1 Introduction

Besides being the lungs of the atmosphere, forests are vital for economic development as well as livelihood sustenance for traditional rural communities particularly those living in forest fringe areas (Kindermann et al., 2008; Ullah et al., 2022a). Globally, forests provide a diverse range of services and livelihood support to almost half of the population in the form employment, food, water conservation, fuel wood and wood for construction, grazing grounds, medicinal herbs, and contribution to local and national economies, while over 1 billion people directly depend on forests (Khan & Khan, 2009; Nazir et al., 2019; Schusser, 2013; Shen et al., 2006). Forests also play significant roles in soil conservation by providing organic content and preventing erosion (Hayat et al., 2021). Moreover, forests are vital for sustaining biodiversity as they are home to 80% of amphibian species, 75% of bird species and 68% of mammal species as well as breeding grounds for fish and shellfish (FAO, 2020).

As of 2020 estimates, 4.06 billion hectares or 31% of the total land area of the earth is covered by forests, which decreased from 4.34 billion hectares in 1990 (FAO, 2020). Pakistan experienced the second highest annual deforestation rate among Asian countries, 2.9%, between 1981 and 1990 (Tole, 1998), which reduced to 1.7% per annum between 1990 and 2000. It recorded an increase to 2.1% in 2005 and again to 2.4% between 2005 and 2010. According to the FAO, any country with less than 10% of the land covered with forests is considered environmentally unstable; Pakistan is therefore in an alarming situation having only 2.5% forest cover and a deforestation rate of 2.1% (FAO, 2010). The impact of deforestation is enormous not only on the biophysical environment but on the socio-economic conditions of the people inhabiting the forested regions (Ahmad et al., 2018; Haq et al., 2019; Hayat et al., 2019; Irshad et al., 2015; Mujahid & Minhaj, 2020; Nazir & Ahmad, 2016). Drastic changes have been witnessed in the past 3 to 4 decades in association with deforestation including ecosystems destruction, damage to natural habitats, soil erosion, increasing intensity and frequency of flash floods, severe and persistent dry weather, soil erosion, tremendous decline in water resources, abandonment of traditional mixed mountain agriculture, and outmigration (Hayat et al., 2021; Hussain et al., 2018; Zeb et al., 2019).

Understanding the severity of the situation, the Government of Pakistan and local communities have been launching a number of afforestation campaigns in the recent past. Social forestry programs and community efforts to restore forests extending from the 1970s through the early 2000s in different forest divisions are worth mentioning (Baig et al., 2008; Khan & Khan, 2010; Kingdom of the Netherlands & NWFP Forest Department, 1997). However, the most important forest restoration program was initiated by the Government of Khyber Pakhtunkhwa (the Northwestern Province of Pakistan) when a new political party, the Pakistan Tehreek-e-Insaf gained majority control of the provincial government for the first time in 2013 (Sajid, 2016). This project, called the ‘Billion Trees Tsunami Afforestation Project’ (BTTAP), was initially restricted to the Khyber Pakhtunkhwa Province only. The term ‘Tsunami’ was introduced by this political party to describe their political rallies as comprising a ‘tsunami of people.’ It became a hallmark of the party, and therefore, the biggest forest restoration projects in the history of the country were likewise referred to as ‘Tree Tsunami Projects.’ In 2018, this party took over the Federal Government as well and launched another larger program called the ‘Ten Billion Trees Tsunami Afforestation Project,’ with a nationwide coverage.

The first phase of the BTTAP started in 2014 followed by another phase in 2016 with a total allocated budget of around PKR. 10,000 million. Besides planting different tree species over an area of 263,153 hectares and aerial seeding over an area of 23,096 hectares, another 306,983 hectares area was allocated for natural regeneration through imposing a ban on any kind of wood harvest and grazing (Government of Khyber Pakhtunkhwa, 2014, 2016, 2017). The provincial government also started aerial surveillance to monitor any violations of the ban on wood harvest and grazing. The objectives of the BTTAP were not limited to new plantations only but also focused on the restoration of deforested areas through imposing bans on grazing, commercial logging, and minimizing fuel wood harvest (Kamal et al., 2018). Ideally, natural regeneration of forests along with planting new trees would have a considerable positive impact on the restoration of vegetation cover.

In the past few years, there have been heated political debates on these reforestation efforts, but little to no scientific research to quantitatively assess the project outcomes. Therefore, we conducted this study to scientifically investigate the ground-level impact of these efforts in forest restoration using remote sensing data validated through high resolution Google Earth images along with field observations. Research on the forest resources of Pakistan is mostly focused on vegetation cover dynamics, causes and underlying drivers of deforestation, socioecological and environmental impacts of deforestation, and biodiversity (Ahmed et al., 2015; Ali et al., 2005; Shah et al., 2022; Tariq & Aziz, 2015; Ullah et al., 2022b). However, due to data limitations, issues related to forest restoration efforts, community and social forestry programs, and forest management and conservation are relatively under-researched. Qualitative and descriptive research pieces are available on the appraisal of the social forestry efforts of the late 20th and early 21st century; however, there is a lack of quantitative information on these afforestation campaigns (Ali et al., 2020; Khan, 2001; Mansoor et al., 2018; Mulder & Rafiq, 1989; Siddiqui et al., 2018). Likewise, the recent afforestation efforts under BTTAP—the focus of this research—are the latest development in forest history, which motivated us to bring it under scientific research.

2 Methods and material

2.1 Study area

Four administrative districts of the Malakand Forest Region in the Khyber Pakhtunkhwa Province namely District Malakand, District Lower Dir, District Upper Dir, and District Swat were selected for this study (Fig. 1). The study area is located on the Northwestern frontier of Pakistan, physiographically in the Eastern Hindukush Mountains divided into two parallel valleys, the Swat River Valley in the east, and the Panjkora River Valley in the west. This region has been one of the main forest hubs of the country and home to two beautiful forest tourism sites, i.e., Kalam in the Swat District and Kumrat in the Upper Dir District. Several tree species can be found in these forests dominated by pine, deodar, and wild olive along with several shrub species. The selected study area holds considerable research potential on every aspect of environmental change because of its geographical location, history of biophysical and sociocultural changes, natural resource base, and forest cover dynamics. The vegetation cover of the study area has gone through several phases of expansion and shrinkage throughout history making it a favorite site to study the spatio-temporal trends and dynamics of forest resources (Khan, 2001). Additionally, this area has been home to experimenting various social, participatory, and community afforestation campaigns in the past (Church, 1993; GoNWFP (Government of North-West Frontier Province) 1995; Mulder & Rafiq, 1989).

Fig. 1
figure 1

Location and physiography of the study area. Inset map of Pakistan shows the study area outlined in green. The boundary shapefiles of the survey of Pakistan were used to create this map. Elevation is colored in the study area according to intervals given in meters above sea level in the provided key. Elevation data were downloaded from https://www.diva-gis.org/Data

Full size image

In the past few decades, this area has been exposed to high rate of deforestation. There are several factors behind the high deforestation rate including poverty, high dependence of the forest fringe population on forest products, illegal timber harvesting, ownership and property rights, unmonitored fuel wood extraction, and clearing forests for the construction of settlements (Rahman et al., 2014). Another major factor that has played a critical role in deforestation is the political change occurred in this region during the late 1960s. Before 1969, this area was under feudal rule and was kept detached by the rulers for their own use. Therefore, the natural environment remained largely intact as there were no link roads and construction activities were limited. When the feudal rulers were captured and this region was merged into the State of Pakistan, large scale development activities were started. Link roads were constructed, and built-up areas started to expand resulting in unprecedented land use land cover changes in the region (Haq et al., 2018).

3 Data and procedures

To better understand the history of vegetation cover dynamics in the study area, and the impact of the trees’ tsunami projects, we performed analysis on satellite imagery spanning about 30 years (1990 to 2021). Normalized Difference Vegetation Index (NDVI) was used to identify vegetation within each satellite image to analyze forest cover change through the study period. NDVI is recognized as one of the best techniques to study forest cover change and has been widely used by researchers throughout the world (Chu et al., 2019; Gao et al., 2021; Haq et al., 2021; Liang et al., 2018; Mane et al., 2022; Othman et al., 2018). This index is a dimensionless, radiometric measure that differentiates the abundance of photosynthetic activity of the landscape. This index is computed in each pixel by dividing the difference between the surface reflectivity values of the near-Infrared and the red electromagnetic bands with the sum of both (Jensen, 2015).

$${\text{NDVI}}=\frac{{\rho }_{{\text{nir}}}-{\rho }_{{\text{red}}}}{{\rho }_{{\text{nir}}}+{\rho }_{{\text{red}}}}$$

As a result, NDVI values span gradually from − 1 to + 1 (non-vegetation to most vegetation). In this study, we used the 32-day, Tier 1, Landsat NDVI, image collections to determine the dynamics of vegetation cover of the study region. The images were acquired from the Google Earth Engine (GEE) platform from 1984 to 2011 (Landsat 5), 1999 to 2003 (Landsat 7), and 2013 to 2021 (Landsat 8). Final NDVI images were the result of the median value of all images for each period. The median value of each year is acquired using GEE in order to avoid redundancy and seasonal impact on vegetation cover. Further the median value in each year also eliminates deciduousness issue in a dry and wet season. Furthermore, the study site is covered by Landsat images from path 151 and rows 35 and 36. Any scenes having more than 10% cloud cover were eliminated from the study. The removal of scenes with more than 10% cloud instead of atmospheric correction was opted to reduce distortions caused by obscured data. Moreover, it helped to overcome the constraints of time, resources, and limited data availability.

Finally, the NDVI results were classified into four major landcover categories based on threshold values according to the need of the study. The threshold values were validated through high resolution Google Earth images (Fig. 2) and field observations. All NDVI values less than 0.2 were classified as other land use categories (built-up area, roads, and barren land). Grasses, shrubs, and agricultural land were grouped into one category with NDVI values ranging from 0.2 to 0.35. Scattered trees and new forests were identified as values ranging between 0.35 and 0.5, while above 0.5 were classified as healthy forests. Temporal Google Earth images also helped identify and validate changes in forest cover through the study period. Several factors, including logistical constraints, such as time, budget, and the vast geographical area, posed challenges for conducting detailed ground truthing across the entire region. Nevertheless, to double check our results, we did collect a limited number of control points from convenient location. However, the availability of high-resolution Google Earth Images has made it easier to validate the results for a vast spatial area. Therefore, instead of the limited number of ground control points, we preferred to mention Google Earth as our main validation technique. Further the results were also compared with the areas of afforestation program and surveys conducted by Ten Billion Tree Tsunami GIS team for further validation.

Fig. 2
figure 2

Time series of Google Earth images. Images were downloaded from Google Earth to validate the change in forest cover through the study period

Full size image

Furthermore, the kappa coefficient was computed using 130 to 150 random samples for each land cover category. Around each sample point, a 20 × 20 m grid area was examined for accuracy assessment. The category of healthy forest area was further validated and categorized based on their covered area extent. The area with more than 10% canopy was categorized in this class and validated through kappa coefficient. The results of accuracy assessment test are given in Table 1.

Table 1 Accuracy assessment for land cover categories through the selected years
Full size table

4 Results and discussions

Accuracy assessment results (Table 1) show that the overall accuracy in 1990 was 83.16% while the kappa coefficient for the same year was 0.73, whereas for 1995, the overall accuracy was 83.95% with kappa coefficient of 0.75. Similarly, in 2000, the overall accuracy was 85.9%, and the kappa coefficient result was 0.78. The 2010 and 2015 overall accuracy was 88.44% and 90.21% whereas kappa results were 0.84 and 0.89, respectively. For the year 2021, the overall accuracy was 91.3%, and the kappa coefficient was 0.91.

Comparing the image classifications over time reveals a dynamic forest cover with different phases of alternating deforestation and reforestation. The initial epoch (i.e., 1990) features a relatively satisfactory extent of vegetation cover in the study area. It can be seen in Fig. 3 that more than half of the land (51%) was green cover with 20% healthy forests and 31% sparse vegetation. Grasses, shrubs, and agriculture covered 26% of the total land, while the remaining 23% was under different land cover categories such as built-up area and bare rocks. A previous study comparing forest cover of 1970s and 2014 in one of the districts of the study area found the situation of forest cover even better during 1975. According to the findings of that study, forests covered 57% of the land area of District Upper and Lower Dir, while 17% of the area was under pastures and rangelands (Haq et al., 2018). Therefore, it can be safely implied that forest cover of the study area decreased by around 6% in the study area during the 2 decades 1970s and 1980s.

Fig. 3
figure 3

Vegetation cover dynamics 1990 through 2000. The maps were created using data from NDVI analysis in GEE

Full size image

The following decade spanning epochs of 1995 and 2000 show a concerning decrease in vegetation cover area. Healthy forest cover decreased from 20% to only 3% during the 5 years from 1990 to 1995 and sparse vegetation or scattered trees cover decreased from 31 to 20%. Grasses, shrubs, and agriculture land increased by 8% while other land cover i.e., built-up land and bare rocks, recorded an increase of 21%. The trend continued with further decrease in vegetation cover during the following 5 years from 1995 to 2000. Continuous increase in the category of other land use indicates that mostly forest cover area has been converted to barren land or used for settlements and infrastructure. This pattern of conversion of forest cover to barren land was also found out by another study by Haq et al. (2019).

Several reforestation campaigns were launched during the 1980s and 1990s to encounter deforestation. The first national forestry project was launched in 1985 jointly funded by the government of Pakistan and the United States Agency for International Development (USAID). Likewise, the Government of Pakistan launched another Social Forestry Program in all the provinces of the country. However, most of these forestry efforts were focused on plantations, while due attention was not paid to a holistic forest management and protection. As a result, deforestation continued despite reforestation efforts (Qamar et al., 2016).

Compared to the situation in 2000 (Fig. 3), the healthy forest cover decreased from 3 to 2% in 2010. Other studies also confirm that during the period between 1980 and 2010, extensive deforestation and loss of forest cover have taken place (Khan et al., 2019). However, the scattered trees cover increased from 17 to 21%. Similarly, a considerable decrease in the category of other land cover and increase in grass, shrubs, and agricultural land were positive signs of vegetation regeneration in the study area. This positive change can be attributed to several factors, but the most important role was played by social forestry programs and forest management campaigns of the late 90s and early 2000s. The policy shift from ‘reforestation’ to ‘holistic forest management’ incorporating forest surveillance, collective community efforts, and a ban on grazing and forest harvest resulted not only in a faster regeneration but also reducing the rate of deforestation (Ali et al., 2006; Schwennesen, 1996; Sheikh, 1997).

Generally, forests in this region are owned by the local community under different kinds of ownership regimes with minimal interference from the state authorities. The ownership regimes and tenure arrangements vary locally from village to village as forests in some areas are considered as common property resources while in other areas, they are held under individual ownership (Rahman et al., 2014). Most of the earlier afforestation campaigns were not up to the mark because those initiatives did not feature community participation. Although the forest restoration efforts undertaken during the 2000s were also initiated by the government, this time the local community was considered as major stakeholders. The sense of agency experienced by the local people in these afforestation programs motivated them to fully participate in planting activities and conservation measures. Incentives were provided to the local populations through engaging them as paid labor in planting activities. Individual landowners were provided free seedlings from state nurseries to plant fast growing trees on their property. People were encouraged to replace the local goat breeds—which are less productive and more harmful for rangelands—with state provided genetic breeds. Forest guards and shepherds were appointed and paid by the government from the local population. These community participatory afforestation programs proved to be very affective in reclaiming the deforested areas (Baig et al., 2008; Khan & Khan, 2009).

In the following 5 years, forest cover continued to grow. Natural regeneration in the forests of the Hindukush mountains is of enormous significance as trees grow fast only if protected from overgrazing and harvesting. Healthy forests recorded a growth of 9% during the years 2010 to 2015, while scattered vegetation, which is dominated by natural regeneration and new plantation, increased by 11% in the same period. The other land cover category decreased from 36 to 29% which implies that a considerable portion of the barren land was converted to vegetation cover (Fig. 4).

Fig. 4
figure 4

Vegetation cover dynamics 2010 through 2021. The maps were created using data from NDVI analysis in GEE

Full size image

Since 2015, vegetation cover of the study area has been growing fast. Healthy forest cover has increased from 11% in 2015 to 35% in 2021, while all the remaining three categories have decreased. The trend shows a speedy conversion of new forests, barren land, and shrubs cover to healthy vegetation cover.

In 2013, when the new government took over and announced the Billion Tree Tsunami Afforestation Project, a total of 267 km2 land in the study area was specified for reforestation, shown in red color (Fig. 5). The areas in the red colored polygons were considered for the plantation of seedlings and areal seeding. The total increase recorded in healthy forest cover in the study area by 2021 is 1951 km2. A considerable growth in forest cover is evident around the specified sites since the first launching of the project.

Fig. 5
figure 5

Forest cover change since the first phase of the project around the specified afforestation sites. Data for afforestation sites were acquired from https://thebilliontreesproject.org/

Full size image

As a synthesis of dynamics, the history of vegetation cover changes in the study area since 1990 has been quantitatively summarized as areas of subclasses for each epoch in Fig. 6 and Table 2. The trend line fit to healthy forests subclass shows a drastic decline during the first 20 years of the study period and a considerable increase in the later 10 years. Healthy forest cover within the study area started at 2249 km2 in 1990 then decreased during the ensuing 2 decades to a minimum of only 265 km2 in 2010. The social forestry campaigns of the early 2000s and the billion trees afforestation program of the provincial government in the 2010s resulted into a substantial increase in healthy forest cover, which is presently about 4000 km2, substantively higher than 1990.

Fig. 6
figure 6

Overall quantitative comparison of classified vegetation coverage within the study area plotted by area of subclasses distinguished by colored bars for each epoch of available imagery from 1990 through 2021. A polynomial trend line is fit to the healthy forests subclass

Full size image
Table 2 Temporal change in vegetation cover from 1990 through 2021
Full size table

The areas which were cleared during the first four phases of the study period i.e., 1990 to 2010, started to recover in 2015. Healthy forests around the villages on gentler slopes and in the vicinity of the low-lying valleys were particularly damaged during the deforestation era. Studies suggest that this trend was mainly because of the expansion of built-up areas triggered by population growth and household dynamics in the 1990s and 2000s (Haq et al., 2019). Moreover, the construction of roads during this period made commercial timber harvest easier and resulted in reckless exploitation of forest resources (Haq et al., 2021). Another factor behind this mass deforestation was the shift in land tenure systems and change in ownership regimes during the deforestation era (Rahman et al., 2014).

During the social forestry campaigns and particularly after the launch of the BTTAP, extensive plantation was carried out over the rangelands and barren slopes around the villages. Moreover, the imposition of bans on commercial timber harvest and restrictions on fuel wood collection and grazing, which was made possible through community participation, resulted in more rapid regeneration of forests after 2010. However, it is observed that when the plants grow up to a certain extent, the government stops paying attention to surveillance and protection of vegetation cover. The same happened during the social forestry programs of the 1990s and early 2000s, when the newly planted areas were kept under surveillance by forest guards and shepherds for a period of 5 years to permit the plants to grow enough to survive uprooting by grazing animals. With the withdrawal of the state-imposed ban and protection measures, the newly grown forests are exposed to overgrazing and extraction of forest products, particularly in the areas where communal ownership is practiced. As hypothesized by Hardin (1968) in his famous ‘tragedy of the commons' metaphor, these forests face ‘competition for more benefits’ from the users, with inadequate attention to conservation or sustainability. Hence, it is feared that the relaxation of bans on cutting and grazing may result in a setback and renewed reduction of forest coverage.

5 Conclusion and policy recommendations

With the development of transportation network and demographic and sociocultural changes, the vegetation cover of the study area was exposed to an unprecedented rate of degradation. Healthy forest cover decreased from 20% in 1990 to only 2% in 2010, while scattered trees cover or sparse vegetation decreased from 31 to 21% over the same period. Later, efforts were made to restore the vegetation cover through social afforestation campaigns. However, due to certain shortcomings, these initial campaigns were not very effective. Community participation was increased in subsequent afforestation efforts during the 2000s. The newly formed provincial government of the cricket star and Ex-Prime Minister of Pakistan in the province of Khyber Pakhtunkhwa in 2013 launched a new forest restoration campaign called the ‘Billion Tree Tsunami Afforestation Project’ that aimed to reclaim formerly deforested areas.

Unlike the previous poorly managed social forestry efforts, this newer reforestation campaign featured the government bearing the responsibility to restore the greenery of the country without conflicting with the local sociocultural dynamics. Hundreds of nurseries were established to plant seedlings across the Khyber Pakhtunkhwa Province in the first phase and nationwide in the second phase. Besides government’s planting, seedlings were provided free of cost to individual landowners in the study area encouraging people to grow and protect their own resources. Large scale areal seeding was undertaken across the entire Hindukush belt. Strict bans were imposed on commercial wood harvesting. It is important to mention here the role of community response to this project which ultimately played a key role in its success. With the limited available resources, it would have not been possible for the government to protect all newly forested areas against grazing and fuelwood harvest without a supportive public response.

Based on the current spatial analyses of vegetation cover, it can be safely concluded that independent of political claims, these projects have been quite successful in restoring the green cover of Pakistan. Recently, considerable improvements have been recorded in vegetation cover of the study area. Our results show that healthy forest cover has increased from 2 to 35% after the launching of the billion trees projects. Scattered vegetation has also increased from 21 to 29%. Compared to 1990, the total green cover of the study area has increased from 51 to 64%. The surveillance and protective measures instituted under these projects enhanced the regeneration of existing vegetation besides expanding forest cover to the barren areas. Grazing grounds and rangelands around villages have also been regaining green cover. However, despite considerable success in restoring forests, these projects are proving to be a short-term political policy without evident legislation to establish them within a long-term state policy. Moreover, with ongoing political change and shuffling of administrative structure, the future of the billion trees afforestation program seems uncertain. It is therefore feared that the recent increase in forest cover may be adversely affected due to relaxation of restrictions on the collection of forest products and interruption of the forest restoration campaign.

We suggest to shift the focus of national forest projects from being politically interested campaigns to being issues of national concern secured as policy through legislation. From the National Forest Policy 1955 to the Provincial Forest Policy 2001 and National Forest Policy 2015 (Hassani, 2021), these policies have been changing as frequently as the governments have been changing. The point is, how can such unsustainable forest policies ensure sustainable forest management? Every few years, we have a new forest policy formulated by short-lived political governments and donor organizations. It should be a contemporary priority to bolster the political will to deal with deforestation as a matter of national interest regardless of change in selected government. Moreover, instead of government and donors driven objectives, forest policy should be founded on the recommendations of scientific research conducted by researchers and academia. Large and growing populations in rural and mountainous areas depend on forest resources, and their well-being should be considered central to any forest management policies. The state’s effort to change forest ownership can be counterproductive to more effective community-level engagement, and therefore, flexible policies need to be formulated that are responsive to local sociocultural conditions and community interests. Reforms are also needed in the forest department at district, provincial, and national levels to ensure corruption-free implementation of forest policies and conservation measures.