Skip to main content

Table 1 Summary of approaches to forest degradation monitoring

From: Current remote sensing approaches to monitoring forest degradation in support of countries measurement, reporting and verification (MRV) systems for REDD+

 

Method

LiDAR

 

Optical

SAR

Mapping approach

Dense time-series tracking

Change detection

Vegetation indices

Data transforms

Spectral Mixture Analysis

Classification

Interferometry

Modelling

Data fusion

Visual interpretation

ICESat GLAS

Airborne LiDAR

MODIS

CBERS

Landsat

SPOT

Sentinel-2

RapidEye

Quickbird

IKONOS

ALOS-1/2 PALSAr-1/2

ENVISAT ASAR

SRTM

TerraSAR-X

TanDEM-X

Cosmo-SkyMed

GeoSAR

(1) Detection and characterisation of degradation type

 Forest disturbance mapping

 

           

            

 

           

            

   

        

            

 

           

            
 

 

          

            
    

    

   

            
    

       

              
    

            

         
    

        

            

 

         

              
  

  

         

           
 

                  

      
     

        

     

  

   

 Identification of canopy gaps

    

             

        
  

           

 

          
 

                     

   

                   

 

   
     

              

 

   
     

                 

   
     

     

               
       

 

 

               

 Proxies

     

   

    

    

       
         

             

   
         

                

     

              

  

   
         

 

               

(2) Quantification of carbon stock changes

 Tracking of secondary and degraded forest dynamics

    

        

            
     

        

     

      
     

 

   

               

 Canopy height change

 

     

   

               
       

  

                
      

             

      
      

               

  

 
      

               

 

  
      

               

 

  

 AGB change

       

   

               
       

   

               
       

   

               
       

   

               
       

 

                
       

 

                
       

            

      
      

              

 

  
      

              

 

  
 

Study specifics

Mapping approach

Scalea

Country

Map resolution (m)

Map accuracy/RMSE

References

(1) Detection and characterisation of degradation type

 Forest disturbance mapping

R

Ethiopia

30

 

[15]

R

USA

30

 

[62]

R

USA

30

98% (prod acc.), 86% (user acc. spatial), 80% (user acc. temporal), 90% (overall acc. LC)

[96]

N

Australia

30

 

[54]

N

USA, Canada

30

44.6 ± 5.8% (omission), 27 ± 4.5% (commission)

[61]

N

Brazil

20–30

 

[90]

R

Brazil

250

 

[80]

R

Indonesia

6.5

91.5%, 0.87 (Kappa)

[19]

R

Central Africa

30

87%

[34]

R

USA

250

88%

[87]

R

Tanzania

10–20

20–25% misclassification (20% cover class)

[37]

R

Brazil

25 m

 

[35]

R

Indonesia

<30

79.6%

[94]

 Identification of canopy gaps

R

Gabon, Congo

2.4

 

[72]

R

Cambodia

10, 30

 

[51]

R

Panama

1

 

[50]

R

Sumatra, Brazil

1–3

93.4% (FAR 2.3% at 95% conf int.)

[36]

R

Congo

<10

53.6%, 100% (user acc.)

[74]

R

Brazil

2

86%

[7]

R

Peru

  

[5]

R

Brazil

5

 

[2]

 Proxies

R

Congo

  

[52]

R

Norway

10

 

[83]

R

PNG

5

 

[95]

R

Congo

 

95% (user acc.), 70.4% (overall acc., L-band), 100% (user acc.), 53.6% (overall acc., X-band)

[74]

R

Amazon

  

[2]

(2) Quantification of carbon stock changes

 Tracking of secondary and degraded forest dynamics

R

Brazil

30

88% (Kappa 0.62)

[29]

R

Australia

25

77.8% (Kappa 0.69)

[59]

R

Costa Rica

20

RMSE RH100 1.34 m (r2 = 0.69, p < 0.001)

Corr. with ISODATA1 and early stage (97%), ISODATA3 and late stage (99%), ISODATA2 and intermediate stage (56%)

RH100/RH75 r2 0.79 (late), r2 0.73 (intermed.), r2 0.72 (early)

RH50 r2 0.3 (intermed.)

[11]

 Canopy height change

R

Brazil

1

 

[2]

R

Gabon

 

r2 = 0.83, RMSE 3.3 m, n = 95

[65]

R

Australia

  

[55]

R

Cameroon, DRC

2

75–82%

[13]

R

Tanzania

  

[84]

N

Uganda

 

0.9 mm (bias), 8–16 mm (bias upslope)

[85]

 AGB change

R

Kalimantan

 

r2 = 0.77 (PPR 54.2 Mg ha−1), r2 = 0.81 (PPR 47.4 Mg ha−1)

[16]

R

Brazil

 

R2 = 0.7, SE 41.5 Mg ha−1

[2]

R

Kalimantan

 

r2 = 0.88, RMSE ± 13.8 Mg 0.13 ha−1

[46]

R

Norway

14

95.7–97.8% (classification of deforestation and untouched classes), 56.3–69.2% (degradation classes)

AGB: SE 5–8.4 Mg ha−1, r2 = 0.88–0.98

SE reduced by 18–84% using LiDAR, largest gains in degradation class (73–84%)

[67]

R

Kalimantan

50/100

SE 53.2 Mg ha−1 (n = 51 @50 m), 49.1 Mg ha−1 (@100 m)

[71]

R

Brazil

30

p < 0.0001, R2 = 0.6, N = 26

[29]

R

Mozambique

 

RMSE 8.7–10.9 MgC ha−1

Mean error 9.8 ± 0.7 MgC ha1, mean absolute bias 1.6 ± 0.1 MgC ha−1

[78]

R

Tanzania

10

67.2 Mg ha−1 (51%), bias −5.5 Mg ha−1 (−4.2%), precision 67 Mg ha−1 (51%)

[84]

N

Uganda

 

±8.5 Mg ha−1 (95% conf int.)

[85]

  1. aScale: ‘N’ National, ‘R’ Regional