The present study was carried out in Palas valley (NW Himalaya), Pakistan, known for its intact tracts of forests and aimed at mapping and monitoring of spatial vegetation patterns using Geographic Information System (GIS), Remote Sensing and statistical modeling techniques. A regional vegetation/landcover map was developed at 250m resolution by classifying MODIS normalized difference vegetation index (NDVI) images of the year 2011 into six land-cover categories (Glaciers, Pastures, Conifers, Broadleaves, Shrubs and Built-up/agriculture) with an accuracy of >92%. Trends (progressive/regressive) in the regional land-cover conditions were then evaluated between the years 2000 to 2011. The statistical tests highlighted conifers as the most negatively affected land-cover type, whereas, built-up/agriculture land-cover types were dominated by progressive land-cover evolution. The tests further revealed that human population had significant role in modifying regional landcover conditions and associated fauna. A total of eight forest vegetation communities were determined in the valley through classification of floristic 1 data (2004-2007). These included Salix denticulata-Bergenia stracheyi-Geum elatum (SAL-BER-GEU), Betula utilis-Abies pindrow –Viburnum grandiflorum (BET-ABI-VIB), Picea smithiana-Abies pindrow- Viburnum grandiflorum (PIC-ABI-VIB), Juglans regia – Aesculus indica – Acer caesium (JUG-AES-ACC), Cedrus deodara-Quercus floribunda- Indigofera heterantha (CED-QUF-IND), Cedrus deodara-Parrotiopsis jacquemontiana – Pinus wallichiana (CED-PAR- PIN), Quercus baloot - Cotoneaster bacillaris – Cedrus deodara (QUB-COT-CED) and Quercus baloot-Olea ferruginea-Cotoneaster bacillaris (QUB-OLE-COT). The spatial distribution of communities was strongly correlated with elevation, aspect and heat load indices (p<0.05). In order to map the vegetation communities, generalized regression models with stepwise backward procedure were fitted for each community to determine its response against a set of predictor variables. The final models were then implemented in geographic information system to produce forest vegetation communities’ maps. The mapping results indicated that potentially 97231.5 ha of the study area (69.27%) is under forest vegetation out of which PIC-ABI-VIB community had greatest contribution (27.69 %). The forested area calculated for years 1992, 2001 and 2010 was 43886.61 (31.27 %), 38420.19 (27.37 %) and 33491.16 (23.86 %) ha respectively. The logistic regression models were used to determine the driving variables of forest/non-forest transitions. The results showed that locations near to roads but away from settlements were particularly vulnerable to deforestation and spatially concentrate in lower reaches. The simulation of forest cover up to year 2020 indicated that forest area may increase in future.