (2001), Forest area and deforestation in Central Africa: Current knowledge and future directions, in W. Weber, A. Vedder, S. Morland, L. White (Eds), African Rainforest Ecology and Conservation, Yale University Press, 119-138p.

Dense humid forest

Agriculture and degraded forest

Forest / savanna transition

Edaphic grass savanna

Wet savannas

Dry Savannas

Logging concessions (GFW)

D.C.D.C.

D.C.

R2 = 0.5036

0

0.2

0.4

0.6

0.8

1

0 20 40 60 80

Numb er of t rees harvested p er 50 - ha parcel)

Prop

ortio

n of

ope

n fo

rest

cla

ss

HABITAT MAPPING

TNP

Central African Republic

Cameroon

Gabon

Republic of Congo

Vegetation map derived from AVHRR images, Laporte et al., 1998

Eq. Guinea

Monitoring Logging in Central Africa

Forest Information Users & Collaborators

Bringing expertise for field surveys:

• Forest and fauna inventories

• Validation of forest maps

Remote Sensing Analysis & Data Set Integration

• Technical backstopping of OSFAC network www.osfac.org

• Remote sensing data acquisition and distribution

• Development of new validation methods (video/inventories)

• Remote sensing analysis & ecological field data integration

• Remote sensing/GIS training (Lope workshop 2000)

• Primary study sites (Tri-National Park area – C.A.R, Rep. of Congo, and Cameroon; Okapi Reserve- D.R.C)

ESTIMATION OF LOGGING INTENSITY

E

A

E2

C

Introduction

References

10 kmPark limits Road network Concession limits

Road density: 0.13 km/km2 in the displayed window

E1

Nouabale-Ndoki Nat. ParkLake Lobeke Res.

Dzanga Sangha

LAND USE / LAND COVER CHANGE AROUND LOGGING TOWNS

D

7 km X 10 km

Landsat-4 TM

1990-12-28

Bands: 4,5,3 (RGB)

Pokola is the base for the CIB logging company, and Ngombe is the base for IFO. Only a few hundred people were living in Pokola in the 1970’s; it was a small fishing village along the Sangha River. The population of both towns has increased rapidly with the arrival of the logging companies. The population of Pokola is currently about 9,000.

Landsat-7 ETM+

2002-02-09

Bands: 4,5,3 (RGB)

Land-cover map 1990

Scale:

Evergreen Gilbertiodendron dewevrei forest (closed canopy)Dense humid riparian forest (closed canopy)Dense humid semi-deciduous forest, type I (closed canopy)Dense humid semi-deciduous forest, type II (semi-open to open canopy)Dense humid semi-deciduous forest, type IIISwamps and marshesSavanna and AgricultureBare soil and RoadsWater

Park boundaryCountry limit

Classification based on dry season satellite imagery Landsat-7 ETM+ image, 09 February 2002

Scale: 5 km

DullesAirport

:

Sound management of forest concessions requires accurate mapping of forest types. In collaboration with WCS and the CIB logging company, maps were produced for the Kaboand Pokola concessions using Landsatimagery. Detailed fauna and flora inventories covering more than 1 million hectares in areahave been completed for the Kaboconcession by WCS and CIB. Using simple GIS techniques, statistics for forest types were derived from the vegetation maps (D1). Inventory data are now being used for validation (D2). These data sets also can beused to better understand wildlifedistributions in relation to logging activities.

D2

MANAGEMENT OF FOREST CONCESSIONS

D1

Stratification of the Kabo and Pokola Forest Management Units

Evergreen G. dewevrei monodominant forest, close canopy (Lim)Tall swamp, closed canopy (MC)Dense humid semi-deciduous forest type I, closed canopy (FD)Dense humid semi-deciduous forest type II, semi-open canopy (FC) Dense humid semi-deciduous forest type III, open canopy (FC)Agriculture (Ag)Bare soil and Roads (BS) Bare soil and Roads (BS)Water

Kabo Concession, Kabo VillageInventory transect #10 super-imposed on the forest

stratification map derived from 2001 Landsat imagery

skid trail

Abandoned logging road

Active logging road/ Bare soil

Aerial Digital Video 2001

Forest Unit Pokola 2002bCurrently logged - Skid trails visible(A) Pan & MS bands 431 merged

(B) Panchromatic bandImage acquired June 2002

Forest Unit IFO 1994Logged in 1994 - No skid trails visible

(C) Pan and MS bands mergedImage acquired Oct 2000

Logs (20 m)

B

Democratic Republic of Congo

Research funded by the NASA/LCLUC & the

USAID CARPE program

Land-cover map 2001

It is important to understand the impacts of logging in land use and land cover change in the central African forests. Deforestation rates across the region are variable, and little is known of logging impacts on forest habitat and wildlife.

Pokola, Kabo, and Ngombe are the main logging towns in the Tri-National Park region in the Rep. of Congo. Ouesso, on the other hand, is the administrative center and the largest town in northern Congo.

Between 1990 and 2001, the “urban” area of Pokola increased most rapidly while agriculture expanded most in Ngombe. Ouesso did not change substantially over this period.

Pokola

11 by 10 km

10 by 10 km

18 by 17 km

Pokola

Kabo

Ngombe

Pokola

Forest to other

Other changes

Land cover change

ForestDegraded ForestAgricultureBare soilsWater

Land cover map legend

Classified Landsat TM 1999Lim

MC

FD

FC (logged)

Bare soils

Harvesting units (50 ha)

Logging started in spring 2002

A(B)

B

Active road

Estimating forest logging intensity, CIB - Forest unit Mokobo 1998

After selective logging, forests regenerate rapidly along skid trails with secondary species like Musanga sp. These regenerating forests often have spectral signatures similar to those of naturally open forests (i.e. short swamps), making automatic classification difficult. The same problem is observed with very high resolution IKONOS imagery at 1 & 4 m resolution. Three years after logging, only major and abandoned secondary roads are visible on the images. Only aerial digital videography and/or field surveys allow us to validate the maps developed from satellite imagery.

Mapping active logging

Marantaceae forest (FC)Forest Unit Pokola 2002b

(C)

Central African Republic

Cameroon

Republic of Congo

Nouabale-Ndoki Nat. Park

Lake Lobeke Res.

DzangaSangha

*Ouesso *

*

**Ngombe

Pokola

Ndoki I

Bayanga

* Bomassa

Ndoki II

*

(A)

500 m

500 m500 m

Cla

ssifi

ed L

ands

at T

M 2

001

Acknowledgements: This research would not have been possible without the assistance of Didier Devers at UMD, and the logistic support of WCS and CIB and the precious time of Steve Blake, Brian Curran, Olivier Desmet, Yves Dubois, Paul and Sarah Elkan, Leon Embon, Mike Fay, Patrick Geffroy, Frederic Glannaz, Jackie Glannaz, Patrice Gouala, Steve Gulick, Christian Guyonvaro, Gregoire Kossa, Tom van Loon, Fiona Maissel, Richard Malonga, Alfred Tira, Luca van Der Walt, Chris Wilks, and many others working to protect these unique ecosystems and to ensure the sustainable use of the forest resources.

Skid trailFelling site

Ouesso

100 kmScale:

Acknowledgements

Kabo*