Abstract:
This research investigates the capabilities of remote sensing and AI-based analysis in the
context of deforestation and reforestation within flood-affected regions, highlighting the
impact of climate change on forest cover. The study utilized Sentinel-2 data to collect in formation for July-August across 2018-2022, and May-June and November-December of
2022 for pre- and post-flood analysis, respectively. Due to a lack of annotated data, the
study employed unsupervised learning techniques, including Kmeans++, and deep em bedded clustering, to analyze the data. Various features such as Normalized Difference
Vegetation Index (NDVI), Soil-Adjusted Vegetation Index (SAVI), Shadow Index (SI),
Advanced Vegetation Index (AVI), and Normalized Difference Water Index (NDWI)
were employed to identify deforested areas in Pakistan due to floods.
The results of the study revealed that Deep Embedded Clustering (DEC) outperformed
traditional machine learning methods like Kmeans++ and Kmeans, indicating its po tential for future research in this field. The November 2022 - December 2022 data was
also used to determine the potential areas of reforestation. The findings of this study
demonstrate the potential of remote sensing and AI-based analysis for deforestation
and reforestation in flood-affected regions. The results obtained indicate a reduction in
tree cover over the years, consistent with official figures reported by the UN and the
Pakistan Forest Institute. Additionally, our findings reveal fluctuations in forest cover
before and after the monsoon season in 2022, highlighting the importance of monitoring
and conservation efforts in flood-affected regions.
The study provides valuable insights for policymakers and environmentalists in the fight
against climate change, highlighting the importance of preserving forest cover and pro moting reforestation efforts. This research can serve as a foundation for future studies
in this field, paving the way for innovative solutions that address the challenges posed
by climate change.
