Accurately characterizing carbon stock is vital for reporting carbon emissions from forest ecosystems. We studied the estimation of biomass using Sentinel-2 remote sensing data in moist temperate forests in the Galies region of Abbottabad Pakistan. Above-ground biomass (AGB), estimated from 60 field plots, was correlated with vegetation indices obtained from Sentinel-2 image-to-map AGB using regression models. Furthermore, additional explanatory variables were also associated with AGB in the geo-statistical technique, and kriging interpolation was used to predict AGB. The results illustrate that the atmospherically resistant vegetation index (ARVI) is the best index (R² = 0.67) for estimating AGB. In spectral reflectance, Band 1(Coastal Aerosol 443 nm) performs better than other bands. Multiple linear regression models calibrated with ARVI, NNIR and NDVI yielded better results (R² = 0.46) with the lowest RMSE (48.53) and MAE (38.42) and were therefore considered better for biomass estimation. On the other hand, in the geo-statistical technique, distance to settlements, ARVI and annual precipitation were significantly correlated with biomass compared to others. In the stepwise regression method, the forward selection resulted in a very significant value (less than 0.000) for ARVI. Therefore, it can be considered best for prediction and used to interpolate AGB through kriging. Compared to the geo-statistical technique, the remote sensing-based models performed relatively well. Regarding potential sites for REDD+ implementation, temporal analysis of Landsat images showed a decrease in forest area from 8896.23 ha in 1988 to 7692.03 ha in 2018. Therefore, this study concludes that the state-of-the-art open-source sensor, the Sentinel-2 data, has significant potential for forest biomass and carbon stock estimation and can be used for robust regional AGB estimation with acceptable accuracy and frequent availability.

A changing climate and global warming have adversely affected Pakistan’s moist and dry temperate vegetation. Abies pindrow (fir) (Royle ex D.Don) Royle and Picea smithiana (spruce) Wall.) Boiss are the two major representative species of the moist and dry temperate forests in Northern Pakistan. The dendroclimatic study of both species is crucial for the assessment of climate variability at various spatial and temporal scales. This study examined the dendroclimatology of fir and spruce, and analyzed the growth–climate relationship along the latitudinal gradient. Two hundred and nineteen samples (ring cores) of the two species were collected from five different sites (Shogran (SHG), Upper Dir (UDS), Bahrain Swat (BSG), Astore Gilgit (NPKA), and Sharan Kaghan (SHA)) in Northern Pakistan. The cores were cross-dated, and chronologies were generated for the species and climatic data (precipitation, temperature, and Palmer Drought Severity Index (PDSI)) correlated with radial growth. The interspecies correlations for fir were calculated as 0.54, 0.49, 0.52, 0.60, and 0.48 for SHG, UDS, BSG, NPKA, and SHA, respectively, whereas in the case of spruce, the interspecies correlations were 0.44 for SHG, 0.55 for UDS, and 0.49 for BSG. Climate variability was observed in the samples of both species, which showed significant drought and humid years at specific intervals. With respect to the correlation between tree-ring width and climatic factors, a positive correlation was observed between fir growth and summer season precipitation, mean temperature, and PDSI in the spring, summer, and autumn seasons. Similarly, the growth of spruce was positively correlated with precipitation (in February, September, and May) and PDSI (in the summer and autumn seasons); however, no correlation was observed between monthly temperature and spruce growth. The relationship of fir and spruce growth with seasonal precipitation and PDSI showed a change from a negative to a positive correlation after 1980, following rapid warming. During the winter and spring, the correlation coefficient between fir radial growth and seasonal temperature showed an initial upward trend followed by a progressive decrease along with increasing latitude. Seasonal variations were observed regarding the correlation coefficient between spruce radial growth and increasing latitude (increasing in winter; a decreasing trend in spring and summer; an initial increase and then a decrease in autumn). In the same way, the correlation of seasonal temperature and PDSI with the radial growth of both species showed increasing trends with increasing latitude, except in the autumn season.

Climatic and topographic factors control the distribution of forests across the globe. The present study investigated the impacts of these factors on spatial distribution of sub-tropical (scrub and pine) and moist temperate forests in Pakistan. The study used Digital Elevation Model (DEM), Sentinel-2 images and climatic data to quantify the impacts of climatic and topographic factors on distribution of forests. The data was statistically analyzed using correlation coefficient (R), liner regression and decision tree. Results specified six forest types during stratification. These types were significantly related to topographic (elevation) and climatic factors. Correlation coefficient (R) indicated strong positive relationship with elevation (R= 0.92) followed by annual mean temperature (R= –0.76). Similarly, annual precipitation indicated positive relation with R value of 0.53. Stepwise linear regression model showed that elevation, precipitation seasonality and annual temperature range were strongly significant with overall R² of 0.85. Decision trees were developed to explore possible interactions of predictors to determine imperative factors. Results of decision trees of both growing methods (Chi-squared Automatic Interaction Detection (CHAID) and Classification and Regression Trees (CRT)) showed elevation was the most important factor that predicted particular forest type. Moreover, other factors such as temperature of the driest quarter, annual precipitation, precipitation seasonality and slope were identified as important factors in CRT. The present study concluded that forest types were strongly influenced by climate and topography. However, elevation was the best predictor, has significant relative importance and can be used for detailed forest stratification.

Remote Sensing (RS) provides the best ways to monitor temporal changes and to understand land use dynamics. Remote sensing analysis can be further enhanced when community perception regarding major drivers of change is integrated. The present study was an attempt to assess the land use land cover changes in the Ishkoman watershed in the Ghizer district. The study explored Landsat-5 and Landsat-8 images to assess the LULC dynamics from 1998 to 2018, and also used questionnaires for community perception regarding LULC changes in the past two decades. Supervised classification was used to monitor changes between 1998 and 2018 and the maximum likelihood technique was used to categorize the pixels into six classes: vegetation/forest area, bare rocks, water bodies, glaciers/snow area, rivers, water, and agriculture. Regarding the questionnaires, the correlation matrix and regression models were developed between independent variables (population, land type cleared, and extra land required for new family members) and dependent variables (land use dynamics factors and socio-economic variables). The results showed that all six land cover classes have shown temporal changes between 1998-2018 and the most significant change was observed in forests and pastures (which decreased from 18.7% to 5.9 %). Similarly, glaciers, water, rivers, and agriculture have changed from 13.1, 6.5, 9.3, 1.5 to 15.8, 4.0, 11.32, 3.1, respectively between 1998-2018. The largest change was observed in bare rocks which increased from 50.2 % to 60.06%. Moreover, temporal NDVI analysis showed a decrease in vegetation cover (conversion to bare rocks) between 1998-2018. The questionnaire results revealed that the highest correlation was shown between population increase and decrease in crop production (R2 = -0.348), whereas the lowest correlation was found in population increase and population access to bus stops (R2 = -0.167). Similarly, the highest correlation was found between access to roads and markets (R2 = 0.349) and dependent variable (land type cleared), whereas the lowest correlobserved in access to water resources (R2 = -0.021). The study concluded that land use land cover has been significantly changed from 1998 to 2018 in the Ishkoman Watershed. The study suggested more in-depth research to examine land use land cover changes at finer scales by using high resolution satellite imagery, and conducting details surveys regarding the underlying anthropogenic causes of land use dynamics.

In the context of the Bonn Challenge, Pakistan started Forest Landscape Restoration (FLR) in 2014. This study assessed growth performance and the survival rate of young plantations and developed linear regression models by using Landsat-8 data. The results showed that fast-growing species such as Eucalyptus camaldulensis and Robinia pseudoacacia have shown a good growth rate as compared to Pinus roxburghii and Cedrus deodara. Landsat-8 vegetation indices include Normalized Difference Vegetation Index (NDVI), Soil Adjusted Vegetation Index (SAVI), Modified Soil Adjusted Vegetation Index (MSAVI), Difference Vegetation Index (DVI), and Green Normalized Vegetation Index (GNDVI), which were correlated with volume (m³). RVI has the highest correlation with R² value of 0.88 followed by NDVI, SAVI, and GDVI with R² value of 0.83. Stepwise linear regression (SLR) showed that MASVI and SAVI have a strong significant relationship with volume compared to the rest of the indices. The simple linear regression model of RVI and volume has the lowest RMSE (1.19 m³/ha) and is considered the best for plantation mapping. The temporal assessment of afforestation (2013-2018) by Landsat-8 images showed that the plantation was successful in the sampled sites. The RVI differencing and threshold measured area under vegetation was 7,309.7 ha in 2013 and was increased to 9,224.9 ha in 2018. The study suggested that Landsat-8 data have the potential for monitoring FLR activities and can be enhanced further when combined with other datasets.

Environmental variables influence the spatial distribution, pattern and structure of vegetation in complex mountainous landscape along varied geographical conditions. This study explored the spatial distribution of four forest types across ecological gradient based on field data, climatic, topographic, and soil variables using stepwise linear regression (SLR), decision trees (DT), random forests (RF), and Maxent modeling. Results showed that climatic variables particularly annual precipitation, precipitation of warmest and coldest quarter have achieved the highest correlation (R = −0.9, 0.8 respectively) for forest types mapping and outperformed other explanatory variables (topographic and edaphic). Among the rest of variables, elevation (R = 0.6), sand contents (R = 0.8) and soil carbon (R = 0.6) contained useful information in order explain forest type spatial distribution. Analysis of regression models revealed that RF has achieved the highest correlation (R² = 0.923) and lowest RMSE 0.54, followed by the SLR model in which R² value has been progressively increased from 0.41 (error 2.02) to 0.917 (0.77) with respect four different predictors models, each separate developed for topographic (n = 5), soil (n-11), climatic (n = 11) and combined of all datasets (n = 27). DT showed that annual precipitation was the most important predictor for forest type classification with risk estimate of 0.412 (std error 0.31) and 0.478 (std error 0.52) for training and validation respectively. Maxent modeling showed impressive predictive performance of all forest types (STPF, MTF and DTF) along ecological gradient with average AUC values of 0.968, 0.918, and 0.940 respectively and climatic variables have highest gain compared to topographic and soil predictors. This study suggests that mapping of forest types through machine learning algorithms may be improved by incorporating other explanatory variables such as microclimate, soil types, nutrients, anthropogenic, demographic factors and spectral indices.