Assessment of change in forests land, carbon stock and carbon emissions of KPK, Pakistan for past three decades using geospatial techniques

Global environmental problems have become more obvious since the industrial era, deforestation being one of them (Iqbal & Khan 2014). Among the countries affected by climate change, Pakistan is at the higher end of the list, and deforestation is one of the deeper problems it is facing right now (Hussain et al. 2019). Fortunately, the plantation of forests plays a considerable part in measures related to climate change mitigation (Hussain et al. 2019). Activities that allow carbon dioxide (CO₂) in the atmosphere lead to the destruction of forests. New forests, however, help in mitigating detrimental impacts created by the change in climate, by eliminating CO₂ from the atmosphere (Khan et al. 2021).

Globally, forests cover up 31% of the terrestrial area as of this time (UN Global Forest Goals Report 2021). The percentage of Pakistan's forest area is unclear. Only some countries have consistent data from proper assessments over time. The UN's Food and Agriculture Organisation (FAO) estimates Pakistan's forests cover 2.2% of the total land. On the other side, Pakistan Forest Institute figures it to be 5.1%. 14.10% of the KPK is covered with forests, which is almost 40% of all forests in Pakistan (Kahlon & Talal 2020).

To expand and maintain their forest area, countries created strategies and plans, often with calculated targets, detailed objectives, processes, and resources. Recently Pakistan started Billion Tree Tsunami Project, which aimed for an increase in the forest area to 6% by 2020 and then to 10% by 2030 (Kahlon & Talal 2020). KP forest department revealed the Billion Tree Tsunami Afforestation Project or BTTAP in 2014, which followed the guidelines of Forest Ordinance 2002 and KP Forest Policy. It was planned in accordance with the Green Growth initiative in Khyber Pakhtunkhwa (KPK) Province to promote the forest sector. It concluded in November 2017 (Kamal et al. 2019).

Environmentalists and naturalists have given extensive attention to the changing aspects of land use and many environmental problems such as flooding, salinity, water quantity and quality degradation, land degradation and its direct effect on surroundings such as local climate, forest cover loss, reducing agricultural stretches, increasing built-up and increasing impermeable surfaces, etc. (Fahad et al. 2020, 2022). Land-use and land-cover (LULC) change is a significant motivator of global change (Butt et al. 2015). Specifically, changes in tropical areas are of major concern due to the widespread and quick changes in the distribution and attributes of forests (Iqbal & Khan 2014). Through in-depth research into environmental changes, researchers are seeing that LULC change due to human activities play a major role in the evolution of global warming (Hossen et al. 2019). It is safe to say that LULC is a product of the social and economic activities of humans and their environment (Chang et al. 2018).

Several anthropogenic activities like tropical deforestation, land misuse, biomass burning, soil erosion, agricultural practices and industrial development release a huge amount of CO₂ into the atmosphere which results in increased temperature which melts glaciers and causes flooding (Fahad et al. 2020). Among these, forests play a critical role as a sink for carbon emissions. Between 1990 and 2010, the total global forest carbon stock fell from 668 gigatons (Gt) in 1990 to 662 Gt in 2010. This was due to the decline in the forest area. In 2020, it remained at 662 Gt (UN Global Forest Goals Report 2021).

LULC related to deforestation and land degradation releases CO₂ and other greenhouse gases which turned out to be an urgent matter in the scientific community (Hossen et al. 2019). In a study that focused on mapping deforestation trends in Western Himalayas, Pakistan, extensive deforestation was revealed in the area from 1990 to 2010. Nearly 170,684 ha (1,706.84 km²) of the forest had been lost, or roughly 0.38% of forest land was degraded every year (Qamer et al. 2016). Another study took place in Margallah Hills National Park. It is a protected area of Islamabad. The results showed a decrease in forested land in 2017, compared to that in 1990. Also, land-use change released substantial carbon into the atmosphere. This implied that land-use change considerably modifies carbon dynamics in the long run (Mannan et al. 2018).

LULC changes were observed in Azad Jammu and Kashmir, Pakistan from 1998 to 2009 (Iqbal & Khan 2014). Forest cover, which covered 23.60% (175.50 km²) area in 1998 had declined to 20.90% (155.50 km²) in 2009. In one study by Ali et al. (2017), a decline was seen in forest cover from 1992 to 2011 in Swat Valley, Pakistan. Total annual deforestation was 1,268 ha from 2001 to 2008 in the valley. The total percent drop in carbon stocks because of deforestation was recorded to be 17.25. Dynamics of LULC change were studied by Hassan et al. (2016) in Islamabad, Pakistan from 1992 to 2012. While the other classes (agricultural land, settlement, water bodies) witnessed an increase, the forest area was reduced to a great extent.

Anthropogenic activities have distinct carbon footprints that are measured as total emissions of GHGs resulting from an activity. A study by Fahad et al. (2020) suggested that in order to meet the global climate targets, quantification of the net global warming potential of agricultural practices requires precise estimates of local, regional and global carbon budgets. They conducted a case study for observing the development of a deep soil carbon profile resulting from a 10-year wheat–cotton and wheat–maize rotation on semi-arid lands. Results showed that by implementing no tillage along with mulch application, lower carbon losses from soil can mitigate the risks associated with global warming. Therefore, it is necessary to reconsider agricultural practices and soil erosion after a land-use change when calculating global carbon footprints (Fahad et al. 2021a, 2021b).

Changes in LULC affect carbon stocks at different intensities, therefore, the valuation of carbon stocks can assist in decision-making, mainly in the context of ecosystem services in areas of interest (Fernandes et al. 2021). Therefore, this study aimed to assess the LULC change patterns over the past three decades (1990–2020) to better understand the merits or demerits of the project, and to help policymakers, environmental planners and administrators with future projects of the same nature. Moreover, a carbon stock assessment was also done to help determine the impact of local tree plantation drives. The main objectives of the study were: (i) to identify the forest cover change dynamics in selected districts of KPK over a defined temporal scale, and (ii) to assess changes in carbon stock and emission from deforestation using the mathematical algorithm as a result of LULC changes.

The climate in the province fluctuates in accordance with altitude. The mountainous regions experience cold winters and cool summers. A broad range of temperatures has been experienced in the province. Precipitation usually occurs from January to April. It is variable but averages roughly 16 inches annually (Raziq et al. 2016).

Satellite imagery provided by Landsat (4-5 & 8) for AOI was secured from the USGS website for the last three decades. The quality of the image especially for those with limited or low cloud cover was the deciding factor for the selection of Landsat satellite images dates. The best imagery found was of September for the years 1990, 2000, 2009 and 2020 (Table 1).

Imageries from satellites may often be distorted geometrically due to differences in acquisition systems and certain movements. Distortions such as geometric and radiometric ones were corrected before processing. All the bands were stacked in ERDAS Imagine software to get the composite image. As the study area was covered by two or more tiles. Mosaic was done to merge images or tiles to create a single image. The study area was extracted from the mosaiced images of 1990, 2000, 2009 and 2020 in ArcMap.

Maximum likelihood supervised classification was performed over an area of interest (AOI) on images for the years 1990, 2000, 2009 and 2020 in ArcMap. LULC was classified into five classes (Table 2). The LULC classes included forest area, an agricultural area, settlement, water bodies and others.

More than 150 training samples/signature files were compiled for each land cover type (Forest area, Agricultural area, Water bodies, Settlement and Others). For guidance, these training areas were identified and visualized in high-resolution images on Google Earth.

An accuracy assessment of the images was carried out to determine the quality of information originating from the data. It was carried out using by taking Reference points based on ground truth data taken from Google Earth and visual analysis. To assess the accuracy of the classification, every LULC map was compared to corresponding reference data. The ground truth data obtained as a result was used to verify the classification accuracy.

A comparison of reference data and classification results was carried out statistically using error matrices. For each classified image (1990, 2000, 2009, 2020), the overall, producer's and user's accuracies were obtained.

The attribute tables of each class for all classified images were used to calculate the area of different land cover over the AOI. This was done in ArcMap. The analysis was helpful to identify various changes occurring in different classes of land use like an increase in urban built-up area or forest-covered area, etc. that took place over the time span of the study.

Carbon stock assessment was done by using emission factors from the Forest reference emission level of KPK obtained from Pakistan Forest Institute, Peshawar (Ali et al. 2020). There are different methods used for calculating carbon sequestration on different scales, globally (Sun & Liu 2020). The formulas and constant values used in this study were taken from a study by Ali et al. (2017) and Khan et al. (2021), based on those reported by the Intergovernmental Panel on Climate Change's Greenhouse Gas Inventory Guidelines Reference Manual which are also standardized by UNFCCC.

Next in 2009 image classification results (Figure 2(c)), the forest area and agricultural land decreased to 1,486 and 1,846 km², respectively. The settlement has kept on increasing in 2009, i.e. 2,731 km² as compared to 2000. Similarly, in 2020 (Figure 2(d)), an increase in forest area was noticed, i.e. 2,323 km² compared with 1990 and 2009. This is due to afforestation made by the KPK government under the project Billion Tree Tsunami. The details of the area for each class are shown in Figure 3. It can be seen clearly in Figure 3, forest area continued to decline from 1990 to 2009. Afterward, a sharp increase is seen in the graph in the forest class. Agricultural land continues to decrease over the study period. While the settlement shows a huge increase with each passing decade.

LULC changes were observed in Azad Jammu and Kashmir, Pakistan from 1998 to 2009. Forest cover, which covered 23.60% (175.50 km²) area in 1998 had declined to 20.90% (155.50 km²) in 2009 (Iqbal & Khan 2014). Ali et al. (2017) conducted a study on LULC changes in Swat Valley, Pakistan. They observed a declining trend in forest cover from 1992 to 2011. Total annual deforestation was 1,268 ha from 2001 to 2008 in the valley. The total percent drop in carbon stocks because of deforestation was recorded to be 17.25.

Dynamics of LULC change were studied by Hassan et al. (2016) in Islamabad, Pakistan from 1992 to 2012. While the other classes (agricultural land, settlement, water bodies) saw an increase, forest area was reduced to a great extent.

The LULC transition statistics of each class for 1990–2000, 2000–2009 and 2009–2020 are shown in Table 3 and discussed below.

Between 1990 and 2000, forest area reduced by 4% relative to the start of the study time span (1990) which was approximately 74 km². By 2009, this decline continued, and it was seen that the forest area reduced by 11%, compared to what was seen in 2000 (Table 3).

Forest areas increased tremendously between 2009 and 2020 by 56%. This may be due to new plantations as a result of BTTP in the study area. Around 836 km² of land was found to be covered with forests. The net change in forest area was recorded as a 32% increase.

Agricultural land decreased in the first decade by 8%, losing about 173 km² of land to increasing settlement. From 2000 to 2009, another 4% decrease was seen as 89 km² of land was lost as an agricultural area. The greatest decrease was seen from 2009 to 2020, whereby 674 km² was converted into other classes. Overall, a 44% decrease was seen in this class over three decades.

Being a category largely influenced by changing seasons, a 38% decrease was seen in water bodies from 1990 to 2000, followed by a 28% increase from 2000 to 2009. A decrease of 54% was seen from 2009 to 2020. This may be due to less rainfall and increased temperatures in the last decade. A net decrease of 63% in the water bodies was recorded over three decades for this class.

The settlement category saw the most expansion over the study period. From 1990 to 2000, 32% (557 km²) of land was converted into a settlement. Another 20% increase was seen from 2000 to 2009 as 455 km² of land changed into developed areas. It increased again from 2009 to 2020 by 17%. Overall, an 86% increase was recorded for this class. Most of the agricultural class and other classes were converted into a settlement over three decades.

From 1990 to 2000, this class decreased by 12%. Another 8% was decreased from 2000 to 2009. Another sharp decrease was seen from 2009 to 2020 by 33%. This decline was probably due to an increase in Settlement and forested land. Overall, the net decrease seen in three decades was 47%.

In 1990, the forest area calculated was 1,758 km². Total wood volume was 1,012 m³. Total DMB was calculated as 435 tons. Carbon stock for 1990 was recorded to be 309 tons per km² and its CO₂ equivalent was 1,134 tons. Forest cover decreased in 2000 (1,684 km²), hereby decreasing wood volume, which was 969 m³. DMB was calculated to be 416 tons. The carbon stock was 296 tons per km². This also resulted in decreased CO₂ equivalent which was 1,086 tons.

In 2009, forest cover decreased further, standing at 1,486 km². The total wood volume was 856 m³. Total DMB was 368 tons. Carbon stock for the year was calculated to be 261 tons per km². The CO₂ equivalent was 958 tons. This implied that land-use change considerably modifies carbon dynamics in the long run (Mannan et al. 2018).

Forest cover increased substantially in 2020, with 2,323 km² area covered with forests. Wood volume was calculated as 1,338 m³. Total DMB increased as well, which was 575 tons. Carbon stock increased greatly (408 tons per km²), resulting in increased CO₂ equivalent, which was 1,498 tons. Table 5 shows the carbon stock assessment for the selected years. In a similar study that focused on mapping deforestation trends in Western Himalayas, Pakistan, extensive deforestation was revealed in the area from 1990 to 2010. Nearly 170,684 ha (1,706.84 km²) of the forest had been lost, or roughly 0.38% of forest land was degraded every year (Qamer et al. 2016).

Table 6 shows the results for annual emissions from forest degradation. From 1990 to 2000, 1.3 tons of carbon stock was degraded. From 2000 to 2009, 3.4 tons of carbon stock degraded. The negative sign here (Table 6) implies a decrease in carbon stock. From 2009 to 2020, carbon stock increased, which has already been explained above. Between 1990 and 2000, the forest area degraded was 74 km², and the resulting annual emissions due to forest degradation were 352 tons. This further increased from 2000 to 2009, when degraded forest area increased to 197 km², and the annual emissions due to forest degradation were calculated to be 2,507 tons. From 2009 to 2020, an increase in carbon stocks was identified in the study area. This was probably due to increased plantations in the region.

In some cases, spectral signatures were confused with bare soil (Others) due to mixed pixels. This caused an underestimation of built-up area at some points and created a slight overestimation in the Settlement class at various sites. As rural settlements are not clustered at one point and are mostly found scattered in hilly areas, so it was much more difficult to separate the scattered residential areas from bare soil because of serious spectral mixing. Google Earth was used as a reference in such cases. Large urban settlements concentrated at one point such as Mardan city and some rural residential areas in plain valleys were easier to classify.

The analysis by using supervised classification in 1990, 2000, 2009 and 2020 suggested that the study area included different topographies (Forest land, Agricultural land, Water bodies, Settlement and Others) which changed over time. Image classification of LULC was carried out by making use of Ground truth points and satellite data for the AOI. The forest area declined by 4% during the period from 1990 to 2000 and 11.7% from 2000 to 2009; however, it increased considerably from 2009 to 2020 by 56%. An increase in the forest cover also resulted in an increased carbon stock from 2009 to 2020. There was a significant change in forest land, agricultural land and settlement due to higher human impact and BTTAP, in the case of an increase in forest land. The results indicated that the tree plantations have proven to be significant in promoting an increase in the carbon stock. If this pattern of new tree plantation continues, along with the protection of existing forests, it may ultimately help combat climate change in the future.

When medium-level spatial resolution data such as provided by Landsat is used, mixed pixels can be a frequent challenge that arises particularly for urban areas that are a diverse mixture of features mainly comprising of vegetation, trees, water, etc. Better quality images, if available, can result in a better classification process.

The findings of this analysis act as a base for studies in forest LULC change in the region and will help policymakers to improve policies to efficiently manage the land resources, i.e. reserving more land for forests and discouraging deforestation all over the country.

The study provides proof that the BTT plantation drive has been successful in increasing carbon stocks to an extent. Further plantations need to be done in the country to combat climate change and achieve carbon sequestration, as forests are the main source our country can currently rely on for this purpose. Management practices and precautionary measures should be taken to minimize the disastrous effects of human activities on forests and soil structure. Land misuse should be minimized, and restoration processes must be carried out for better crop production and to avoid/minimize forest degradation and soil erosion (Fahad et al. 2020).

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.