Obtaining high-quality information on forest and land use is essential to analysis of climate change, sustainable forest and land use planning. Papua New Guinea’s (PNG) forest and land cover/land use has been well documented using different methods, land classifications and forest definitions. These studies have delivered significant results indicating a general decline in the forest extent, with the drivers of land use changes attributed to demographic and economic development. This study is a component of the larger National Forest Inventory for PNG within which we sought to stratify and quantify forest and land use change by applying a systematic point-based sampling approach utilizing Open Foris—Collect Earth and freely available satellite images. A total of 25,279 sample points was assessed to determine the PNG’s forest extent and the forest change drivers based on the national forest definition. Analysis revealed that in 2015, about 78% of the country was covered with 12 forest types, and more than 23% of the total forest area has been degraded through anthropogenic activities. Analysis also revealed that between 2000 and 2015, about 0.66% of the total forest area was deforested, and subsistence agriculture was the main driver (0.45%), followed by palm oil planting (0.23%). During the same period, about 6.6% of the total forest area was degraded, and commercial logging was the main cause (6.1%). Apart from Global Forest Watch, this study established more forest than previously estimated in earlier studies. This is due to the fundamental differences in the purposes and methodologies used.

Other land use and agriculture have reduced the global forest since 1990 [1]. Anthropogenic deforestation and forest degradation are responsible for emitting 7–14% of the total CO2 emissions [2]. Forest, however, has a potential to mitigate climate change cost-effectively [3]. For instance, about 1.7 GtCO2e could be sequestrated annually by restoring 350 million hectares of degraded terrestrial [4].

A focus on forests therefore is one of the key aspects of international climate change agreement [5]. As a party to the United Nations Framework Convention on Climate Change (UNFCCC), Papua New Guinea (PNG) is fully committed in identifying appropriate actions that might contribute to combat climate change mitigation and adaptation and adapt to its effects. PNG is obligated to relay its efforts through biennial update reports (BUR) [6], national communications [7], national determined contributions (NDC) [8] and forest reference/emission levels (FRL/FREL) [9] to the UNFCCC. Undated forest and land use change information therefore is necessary for reporting climate change efforts [10]. PNG’s BUR, NDC and FRL were developed using the results of this study.

Furthermore, updated information on PNG’s forest trends will allow policy makers to better understand the magnitude of the issue, track progress and realign priorities to alleviate climate change and make better decisions in managing the forest resources toward balancing environmental conservation and socioeconomic developments.

PNG’s forest land use/land cover has been well documented in studies extending from 1975 to 2015 [11, 12, 13, 14, 15, 16]. These investigations indicate that a decline in the forest area occurred mostly due to logging and subsistence agriculture. Given the continuous change in the forest landscape, this foundation of knowledge constantly requires up-to-date information.

The aim of this study was to determine the forest extent in 2015 and identify the drivers of forest change between 2000 and 2015. It focuses only on anthropogenic deforestation and forest degradation to identify the impacts of human activities on the forest. A point sampling approach with augmented visual assessment was applied to collect forest and land use data using the Collect Earth (CE) software. The assessment was based on the PNG national forest definition [17] and the International Panel on Climate Change (IPCC) Good Practice Guidelines (GPG) for Land Use, Land Use Change and Forestry (LULUCF) [18].

Gamoga [19] and the PNG Forest Authority report [20] described in detail the materials and methods used. A summary account is presented below.

Sampling Design and Intensity

A probabilistic stratified-systematic sampling design was created in QGIS [21, 22] to facilitate area estimation and proportional land compositions. The sampling intensity relies on 0.04 × 0.04° grid (4.44 × 4.44 km) and 0.02 × 0.02° grid (2.22 × 2.22 km) for the three smaller provinces having fewer than 500,000 ha of landmass. In this way, a total of 25,279 square sampling plots were generated and overlaid on PNG land boundaries.

The sample plot size of 100 m × 100 m is consistent with the minimum mapping area required to apply PNG’s national forest definition. Within each sampling square plots, there are 25 sampling points equivalent to about 4% of the plot. These sampling points help quantify and characterize land use within the sample plot.

These sampling plots are automatically organized by the customized PNG Collect Earth in subnational units and arranged along a 4° grid (WGS 1984 datum).

Land Use Assessment Approach

The assessment was conducted by the Papua New Guinea Forest Authority (PNGFA) field officers. These officers were guided by a set of assessment protocols and rules for consistent judgements on the types of the land use observed [19, 20].

Bey et al. [23] well described the application of CE. The PNG customized CE software facilitated the forest and land use change data collection through Google Earth engine using high- and low-resolution satellite images (table 1). Once the areas of interest were activated in CE, the satellite images were simultaneously executed in the Google Earth Engine over the area of interest. Landsat 7 or 8 was used to detect land use change, while the high-resolution images including the supporting resources materials (PNG Forest Base Map 2012 [15] and the Global Forest Cover data set from Hansen [24]) were used as reference points for land use identification and classification. These data sets were converted to XML files and overlaid on the area of interest to assist with visual interpretation.

Table 1.

Satellite Imagery, Source, Type, Year and Purpose.

SourceImagery TypeResolutionAcquisition YearPurpose
Google Earth World-View, QuickBird, Ikonos, SPOT, Landsat, etc. High (0.5–2.5 m), low (30 m) 2000–2015 (to date) Land use and disturbance 
Bing Maps World-View, QuickBird, Ikonos, SPOT, Landsat, etc. High (0.5–2.5 m), low (30 m) 2000–2005, 2007–2015 (to date) Land use and disturbance 
Google Earth Engine Landsat 7 (Annual Greenest Pixel) Low (30 m) 1999–2013 Historical land use change 
Landsat 8 (Annual Greenest Pixel) 2014–2015 Check current situation 
SourceImagery TypeResolutionAcquisition YearPurpose
Google Earth World-View, QuickBird, Ikonos, SPOT, Landsat, etc. High (0.5–2.5 m), low (30 m) 2000–2015 (to date) Land use and disturbance 
Bing Maps World-View, QuickBird, Ikonos, SPOT, Landsat, etc. High (0.5–2.5 m), low (30 m) 2000–2005, 2007–2015 (to date) Land use and disturbance 
Google Earth Engine Landsat 7 (Annual Greenest Pixel) Low (30 m) 1999–2013 Historical land use change 
Landsat 8 (Annual Greenest Pixel) 2014–2015 Check current situation 

Land Use Classification

The six IPCC GPGs for LULUCF categories [18] were used as the main land classes with country-specific subcategories. This study, however, was focused on forestland and cropland. The PNG’s national forest definition was used to classify forestland. The PNG’s forest is defined as land spanning ≥1 ha with trees ≥3 m high and the canopy cover of ≥10% [17]. This interpretation excludes land that is predominantly under agriculture or urban land use.

Forest has been classified into subdivisions based on the natural vegetation types and plantations. The classification of forest vegetation types is based on the structural formation and is described by Hammermaster and Saunders [25]. Brief descriptions are found in table 2.

Table 2.

Forest Vegetation Classification in PNG and Their Short Descriptions.

Forest TypesShort Description
(a) Natural forest 
Low altitude forest on plains and fans Occur below 1,000 m above sea level 
Low altitude forest on uplands Occur below 1,000 m above sea level 
Lower montane forest Occur between 1,000 and 2,000 m above sea level 
Montane forest Occur between 2,000 and 3,000 m above sea level 
Dry seasonal forest Restricted to southern part of PNG in a low rainfall area (1,800–2,500 mm) 
Littoral forest Occur in dry or inundated beach forest 
Seral forest Occur in river line, upper stream, river plains and volcano blast area 
Swamp forest Found in swamp and inundated areas 
Woodland Low and open canopy tree layer 
Savanna Low (<6 m) and open tree layer in low rainfall area with a marked dry season. Restricted to southern part of PNG 
Scrub Community of dense shrubs up to 6 m 
Mangrove Estuarine vegetation found along coastline and in the deltas of large rivers 
(b) Plantation forest 
Eucalyptus deglupta, Araucaria cumminghamii, Araucaria hunstanii, Pinus species, Acacia species, Terminalia Planted forest 
Forest TypesShort Description
(a) Natural forest 
Low altitude forest on plains and fans Occur below 1,000 m above sea level 
Low altitude forest on uplands Occur below 1,000 m above sea level 
Lower montane forest Occur between 1,000 and 2,000 m above sea level 
Montane forest Occur between 2,000 and 3,000 m above sea level 
Dry seasonal forest Restricted to southern part of PNG in a low rainfall area (1,800–2,500 mm) 
Littoral forest Occur in dry or inundated beach forest 
Seral forest Occur in river line, upper stream, river plains and volcano blast area 
Swamp forest Found in swamp and inundated areas 
Woodland Low and open canopy tree layer 
Savanna Low (<6 m) and open tree layer in low rainfall area with a marked dry season. Restricted to southern part of PNG 
Scrub Community of dense shrubs up to 6 m 
Mangrove Estuarine vegetation found along coastline and in the deltas of large rivers 
(b) Plantation forest 
Eucalyptus deglupta, Araucaria cumminghamii, Araucaria hunstanii, Pinus species, Acacia species, Terminalia Planted forest 

Note: PNG = Papua New Guinea.

The cropland category in PNG has been categorized into subsistence and commercial agriculture. The subsistence agriculture involves a temporary cultivation of land on a rotational basis where the cultivated land is abandoned for a few years then recultivated once it has naturally restored its fertility. Commercial agriculture includes large-scale palm oil and other types of plantations (coconuts and cocoa). They can be either active or abandoned.

Assessment of Forest Disturbance by Human Activities

The key features used in identifying forest disturbances or degradation are shown in table 3.

Table 3.

Key Forest Disturbance Features.

Disturbed ForestKey ElementsRemarks
Logged forest Network of logging roads Easy to see 
Gardening Isolated patches of temporary forest clearings at the edge of cropland areas Challenging to see in Landsat 8 & 7 images 
Fire Burnt forest Challenging to detect on Landsat images 
Others Mining clearings and those not identified As above 
Disturbed ForestKey ElementsRemarks
Logged forest Network of logging roads Easy to see 
Gardening Isolated patches of temporary forest clearings at the edge of cropland areas Challenging to see in Landsat 8 & 7 images 
Fire Burnt forest Challenging to detect on Landsat images 
Others Mining clearings and those not identified As above 

Data Cleaning and Quality Checking

Data cleaning involved the correction of integrity errors (blank records, incorrect input values and odd values), while the quality check involved the correction of land use classification and land use change errors. Quality check and analysis was made very easy by Saiku [26] with drag-and-drop function. For example, to determine whether the forest types were correctly identified based on elevation, a simple query was developed by drag-and-drop of the “Elevation range” to COLUMNS then “Land use subdivision” to ROWS and “Land use category” to FILTER to get only the forestland subdivision thus resulting in forest types against elevation range appearing in the Saiku interface (figure 1). If a forest type and the elevation range were inconsistent with table 2, then it was considered that the forest types were wrongly interpreted.

Figure 1.

Saiku interface showing forest types at elevation range.

Figure 1.

Saiku interface showing forest types at elevation range.

Close modal

This study mostly utilized the Saiku export function to export Saiku-generated results to Common Separated Value file format. This allowed analysis to be performed in Microsoft Excel using analysis functions that are not available in Saiku.

Uncertainty Analysis

Both qualitative and qualitative analyses were done. In terms of qualitative analysis, the major sources of expected errors in estimating past land use trends are classification (random and systematic error) and sampling errors (random error). The “classification error” uncertainty was reduced by defining the land use divisions based on the existing classification system described under land use classification section. The sampling assessment errors include both sampling and human errors. The sampling error emanates from unrepresentative sampling, while human error is related to misinterpretation of historical forest and land use and land use change [27].

With quantitative analysis, we used the spreadsheet developed by Food and Agriculture Organization of the United Nations (FAO) with the standard error of an area estimate being: A * sqrt(pi * (1 − pi)/(n − 1)) for the Land Use Category and Conversion during 2000–2015 (activity data) [28, 29]. The equation was taken from Chapter 3, Volume 4 (AFOLU) of 2006 IPCC Guidelines, pp. 3.33–3.34 [30], where the area estimate of each land use category is calculated by multiplying the total area A, for which land categories are to be estimated, by the proportion of sample plots in the specific land category. The percentage of uncertainty associated with the area estimate is calculated as ±1.96 times the standard error of Ai divided by Ai.

The standard error of an area estimate is obtained as

A*pi(1pi)n1,
1

in which pi is the proportion of points in the particular land use category (stratum) i; pi=nin, A is the total area of PNG, n is the total number of sample points and ni is the number of points under a particular land use category.

Of the 25,279 sample plots, 70 were over the sea hence treated as no data. Consequently, the results generated are from the 25,209 sample plots.

Land Use Composition in 2015

Table 4 shows the land use composition of PNG classified as per the IPCC categories. Nearly 80% of PNG is under forestland followed by cropland (11.2%), grassland (5.3%), wetland (5%), settlement (0.8%) and other land (0.1%).

Table 4.

PNG Land Use Composition in 2015 Under IPCC Land Use Categories.

IPCC Land Use CategoryArea (ha)
Forestland 35,963,273 
Cropland 5,158,633 
Grassland 2,442,680 
Wetland 2,126,505 
Settlement 388,495 
Other land 59,277 
Total 46,138,863 
IPCC Land Use CategoryArea (ha)
Forestland 35,963,273 
Cropland 5,158,633 
Grassland 2,442,680 
Wetland 2,126,505 
Settlement 388,495 
Other land 59,277 
Total 46,138,863 

Notes: PNG = Papua New Guinea; IPCC = International Panel on Climate Change.

Forest Status in 2015

PNG has 12 natural forest types. The three most dominant ones are low altitude forest on uplands (30.9%), low altitude forest on plains and fans (24.8%) and lower montane forest (22.3%). In 2015, 23.7% of the total forest was disturbed or degraded. Logging caused 10.8% of forest degradation followed by gardening (8.3%), fire (3.1%) and others (1.6%; table 5).

Table 5.

Forest Types and the Human Impacts in 2015.

Human Impacts (ha)
Forest TypesLoggingFireGardeningOtherNone (ha)Total (ha)
Low altitude forest on plains and fans 2,379,795 160,449 645,816 119,804 5,621,495 8,927,359 
Low altitude forest on uplands 1,230,894 88,256 983,856 121,922 8,702,804 11,127,733 
Lower montane forest 33,240 128,388 1,126,124 51,318 6,666,762 8,005,831 
Montane forest 19,477 10,207  361,131 390,815 
Dry seasonal forest 100,097 96,172 31,403 80,471 2,043,166 2,351,310 
Littoral forest 3,927 1,957 9,810  130,533 146,226 
Seral forest 7,814 5,888 11,761 7,800 287,277 320,540 
Swamp forest 77,363 37,320 99,227 49,212 2,199,666 2,462,788 
Savanna 276,843 3,905 13,674 329,467 623,889 
Woodland 15,681 238,518 46,938 74,560 680,067 1,055,764 
Scrub 4,424 29,384 3,918 3,925 178,511 220,161 
Mangrove 5,890 1,942 15,628 33,346 225,044 281,850 
Plantation 11,821 13,661 1,988 7,828 13,710 49,008 
Total 3,870,945 1,098,253 2,990,581 563,859 27,439,635 35,963,273 
Human Impacts (ha)
Forest TypesLoggingFireGardeningOtherNone (ha)Total (ha)
Low altitude forest on plains and fans 2,379,795 160,449 645,816 119,804 5,621,495 8,927,359 
Low altitude forest on uplands 1,230,894 88,256 983,856 121,922 8,702,804 11,127,733 
Lower montane forest 33,240 128,388 1,126,124 51,318 6,666,762 8,005,831 
Montane forest 19,477 10,207  361,131 390,815 
Dry seasonal forest 100,097 96,172 31,403 80,471 2,043,166 2,351,310 
Littoral forest 3,927 1,957 9,810  130,533 146,226 
Seral forest 7,814 5,888 11,761 7,800 287,277 320,540 
Swamp forest 77,363 37,320 99,227 49,212 2,199,666 2,462,788 
Savanna 276,843 3,905 13,674 329,467 623,889 
Woodland 15,681 238,518 46,938 74,560 680,067 1,055,764 
Scrub 4,424 29,384 3,918 3,925 178,511 220,161 
Mangrove 5,890 1,942 15,628 33,346 225,044 281,850 
Plantation 11,821 13,661 1,988 7,828 13,710 49,008 
Total 3,870,945 1,098,253 2,990,581 563,859 27,439,635 35,963,273 

Forest Change Between 2000 and 2015

Deforestation in the context of this study refers to the conversion of forestland to other land use. About 0.66% of the total forest in 2015 was deforested over 15 years (table 6). More than three quarters (78%) of deforestation occurred at low altitude forest on plains and fans and low altitude forest on uplands. These forest types comprise more than half (56%) of the nation’s forest.

Table 6.

Forest Types Converted to Other Land Use Between 2000 and 2015.

Hectares
Forest TypesSubsistence AgricultureOil PalmOthersTotal
Low altitude forest on plains and fans 66,480 67,334 5,898 139,712 
Low altitude forest on uplands 41,526 13,867 1,963 57,356 
Lower montane forest 41,208 — 1,959 43,167 
Dry seasonal forest 1,963 — — 1,963 
Swamp forest 5,958 — — 5,958 
Savanna — — 1,315 1,315 
Woodland 3,919 — — 3,919 
Total 161,055 81,201 11,134 253,391 
Hectares
Forest TypesSubsistence AgricultureOil PalmOthersTotal
Low altitude forest on plains and fans 66,480 67,334 5,898 139,712 
Low altitude forest on uplands 41,526 13,867 1,963 57,356 
Lower montane forest 41,208 — 1,959 43,167 
Dry seasonal forest 1,963 — — 1,963 
Swamp forest 5,958 — — 5,958 
Savanna — — 1,315 1,315 
Woodland 3,919 — — 3,919 
Total 161,055 81,201 11,134 253,391 

Others include settlement, coconut, and cocoa. The data on these are not significant so lumped together.

Subsistence agriculture and commercial agriculture, especially for oil palm, were the main factors behind deforestation between 2000 and 2015, removing about 63.3% and 32% of the total forest deforested, respectively. There had been a rapid increase in deforestation from 2011 to 2013 caused by subsistence agriculture and by oil palm development from 2010 to 2015 (table 7).

Table 7.

Annual Deforestation by Cropland Types.

Year of ChangeSubsistence Agriculture (ha)Oil Palm (ha)Others (ha)Total (ha)
2001 7,894 3,927 — 11,821 
2002 8,029 — — 8,029 
2003 3,926 5,887 — 9,813 
2004 6,947 — — 6,947 
2005 4,471 5,890 — 10,361 
2006 4,014 — — 4014 
2007 9,859 — — 9,859 
2008 9,837 —  9,838 
2009 12,246 1,963  14,211 
2010 7,941 5,912  13,856 
2011 18,204 3,924 1,315 23,443 
2012 15,947 9,855 3,941 29,744 
2013 25,766 9,992 3,919 39,677 
2014 10,287 17,930 — 28,218 
2015 15,687 15,921 1,959 33,566 
Total 161,055 81,201 11,134 255,361 
Year of ChangeSubsistence Agriculture (ha)Oil Palm (ha)Others (ha)Total (ha)
2001 7,894 3,927 — 11,821 
2002 8,029 — — 8,029 
2003 3,926 5,887 — 9,813 
2004 6,947 — — 6,947 
2005 4,471 5,890 — 10,361 
2006 4,014 — — 4014 
2007 9,859 — — 9,859 
2008 9,837 —  9,838 
2009 12,246 1,963  14,211 
2010 7,941 5,912  13,856 
2011 18,204 3,924 1,315 23,443 
2012 15,947 9,855 3,941 29,744 
2013 25,766 9,992 3,919 39,677 
2014 10,287 17,930 — 28,218 
2015 15,687 15,921 1,959 33,566 
Total 161,055 81,201 11,134 255,361 

Others include settlement, coconut and cocoa. The data on these are not significant so lumped together.

While clearing of forest for subsistence agriculture occurred throughout the country, most clearings occurred in West Sepik. Forest clearance for oil palm was concentrated in the West Sepik, West New Britain, East New Britain and Madang provinces. The first three provinces experienced high deforestation due to oil palm development and subsistence agriculture (figure 2).

Figure 2.

Forest converted to cropland types in provinces between 2000 and 2015. Note: *AROB = Autonomous Region of Bougainville.

Figure 2.

Forest converted to cropland types in provinces between 2000 and 2015. Note: *AROB = Autonomous Region of Bougainville.

Close modal

Forest degradation in the context of this study denotes the conversion of primary or intact forest to secondary forest through significant anthropogenic activities.

Table 8 shows the area of forest types been degraded or disturbed between 2000 and 2015. Most disturbances occurred in low altitude forest on plains and fans and low altitude forest on uplands and logging was primarily responsible (91.8%). Other disturbances were not significant during the same period.

Table 8.

Forest Types Disturbed or Degraded by Human Activities Between 2000 and 2015.

Forest Degradation/Disturbance (Ha)
Forest TypesLoggingFireGardeningOtherTotal% Forest Disturbed
Low altitude forest on plains and fans 1,299,683 15,676 52,030 12,709 1,380,097 58.1 
Low altitude forest on uplands 744,803 43,198 11,251 799,252 33.7 
Lower montane forest 17,628 1,988 17,794 37,410 1.6 
Dry seasonal forest 66,732 1,963 3,925 72,620 3.1 
Littoral forest 1,963 1,963 3,927 0.2 
Seral forest 3,926 1,939 5,865 0.2 
Swamp forest 27,567 7,910 3,970 39,447 1.7 
Woodland 3,925 5,865 9,790 0.4 
Scrub 3,967 3,967 0.2 
Mangrove 3,885 3,885 0.2 
Forest plantations 9,858 1,953 5,867 17,678 0.7 
Total 2,180,053 19,617 134,608 39,662 2,373,940 100.0 
Forest Degradation/Disturbance (Ha)
Forest TypesLoggingFireGardeningOtherTotal% Forest Disturbed
Low altitude forest on plains and fans 1,299,683 15,676 52,030 12,709 1,380,097 58.1 
Low altitude forest on uplands 744,803 43,198 11,251 799,252 33.7 
Lower montane forest 17,628 1,988 17,794 37,410 1.6 
Dry seasonal forest 66,732 1,963 3,925 72,620 3.1 
Littoral forest 1,963 1,963 3,927 0.2 
Seral forest 3,926 1,939 5,865 0.2 
Swamp forest 27,567 7,910 3,970 39,447 1.7 
Woodland 3,925 5,865 9,790 0.4 
Scrub 3,967 3,967 0.2 
Mangrove 3,885 3,885 0.2 
Forest plantations 9,858 1,953 5,867 17,678 0.7 
Total 2,180,053 19,617 134,608 39,662 2,373,940 100.0 

Figure 3 shows annual forest disturbance or degradation from 2000 to 2015. That arising from logging is quite significant compared with other influences. Logging increased slowly, peaked in 2011 and then slowed down.

Figure 3.

Annual forest degradation/disturbance.

Figure 3.

Annual forest degradation/disturbance.

Close modal

Forest Cover in Comparison With Past Studies

Similar studies conducted in PNG show different extents of forest (table 9). Short commentaries on the reasons and what this study has established are offered under the “Forest Extent” section.

Table 9.

Extent of Forest in Different Studies in Papua New Guinea.

StudiesForest Area (mil. ha)Forestland/Cover (%)Period/YearReference Number
PNG Resource Information System 31.9 69.5 1953–1972 [11, 40
Forest Inventory Mapping System 26.1 56.2 1975–1996 [13
The State of the Forest of Papua New Guinea. Mapping the Extent and Condition of Forest Cover and Measuring the Driver of Forest Change in the Period 1972–2002 33.0 71.0 1972–2002 [14
PNG Forest Base Map 37.2 80.6 2012 [15
Global Forest Watch 42.9 93.4 2001–2015 [44
The State of the Forest of PNG 2014. Measuring Change Over Period 2002–2014 27.8 60.4 2002–2014 [16
PNG Forest Resource Assessment (FRA) 2015 Report 33.6 72.5 2010–2015 [50
This study 2015 35.9 77.9 2015  
StudiesForest Area (mil. ha)Forestland/Cover (%)Period/YearReference Number
PNG Resource Information System 31.9 69.5 1953–1972 [11, 40
Forest Inventory Mapping System 26.1 56.2 1975–1996 [13
The State of the Forest of Papua New Guinea. Mapping the Extent and Condition of Forest Cover and Measuring the Driver of Forest Change in the Period 1972–2002 33.0 71.0 1972–2002 [14
PNG Forest Base Map 37.2 80.6 2012 [15
Global Forest Watch 42.9 93.4 2001–2015 [44
The State of the Forest of PNG 2014. Measuring Change Over Period 2002–2014 27.8 60.4 2002–2014 [16
PNG Forest Resource Assessment (FRA) 2015 Report 33.6 72.5 2010–2015 [50
This study 2015 35.9 77.9 2015  

Note: PNG = Papua New Guinea.

Uncertainty Analysis

The uncertainties of other land, settlement (table 10), nonforestland (table 11), nonforestland remaining nonforestland and deforestation (tables 12 and 13) are significant due to only small areas and number of samples sampled. No samples were collected from nonforestland converted to forestland (table 12).

Table 10.

Uncertainty Analysis: Uncertainty of Area Estimate of Each Land Use Category—Current Land Use.

Current Land CategorySample Size (Count)AreapiArea [Ai] (ha)Standard Error (Proportion)Standard Error (mil. ha)Confidence Intervals (ha)Uncertainty (%)
Forest 19,314 35,963,273 0.766154945 35,349,518.07 0.002665957 123,004.24 ± 241,088.3 ± 0.67 
Cropland 3,191 5,158,633 0.126581776 5,840,339.24 0.002094245 96,626.09 ± 189,387.1 ± 3.67 
Grassland 1,317 2,442,680 0.052243246 2,410,443.99 0.001401504 64,663.80 ± 126,741.0 ± 5.19 
Other land 34 59,277 0.001348725 62,228.62 0.000231153 10,665.13 ± 20,903.7 ± 35.26 
Wetland 1,105 2,126,505 0.043833552 2,022,430.23 0.001289439 59,493.26 ± 116,606.8 ± 5.48 
Settlement 248 388,495 0.009837756 453,902.89 0.00062163 28,681.30 ± 56,215.4 ± 14.47 
Total 25,209 46,138,863       
Current Land CategorySample Size (Count)AreapiArea [Ai] (ha)Standard Error (Proportion)Standard Error (mil. ha)Confidence Intervals (ha)Uncertainty (%)
Forest 19,314 35,963,273 0.766154945 35,349,518.07 0.002665957 123,004.24 ± 241,088.3 ± 0.67 
Cropland 3,191 5,158,633 0.126581776 5,840,339.24 0.002094245 96,626.09 ± 189,387.1 ± 3.67 
Grassland 1,317 2,442,680 0.052243246 2,410,443.99 0.001401504 64,663.80 ± 126,741.0 ± 5.19 
Other land 34 59,277 0.001348725 62,228.62 0.000231153 10,665.13 ± 20,903.7 ± 35.26 
Wetland 1,105 2,126,505 0.043833552 2,022,430.23 0.001289439 59,493.26 ± 116,606.8 ± 5.48 
Settlement 248 388,495 0.009837756 453,902.89 0.00062163 28,681.30 ± 56,215.4 ± 14.47 
Total 25,209 46,138,863       
Table 11.

Uncertainty Analysis: Land Use Data Without Verification—Current Land Use.

Land CategorySample Size (Count)Area (ha)piArea [Ai] (ha)Standard Error (Proportion)Standard Error (mil. ha)Confidence Intervals (ha)Uncertainty (%)
Current forestland 19,314 35,963,273 0.766 35,349,518 0.002666 123,004 ± 241,088.3 ± 0.67 
Current nonforestland 5,895 10,175,590 0.234 10,789,345 0.002666 123,004 ± 241,088.3 ± 2.37 
Total 25,209 46,138,863       
Land CategorySample Size (Count)Area (ha)piArea [Ai] (ha)Standard Error (Proportion)Standard Error (mil. ha)Confidence Intervals (ha)Uncertainty (%)
Current forestland 19,314 35,963,273 0.766 35,349,518 0.002666 123,004 ± 241,088.3 ± 0.67 
Current nonforestland 5,895 10,175,590 0.234 10,789,345 0.002666 123,004 ± 241,088.3 ± 2.37 
Total 25,209 46,138,863       
Table 12.

Uncertainty Analysis: Land Use and Land Use Change Data Without Verification—Forest to Nonforest.

Land CategorySample Size (Counts)Area (ha)PiArea [Ai] (ha)Standard Error (Proportion)Standard Error (mil. ha)Confidence Intervals (ha)Uncertainty (%)
Nonforestland converted to forestland — — 0.000 0.000000 ± 0.0 
Forestland converted to nonforestland 139 262,197 0.006 254,405 0.000466 21,519 ± 42,177.7 ± 16.09 
Nonforestland remaining nonforestland 5,756 9,913,393 0.228 10,534,940 0.002644 121,982 ± 239,084.7 ± 2.41 
Total 25,209 46,138,863       
Land CategorySample Size (Counts)Area (ha)PiArea [Ai] (ha)Standard Error (Proportion)Standard Error (mil. ha)Confidence Intervals (ha)Uncertainty (%)
Nonforestland converted to forestland — — 0.000 0.000000 ± 0.0 
Forestland converted to nonforestland 139 262,197 0.006 254,405 0.000466 21,519 ± 42,177.7 ± 16.09 
Nonforestland remaining nonforestland 5,756 9,913,393 0.228 10,534,940 0.002644 121,982 ± 239,084.7 ± 2.41 
Total 25,209 46,138,863       
Table 13.

Uncertainty Analysis: Land Use and Land Use Change Data Without Verification—Deforestation and Forest Degradation.

Land Use Change StratificationPlot CountAreaPiArea [Ai] (mil. ha) [A*pi]Standard Error (Proportion)Standard Error (mil. ha)Confidence Intervals (mil. ha)Uncertainty (%)
Stable forest 18,065 33,589,333 0.718 33,136,741 0.002834 130,735 ± 256,241.2 ± 0.76 
Stable nonforest 5,756 9,913,393 0.228 10,534,940 0.002644 121,982 ± 239,084.7 ± 2.41 
Deforestation 139 262,197 0.006 254,405 0.000466 21,519 ± 42,177.7 ± 16.09 
Forest degradation 1,249 2,373,940 0.050 2,285,987 0.001367 63,062 ± 123,601.2 ± 5.21 
All classes 25,209 46,138,863       
Land Use Change StratificationPlot CountAreaPiArea [Ai] (mil. ha) [A*pi]Standard Error (Proportion)Standard Error (mil. ha)Confidence Intervals (mil. ha)Uncertainty (%)
Stable forest 18,065 33,589,333 0.718 33,136,741 0.002834 130,735 ± 256,241.2 ± 0.76 
Stable nonforest 5,756 9,913,393 0.228 10,534,940 0.002644 121,982 ± 239,084.7 ± 2.41 
Deforestation 139 262,197 0.006 254,405 0.000466 21,519 ± 42,177.7 ± 16.09 
Forest degradation 1,249 2,373,940 0.050 2,285,987 0.001367 63,062 ± 123,601.2 ± 5.21 
All classes 25,209 46,138,863       

Drivers of Forest Change Between 2000 and 2015

Industrial logging, subsistence agriculture and oil palm were the major drivers of forest change identified by [14, 16, 31], and this study also established that they remained the major causes of forest change between 2000 and 2015.

Forest Degradation

Natural resource extraction is central to PNG’s development. Its forest industry is predominantly based on round log export from natural forest and is one of the major economic contributors along with the mining, petroleum and the agriculture sectors. More than 90% of logs harvested are exported as round logs [32]. A total of 31,620 thousand cubic meters of round logs valued at US$1,800 million was exported between 2004 and 2015 [33]. The average annual forest product foreign earnings was about US$292 million between 2011 and 2014 [34]. For these reasons, logging has been the main driver of forest degradation [14, 19, 20] and continued to be the major driver between 2000 and 2015. PNG’s FRL projected that the emissions from deforestation and forest degradation would continue to increase from 2013 to 2018 under business as usual [29]. Over 90% of these emissions came from forest degradation and logging was the main cause.

Logging in intact forest peaked in 2011 and then significantly reduced from 2012 (figure 3). The log export volume in 2011 was the highest ever recorded due to large volumes being sourced from forests cleared under Special Agriculture and Business Leases (SABLs) [32], through which Forest Clearing Authorities (FCAs) were issued, mainly for oil palm developments. FCAs are permits authorized by PNGFA for clearing forest under the Forestry Act 1991 [35]. SABLs thus contributed to the acceleration of forest degradation. Nelson et al. [36] revealed that 36 oil palm proposals covering about 948,000 ha were submitted for consideration, while Partners with Melanesians and Transparency International Inc. (PNG; 2014) reported that 25 large-scale FCAs covering over 1 million hectares of forest had been granted by the PNGFA [37]. Issuance of FCAs started around 2009–2010 and about 5.855 million m3 was harvested between 2010 and 2015 [38]. These figures suggest about 344,412 ha of forest degraded or logged based on the average timber stand density of 17 m3/ha [39] and that only 18.5% was cleared and planted (table 6). This situation justifies the conclusion of Nelson et al. [36] that most developers cleared forest with no intention of cultivating oil palm.

Gardening also contributes to forest degradation. Gardening areas are isolated patches of temporary forest clearings usually at the edge of fallow land (cropland) and once human activities are intensified will eventually become cropland or other land use. Between 2000 and 2015, only 0.4% of the total forest was disturbed (table 8).

Fire used for hunting and subsistence agriculture has a long history in PNG [40]. Occurrences of fire are evident in rainforest soil [41]. About 3.1% of the total forest was burned in 2015 [19, 20] but was not significant over the period 2000 and 2015 (0.1%). While fires occurred in all forest types, they were prevalent in savanna, woodland and scrub forests in 2015. Savanna forest however was most affected. National Aeronautics and Space Administration 2015 Satellite photo showed the locations of fires in PNG with most fires being detected in the southern part of the country where the savanna forest is dominant [42].

Deforestation

Settlement covers about 0.8% of the total land area and accommodates over 7 million of PNG’s population. Over 80% of people live in rural areas [43] and depend mostly on subsistence agriculture for their livelihood. Subsistence agriculture cover over 80% of the total cropland or 9.8% of the total PNG land area in 2015. This heavy dependence makes subsistence agriculture the main driver of deforestation. About 3.6 million hectares (11%) of the intact forest in 1972 had been cleared for subsistence agriculture by 2002 [14]. Deforestation by subsistence agriculture has been widespread and is increasing at 0.03% annually (table 7). The Global Forest Watch (GFW) [44] also indicates increasing trends, likely correlated with population increase. The PNG population increased from 5.1 million in 2000 to 7.2 million in 2011, about 40% increase in 11 years [45]. The growing population has been projected to double in the next 25 years [46]. This gain will most certainly impact the forest.

Over the last decade, palm oil has generated about 39% of agricultural export earnings and is the most successful agricultural crop in PNG, supporting over 160,000 of the rural population [47]. In 2012, there were 144,183 ha of commercial oil palm operated by two companies with established processing plants and 19,777 ha of small holders. These lands have been expanding at a rate of 3,000 ha/year over the last decade [48]. This study found more than 320,000 ha in 2015. More than 50% of oil palm plantations were established by companies without mills in place or in the process of establishing one. They are the ones possibly responsible for the increased deforestation between 2010 and 2015 (table 7). The oil palm annual average deforestation rate, however, was low (0.02%) compared with that of subsistence agriculture (0.03%) between 2000 and 2015 because deforestation from oil palm expansion occurred only in four provinces, while subsistence agriculture is widespread (figure 2). Raschio et al. [49] estimated that oil palm expansion in forestland will amount to between 53,000 and 99,180 ha in Madang, West Sepik, East Sepik and New Britain Island by 2024.

Forest Extent

This study found more forest in PNG than previously estimated by Saunders, 1975 [11], McAlpine et al. [13], Shearman et al. [14], the FAO [50] and Bryan and Shearman [16]. Table 9 shows contradictions among these studies. The fundamental differences are discussed below. Generally, the differences are due to technical variations in measuring forest cover and forest cover change [31, 51] and also reflect different forest definitions and classifications used [16]. Further, they can be attributed to the use of the terms “land use” and “land cover,” which are sometimes used interchangeably. Land use simply concerns how the land is used by the people, while land cover relates to the physical land type [52]. Therefore, understanding the context in which definitions are used to determine the land cover/land use in the past studies on PNG’s forest and land cover is essential to understand the various results.

In the PNG Resource Information System (PNGRIS), land is classified into cultivated (existing land use and fallow land mostly covered with secondary vegetation at various stages of regrowth), uncultivated (grassland, sago stands and savanna woodland) and unused (forest) [11, 40]. This division determines that only intact forest is considered as “forest,” while the disturbed forest is classified as cultivated land. If the forest were categorized as “land use,” then both intact and disturbed (degraded) forest would be included. This being so, about 7.4% was considered as low land use intensity and 12.5% as extremely low and very low land use intensity. Some of these areas could be assumed as disturbed (degraded) forest in the context of the land use and the national forest definition. Hence, if these disturbed vegetated areas were included, the forest cover would be around 89%.

McAlpine et al. [13] reclassified the forest in PNGRIS by excluding the areas of significant disturbance (significant land use intensity) and 3.9 million hectares of woodlands, mangroves, savanna, some areas of swamp forest, dry seasonal, alpine, littoral and seral forests including 1.9 million hectares of logged but regenerating forest because they were misclassified as forest [31]. This resulted in the gross forest area reducing to 26.1 million hectares in the Forest Inventory Mapping System (FIMS) in 1996. The reclassification of the forest was based on the “closed canopy” definition where only the area of trees with “touching or overlapping crowns” were classed as forest [25]. Furthermore, the main purpose of the FIMS data set was to determine the potential timber production areas thus resulting in the exclusion of about 5.8 million hectares of open canopy vegetated land classified under our study as forest.

Shearman et al. [14] indicated that the broad category of rainforest was about 86% in 2002; however, when they applied the forest thresholds of ≥70% canopy cover and ≥5 m tree height, the forest cover was reduced to 71% forest cover in 2002. Consequently, about 7.9 million hectares of the vegetated land classed in our study as forest were excluded.

Bryan and Shearman [16] found 71% of PNG covered with some form of forested landscape but, when they applied the same forest definition used by Shearman et al. [14], the forest was reduced to 60.4% of “closed canopy forest” in 2014. As a consequence, about 8.4 million hectares of dry evergreen, swamp, mangrove and secondary forest were excluded. The woodland and savanna were also omitted. These vegetation types are classed as forest in the present study.

The methodologies on global scale forest assessment using historical satellite images by applying supervised learning algorithms have been well developed in the last decade [24, 53]. Such information is freely accessible by the public through web platforms such as GFW [44]. The GFW estimates 93.4% of forest cover based on ≥30% tree canopy cover threshold [42, 44], which is considerably higher than our findings (78% national forest cover) despite the lower canopy cover threshold of this study (≥10%). This is due to the fundamental differences in the purposes and methodologies between the two studies. While GFW assesses tree cover, our study assesses land use. For example, 95.3% of cocoa and coconut plantation, 88.9% of subsistence farmland and 85.4% of oil palm plantation were identified as forest in GFW. Even 70.9% of settlement in this study was identified as forest in GFW due to some tree cover in the land use; 77.7% of village and hamlet and 57.1% of large settlement (towns and cities) were identified as forest in GFW.

On the other hand, the GFW estimates the area of tree cover loss significantly lower than this study. Total tree cover loss between 2000 and 2015 was 983,977 ha in the GFW, while this study estimates the total degraded forest area in the same period as 2,373,940 ha. This study assessed the forest degradation due to commercial logging activities by assessing the expansion of the logging road network. When the sample point was close enough (approximately 0.6–1.0 km) to the logging road within the logging concessions, we assessed the sample point as forest degradation even if we did not identify actual forest canopy damage. This is to capture the light disturbance, which is difficult to identify using the Landsat images. The GFW does not identify forest losses unless the canopies were destructed. This study provides a conservative estimation of forest cover and a larger estimation of forest loss and disturbance comparing to the GFW. Many countries have also reported significantly different deforestation assessments comparing to the GFW tree cover loss [24]. The GFW provides very useful information, which significantly improves our understanding of the extent and dynamics of the forest on a global scale. However, additional effort is required for assessing forest and its change over time at the country level. This study compared all the 25,279 assessment points against the GFW data as part of the QA/QC protocols, and it significantly improved the accuracy of this study.

The PNG Forest Resource Assessment 2015 report [50] was compiled by the FAO national correspondent (PNGFA) in 2012. This report indicated the lowest deforestation and highest forest degradation rates compared with the past studies. There was an exponential increase in the “other naturally regenerated forest” in 2015. The exponential increase is attributed to the application of simple linear regression using four data points/years (1990, 2000, 2005 and 2010) to predict “other naturally regenerated forest” in year 2015 hence not realistic. The area of forest in the Forest Base Map [15] and this study are similar due to the same forest definition and forest classes used.

Forest degradation is the main driver of forest change with the main driver been commercial logging. Deforestation is not significant compared with forest degradation. The major driver is subsistence agriculture, followed by palm oil. Both deforestation and forest degradation are in an increasing trend. Disturbances caused by natural phenomena were not assessed and should be tackled in future assessments to understand the overall impact on forests.

PNG has more forest than previously estimated based on the national forest definition. This doesn’t mean that PNG’S forest is increasing. It is due to the different forest definitions and methodologies used between the earlier studies. The understanding of how a forest is defined and applied therefore is important to reduce ambiguity and debate.

Furthermore, forest definition has an influence on the forest extent and the rate of forest change. Definitions with high threshold values can exclude some important forest areas that have significant ecological functions and carbon emission and removal potentials, which can either be under or over estimated. Harmonization of the forest definitions in the country therefore is not only necessary for consistent national and international reporting but to develop policies that support the livelihoods of forest dependent communities including addressing climate change and other values of forest.

  1. How best can forest and land use be monitored?

  2. What are the implications of applying different methods and forest definition for the assessment of forest and land use change?

  3. What are the causes of deforestation and forest degradation in PNG?

GG supervised the activity data collection, data cleaning, quality check, analysis of the results and overall drafting of the article. RT provided guidance and made critical overall input to the article. HA provided guidance on activity data collection, data cleaning, quality check and made input on the GFW discussions. MH provided guidance on the overall design of the Collect Earth project file, data cleaning and quality check including conducting the uncertainty analysis and GFW data comparisons with this study. OI provided technical support on design aspects and facilitation of the Collect Earth assessment, data cleaning and quality check. All authors developed the forest and land use change assessment methodology.

We extend our sincere appreciation to Dr. Paul Dargusch, Professor Fabio Attorre and Dr. David Wadley for valuable input and guidance. Also acknowledged are the FAO Open-Forist Collect Earth team in Rome for their technical advice and guidance in customizing the Papua New Guinea Collect Earth to achieve the objectives of this study. We take the opportunity to acknowledge those Papua New Guinea Forest Authority officers both at headquarters and field offices who assisted in the collection of land use activity data and, likewise, other parties who offered support and guidance.

The authors have no competing interests to declare.

This study was funded by the United Nations Reduce Emissions from Deforestation and Forest Degradation (UNREDD) program and the European Union (EU). This study was also supported by the Italian Development Cooperation (DGCS) through the FAO-Mountain Partnership Secretariat.

1
Joint Research Centre
.
The European Commission’s Science and Knowledge Service [Internet]
.
2020
[cited 12 July 2021]
.
Available
: https://ec.europa.eu/jrc/en/science-update/deforestation-and-forest-degradation-major-threat-global-biodiversity.
2
Joint Research Centre
.
The European Commission’s Science and Knowledge Service [Internet]
.
2014
[cited 12 July 2021]
.
Available
: https://ec.europa.eu/jrc/en/science-update/reporting-greenhouse-gas-emissions-deforestation-and-forest-degradation-pan-tropical-biomass-maps.
3
Gibs
D
,
Harris
N
,
Seymour
F
.
By the Numbers: The Value of Tropical Forests in the Climate Change Equation [Internet]
.
2018
[cited 13 July 2021]
.
Available
: https://www.wri.org/insights/numbers-value-tropical-forests-climate-change-equation.
4
IUCN
.
Forests and Climate Change [Internet]
.
2021
[cited 12 July 2021]
.
Available
: https://www.iucn.org/resources/issues-briefs/forests-and-climate-change.
5
United Nations Framework Convention on Climate Change
.
The Paris Agreement [Internet]
.
2015
[cited 19 July 2021]
.
Available
: https://unfccc.int/process-and-meetings/the-paris-agreement/the-paris-agreement.
6
United Nations Framework Convention on Climate Change
.
Biennial Update Report Submissions from Non-Annex I Parties [Internet]
.
2020
[cited 19 July 2021]
.
Available
: https://unfccc.int/BURs.
7
United Nations Framework Convention on Climate Change
.
National Communication Submissions from Non-Annex I Parties [Internet]
.
2020
[cited 19 July 2021]
.
Available
: https://unfccc.int/non-annex-I-NCs.
8
United Nations Framework Convention on Climate Change
.
Nationally Determined Contributions (NDCs) [Internet]
.
2016
[cited 13 July 2021]
.
Available
: https://unfccc.int/.
9
United Nations Framework Convention on Climate Change
.
Forest Reference Emission Levels [Internet]
.
2021
[cited 19 July 2021]
.
Available
: https://redd.unfccc.int/fact-sheets/forest-reference-emission-levels.html.
10
United Nations Framework Convention on Climate Change
.
National Reports from Non-Annex I Parties [Internet]
.
2021
[cited 24 July 2021]
.
Available
: https://unfccc.int/national-reports-from-non-annex-i-parties.
11
Saunders
JC
. Agricultural Land Use of Papua New Guinea (Map With Explanatory Notes).
Scale 1: 100,000. PNGRIS Publication 1
.
Canberra, Australia
:
AIDAB
;
1993
.
12
Allen
B
,
Bourke
R
,
Hide
R
.
The sustainability of Papua New Guinea agricultural systems: the conceptual background
.
Glob Environ Change
.
1995
;
5
:
297
312
.
13
McAlpine
J
,
Quigley
J
. Forest resources of Papua New Guinea.
Summary statistics from the Forest Inventory Mapping (FIM) System. Coffey MPW Ltd for the Australian Agency for International Development and the Papua New Guinea National Forest Service
;
1998
.
14
Shearman
PL
,
Bryan
JE
,
Hunnam
P
et al. The State of the Forest of Papua New Guinea. Mapping the Extent and Condition of Forest Cover and Measuring the Driver of Forest Change in the Period 1972–2002.
Port Moresby, Papua New Guinea
:
University of Papua New Guinea, Port Moresby
;
2008
.
15
Turia
R
,
Malan
P
,
La’a
P
. Papua New Guinea Forest Forest Base Map & Atlas.
Boroko, Papua New Guinea
:
PNG Forest Authority & JICA
;
2019
.
16
Bryan
JE
,
Shearman
PL
. The State of Forest of Papua New Guinea 2014: Measuring the Change Over the Period 2002–2014.
Port Moresby, Papua New Guinea
:
University of Papua New Guinea
;
2015
.
17
The Government of PNG
. PNG National Forest Definition.
Port Moresby, Papua New Guinea
:
National Executive Council
;
2014
.
18
Penman
G
,
Hiraishi
K
,
Kruger
PR
et al. Intergovernmental Panel on Climate Change Good Practice Guidance for Land Use, Land-Use Change and Forestry.
Japan
:
Institute for Global Environmental Strategies (IGES) for the IPCC
;
2003
.
19
Gamoga
G
. Measuring forest land use change in Papua New Guinea between 2000–2015.
Master Science Thesis Paper
.
Lae, Papua New Guinea
:
Forestry Department, Papua New Guinea University of Technology
;
2018
.
20
PNG Forest Authority
. Forest and Land Use Change in Papua New Guinea 2000–2015.
Port Moresby
,
Papua New Guinea
;
2019
.
21
FAO
.
Collect Earth 1.1.1 User Manual. A Guide to Monitoring Land Use Change and Deforestation With Free and Open-Source Software [Internet]
.
2015
June [cited 24 July 2021]
.
Available
: http://reddplus.mn/eng/wp-content/uploads/2018/08/Collect_Earth_User_Manual_20150618_highres_full.pdf.
22
QGIS
.
A Free and Open Source Geographic Information System [Internet]
.
2019
.
Available
: https://www.qgis.org/en/site/.
23
Bey
A
,
Sánchez-Paus Díaz
A
,
Maniatis
D
et al.
Collect earth: Land use and land cover assessment
.
Remote Sens
.
2016
;
8
:
807
.
doi:10.3390/rs8100807
24
Hansen
MC
,
Potapov
PV
,
Moore
R
et al.
High-resolution global maps of 21st-century forest cover change
.
Science
.
2013
;
342
:
850
853
.
25
Hammermaster
ET
,
Saunders
JC
. Forest Resources and Vegetation Mapping of Papua New Guinea.
Canberra, Australia
:
Australian Agency for International Development
;
1995
.
26
Barber
T
.
Saiku Documentation. Release 3.x. [Internet]
.
2018
.
Available
: https://saiku-documentation.readthedocs.io/en/latest/.
27
Potapov
P
,
Hansen
M
,
Stehman
S
,
Loveland
T
,
Pittman
K
.
Combining MODIS and Landsat imagery to estimate and map boreal forest cover loss
.
Remote Sens Env
.
2008
;
112
(
9
):
3708
3719
.
28
Government of Papua New Guinea
.
Biennial Update Report Submissions From Non-Annex I Parties [Internet]
.
2018
[cited 14 July 2021]
.
Available
: https://unfccc.int/BURs.
29
The Government of PNG
. Papua New Guinea’s National REDD+ Forest Reference Level.
Modified Submission for UNFCCC Technical Assessment in 2017
.
Port Moresby
;
2017
.
30
International Panel on Climate Change
.
2006
IPCC Guidelines for National Greenhouse Gas Inventories Volume 4 Agriculture, Forestry and Other Land Use [Internet]
.
IPCC
;
2006
[cited 25 July 2016]
.
Available
: https://www.ipcc-nggip.iges.or.jp/public/2006gl/vol4.html.
31
Filer
C
,
Keenan
RJ
,
Allen
BJ
et al.
Deforestation and forest degradation in Papua New Guinea
.
Ann For Sci
.
2009
;
66
:
813
.
32
Lawson
S
. Illegal Logging in Papua New Guinea.
London, UK
:
Chatham House
;
2014
.
33
Bank of PNG
.
Quarterly Economic Bulletin [Internet]
.
2019
.
Available
: https://www.bankpng.gov.pg/publications-presentations/quarterly-economic-bulletin/.
34
Oxford Business Group
.
[Internet]
.
2020
.
Available
: https://oxfordbusinessgroup.com/analysis/pngs-export-raw-logs-supports-growth.
35
Government of Papua New Guinea
. Forestry Act 1991.
Port Moresby, Papua New Guinea
:
PNG Forest Authority
;
1991
.
36
Nelson
PN
,
Gabriel
J
,
Filer
C
et al.
Oil palm and deforestation in Papua New Guinea
.
Conserv Lett
.
2013
;
7
:
188
195
.
37
Partners with Melanesians and Transparency International Inc. (PNG)
. Why the Forest Clearance Authorities Are Legally Invalid and Should Be Cancelled.
Port Moresby, Papua New Guinea
:
Partners with Melanesians and Transparency International Inc. (PNG)
;
2014
.
38
PNG Forest Authority
. Status of Various FCA Application by Province.
Internal Report
.
Port Moresby
,
Papua New Guinea
;
2015
.
39
PNG Forest Authority
. Current State of the Forest Sector.
Port Moresby
,
Papua New Guinea
;
2020
.
40
Bourke
RM
,
Hardwood
T
. Food and Agriculture in Papua New Guinea.
Canberra, Australia
:
Anu Press. Australian National University
;
2009
.
41
Haberle
SG
,
Hope
GS
,
Kaarsa
D
.
Biomass burning in Indonesia and Papua New Guinea: natural and human induced fire events in the fossil record
.
Palaeogeogr Palaeoclimatol Palaeoecol
.
2001
;
171
:
259
268
.
43
The World Bank Group
.
Rural Population—Papua New Guinea [Internet]
.
2019
[cited 7 November 2019]
.
Available
: https://data.worldbank.org/indicator/SP.RUR.TOTL.ZS?locations=PG.
44
Global Forest Watch
.
Forest Monitoring Designed for Action [Internet]
.
[cited 6 November 2019]
.
Available
: https://www.globalforestwatch.org/.
45
National Statistics Office
. National Population and Housing Census 2011.
Port Moresby, Papua New Guinea
:
National Statistics Office
;
2015
.
46
Cuthbert
JJ
,
Bush
G
,
Chapman
M
et al. Analysis of National Circumstances in the Context of REDD+ and Identification of REDD+ Abatement Levers in Papua New Guinea.
Report produced by the Wildlife Conservation Society (Goroka, Papua New Guinea), for Papua New Guinea’s UN-REDD National Programme
;
2016
.
47
ITS Global
. Economic Benefits of Palm Oil in Papua New Guinea.
Melbourne
;
2011
.
48
PNG Oil Palm Council
. Palm Oil Industry Statistics 2012. In:
Filer
et al.
Deforestation and forest degradation in Papua New Guinea
.
Ann For Sci
.
2009
;
66
:
813
.
Port Moresby, Papua New Guinea
:
Papua New Guinea Palm Oil Council
;
2013
.
49
Raschio
G
,
Alei
F
,
Alkam
F
. Draft report on Future Deforestation Modelling and Land Suitability Assessment for Oil Palm.
Agriculture of Mapping Assessment in Papua New Guinea
.
Port Moresby, Papua New Guinea
:
UNDP
;
2016
.
50
FAO
. Country Report Papua New Guinea.
Rome, Italy
:
Food and Agriculture Organization of the United Nations
;
2015
.
51
Shearman
PJ
,
Bryan
JB
,
Mackey
B
et al.
Deforestation and degradation in Papua New Guinea: a response to Filer and colleagues, 2009
.
For Sci
.
2010
;
67
:
300
.
52
Coffey
R
.
Michigan State University [Internet]
.
2013
.
Available
: http://msue.anr.msu.edu/news/the_difference_between_land_use_and_land_cover.
53
Sandker
M
,
Carrillo
O
,
Leng
C
et al.
The Importance of High–Quality Data for REDD+ Monitoring and Reporting [Internet]
.
2021
[cited 27 January 2021]
.
Available
: https://doi.org/10.3390/f12010099.