Forest Disturbance Mapping Using Radar Imagery.

Abstract

Tropical forests play a crucial role in mitigating global warming and addressing climate change. However, the increasing degradation of these forests due to different natural and anthropogenic factors necessitates the establishment of robust monitoring systems. While earth observation techniques offer a means to monitor large areas, cloud cover poses a challenge to optical systems in the tropics. In such circumstances, synthetic aperture radar (SAR) imaging turns out to be a useful tool for forest monitoring.

This study presents a methodology based on the Cumulative Sum algorithm, utilizing VV, VH, a combination of VV+VH, and the normalization of the same set of polarization of the Sentinel-1 GRD data for year 2022 in Gabon. The proposed method effectively detects the spatial and temporal occurrences of logging events in Gabon. The achieved overall accuracy of the method reaches 68.73%. Importantly, it demonstrates superior performance in identifying smaller disturbances compared to existing RADD alerts.

The findings of this study underscore the efficacy of the proposed method in detecting forest disturbances within tropical regions. This approach holds significant promise for enhancing monitoring efforts and supporting conservation initiatives in these critical ecosystems.

Keywords:Tropics, Forest disturbance, CUSUM, Sentiel-1, Monitoring.

Aim

To evaluate the effectiveness of the Cusum method on Sentinel-1 data for detecting forest disturbance in the tropical forest of Gabon in 2022.

Objectives

1. To detect the areas of forest disturbance.
2. To evaluate the performance of the Cusum method.
3. To address the implications of the findings for forest management.

Research Question

How effective is the utilization of time-series Sentinel-1 Synthetic Aperture Radar (SAR) data with normalization of polarization in detecting forest disturbances in Gabon?

Study Area

Study site is in Gabon

Methodology

When we plot the backscatter intensity of Sentinel 1 SAR data in timeseries it is difficult to differentiate between disturbed and undisturbed pixels

Equation used to calculate CUSUM of timeseries data

Equation used to calculate CUSUM for normalized polarization data

Following figure shows how CUSUM helps in differentiating disturbed and undisturbed pixels

The diagram below represents the methodological steps followed in this research project

Results

The resulting canopy cover changes detected by the CUSUM approach are presented in the figure below. .

Table below provides a comprehensive analysis of the accuracy achieved using different parameters and polarization combinations for felled tree detection.

Conclusions

In conclusion, this study employed a method that effectively detected annual forest loss resulting from forest disturbances in Gabon using time-series Sentinel-1 Synthetic Aperture Radar (SAR) data. The utilization of CUSUM values derived from backscatter data in time-series analysis demonstrated the radar's capability to detect structural changes occurring in the forest canopy. The innovative approach of normalizing polarization data yielded improved results compared to the direct utilization of polarization data, however the improvement was not statistically significant. The overall accuracy achieved for VH polarization was 67.30%, while for normalized VH polarization, it was slightly higher at 68.44%.

Notably, the algorithm employed in this study outperformed the currently available RADD alerts by detecting a greater number of disturbances. The results of our tests indicate that RADD alerts outperform CUSUM in near real-time detection due to its capability to detect variations with a shorter baseline. RADD alerts are particularly valuable for timely detection of forest disturbances. On the other hand, CUSUM excels in generating yearly change maps, providing a comprehensive view of forest disturbance patterns over longer time frames.

Future research should focus on using CUSUM method with forthcoming L-band NISAR and P-band BIOMASS SAR data. Furthermore, researchers can explore the applicability of this method in detecting forest regrowth and its adaptability across various forest types. The production of yearly canopy change maps facilitated by this method will offer valuable insights for forest conservation initiatives, such as the United Nations' REDD+ program.

References

Aquino, C. et al. (2022) “Reliably mapping low-intensity forest disturbance using satellite radar data,” Frontiers in Forests and Global Change, 5. Available at: https://doi.org/10.3389/ffgc.2022.1018762

Ruiz-Ramos, J. et al. (2020) “Continuous Forest monitoring using cumulative sums of sentinel-1 timeseries,” Remote Sensing, 12(18), p. 3061. Available at: https://doi.org/10.3390/rs12183061

Ygorra, B. et al. (2021) “Monitoring loss of tropical forest cover from sentinel-1 time-series: A CUSUM-based approach,” International Journal of Applied Earth Observation and Geoinformation, 103, p. 102532. Available at: https://doi.org/10.1016/j.jag.2021.102532.