We propose a novel retrieval-augmented strategy for multi-resolution geo-embeddings, called RANGE. Our method is based on the intuition that the visual features of a location can be estimated by aggregating visual features from multiple similar-looking locations.

We developed Fields of The World (FTW), the largest global dataset for crop field boundary segmentation, to advance more accurate and generalizable ML models for agricultural monitoring.

In this review we found that while AI holds great promise for improving urban health research, there's a critical need to integrate spatial equity into these studies. Only 10% of the studies we reviewed explicitly considered spatial inequities, highlighting an important gap in the field.

TaxaBind is a suite of multimodal models useful for downstream ecological tasks covering six modalities: ground-level image, geographic location, satellite image, text, audio, and environmental features.

We analyzed three decades (1990–2020) of Landsat imagery to assess forest cover change and fragmentation in Azad Jammu and Kashmir, revealing gradual decline, increased patchiness, and critical hotspots to guide conservation and UN-REDD⁺ strategies.

We conducted the first global review of Species Distribution Modelling in mountainous environments (1997–2022), highlighting the growing role of remote sensing and machine learning in supporting forest management under climate change.

We develop a probabilistic, multi-scale, and metadata-aware embedding space that connects audio, text, and overhead imagery. This enables the creation of dynamic, multi-scale soundscape maps for any geographic region, along with uncertainty estimates for the mapping.

We extended deep learning approaches combining GEDI LiDAR and satellite imagery to predict forest vertical structural diversity across complex landscapes, advancing understanding of carbon sequestration, biodiversity, and ecosystem processes.

We present a generative model that synthesizes realistic, multi-temporal satellite imagery conditioned on location, text, or layout, enabling simulations of past, present, and future Earth surface scenarios.

We highlighted the importance of field boundaries as core geospatial datasets, essential for agricultural assessments, climate policies, and MRV of greenhouse gas emissions and sustainable land management.

We examined COPD incidence in Pakistan (2019–2020) using spatial scan statistics and machine learning, revealing pollution-linked clusters and highlighting the urgent need for air quality interventions.

We applied GIS, remote sensing, and AHP to identify suitable solid waste disposal sites in D.G. Khan, providing policymakers with a spatial framework to reduce environmental and health risks.

We developed GeoBind, a deep-learning model that uses satellite imagery to align text, image, and audio into a shared embedding space, enabling versatile multimodal reasoning about geographic locations.

We analyzed multi-year Landsat data to assess urban heat island dynamics in Punjab’s districts, showing how land cover change drives surface temperature variations and forecasting future scenarios with CA modeling.

We introduced Sat2Cap, a weakly supervised framework for zero-shot mapping that learns textual concepts from satellite imagery via contrastive alignment, enabling scalable multimodal mapping without text-labeled data.

We mapped groundwater potential zones in Lahore using ten geospatial parameters and weighted overlay analysis, providing a validated framework to guide sustainable agricultural water management.

We developed a species distribution model that integrates taxonomic hierarchy via large language models, enabling zero-shot range prediction and introducing a novel proximity-aware evaluation metric that outperforms existing baselines.

We analyzed multi-year Landsat data to assess water quality dynamics in Tarbela reservoir, finding long-term deterioration but signs of environmental healing in 2020, with policy implications for sediment and turbidity management.

We applied MaxEnt modeling with 26 climatic, biophysical, and topographic variables to predict the potential distribution of four native tree species in AJK, identifying hotspots for conservation and sustainable forest management.

We performed landslide susceptibility analysis in Astore, Pakistan using GIS, remote sensing, and AHP, identifying five risk zones with slope, lithology, and land cover as key triggering factors.

We performed landslide susceptibility analysis in Astore, Pakistan using GIS, remote sensing, and AHP, identifying five risk zones with slope, lithology, and land cover as key triggering factors.

We analyzed monthly streamflow trends across 13 HKH sub-basins in Pakistan using innovative polygon trend analysis, revealing increased flows in glaciated basins and widespread decreases downstream, with implications for floods, hydropower, and water management.

We highlighted the rapid urban sprawl of Lahore, showing significant loss of tree cover (1990–2017) alongside expanding built-up areas, underscoring urgent needs for urban greening and sustainable planning.

We analyzed 1990–2017 Landsat data with GEE and CART classification to map urban sprawl, LST, and land cover change in Lahore, revealing rapid urban growth, tree and green space loss, and rising temperatures, with implications for sustainable urban planning.

We assessed soil erosion dynamics across Pakistan (2005–2015) using RUSLE and six influencing factors, revealing intensified erosion driven by land use changes, with implications for targeted conservation policies.

We combined Sentinel-1 SAR and Sentinel-2 optical time-series data with a random forest classifier to achieve 97% accuracy in crop type mapping, demonstrating the power of integrating radar and optical imagery for precision agriculture.

We synthesized 73 studies (1993–2021) on forest mapping in Pakistan, highlighting the dominance of supervised classification methods, limited use of active remote sensing, and the need for advanced tools to support REDD⁺ and sustainable forest management.

We reviewed 44 global studies (2004–2019) on forest aboveground biomass estimation using high-resolution optical imagery, highlighting common sensors, modeling approaches, strengths, and gaps, with implications for REDD⁺ monitoring and real-time biomass assessment.

We assessed agricultural land suitability in AJK, Pakistan using GIS, AHP, and eight biophysical and climatic criteria, identifying only ~13% as highly suitable and highlighting priorities for sustainable land use planning.

We applied Change Vector Analysis with Landsat data (2000–2017) in Duy Tien, Vietnam to capture both categorical and radiometric land cover dynamics, revealing vegetation–soil index shifts and rapid urban expansion.

We integrated GIS and AHP to zone flood hazards in the Lam river basin, Vietnam showing that adding relative slope length as a sixth factor improved map accuracy and alignment with historic flood events.

We used Landsat-8 and ASTER DEM with GIS-based modeling to identify suitable tea cultivation zones in Mansehra and Abbottabad, finding ~13% highly suitable land to reduce Pakistan’s heavy tea imports.

We analyzed land use and land cover change along River Ravi in Lahore (1990–2017), showing tripled urban growth, loss of water bodies and green areas, and alarming biodiversity decline linked to urbanization.

We developed and field-tested a low-cost UAV system for 3D mapping and air quality monitoring in open-pit mines, demonstrating accurate topographic modeling and pollutant detection at Vietnam’s Coc Sau coal mine.

We analyzed the November 2016 smog event in Lahore using MODIS and Landsat imagery with GIS, showing dense smog cover severely reduced visibility to nearly 100 meters.

We designed three optimized observation lines for monitoring mining-induced deformation in Mong Duong colliery, proposing an improved method that adjusts the angle of draw to ensure accurate control points and compliance with Vietnam’s mine surveying standards.

We analyzed Landsat imagery (1990–2017) to assess Lahore’s urban expansion and tree cover loss, revealing that green depletion from 2010–2017 exceeded the decline of the previous two decades.

We quantified aboveground carbon storage in Ayubia National Park using field data, remote sensing, and GIS, estimating 252 tC/ha and demonstrating a scalable approach for carbon monitoring in Pakistan’s forests.

We developed a GIS-based framework for real-time water loss assessment and leakage detection in Faisalabad’s pipeline network, enabling accurate mapping of leak locations to support sustainable urban water management.

We applied GIS and AHP to identify and prioritize suitable landfill sites in Sahiwal, integrating land-use, environmental, and proximity criteria to ensure sustainable and eco-friendly solid waste management.

We analyzed Punjab University’s transit system using GIS-based service area modeling, revealing that only 26% of users live within an 8-minute walk to bus stops, highlighting critical gaps for future transport planning.

We conducted drainage morphometry analysis of the Haro River basin using DEM and ArcGIS, identifying dendritic to semi-dendritic patterns and elongated basin shape with low flood potential.

We assessed accessibility of 11 neighborhood parks in Sheikhupura using GIS network analysis, finding they serve less than 11% of the population and highlighting gaps in spatial distribution for sustainable urban planning.

We conducted drainage morphometry analysis of the Haro River basin using DEM and ArcGIS, identifying dendritic to semi-dendritic patterns and elongated basin shape with low flood potential.