The mangrove in Madagascar contributes to 2% of the global volume. However, in the past two decades, the deforestation of mangrove became a major concern in the north-western Ambanja and Ambaro Bays (AAB), one of the largest mangrove ecosystems in Madagascar. The mangrove ecosystem was degraded or deforested due to anthropogenic reasons (Jones et al., 2016). To understand how the deforestation changed over time, I created an animation (Fig. 1) to show the expansion of the mangrove deforestation areas from 2000 to 2013.

The original data, mangrove deforestation shapefiles, were provided by Trevor Johns. I plotted the shapefiles using matplotlib python module and exported the images with JPEG format. The animation was created from these JPEG files using imageio module. The result shows that the areas of mangrove deforestation kept increasing over time from 2000 to 2013.

 

Reference:

Jones, T., Ratsimba, H., Carro, A., Ravaoarinorotsihoarana, L., Glass, L., & Teoh, M. et al. (2016). The Mangroves of Ambanja and Ambaro Bays, Northwest Madagascar: Historical Dynamics, Current Status and Deforestation Mitigation Strategy. Estuaries Of The World, 67-85. doi: 10.1007/978-3-319-25370-1_5

Introduction

The boreal forest is located in Arctic and sub-Arctic regions. It contributes to 32% of the global forest and 77% of Canada’s forest. The removal of boreal forest vegetation will increase the atmospheric concentrations of CO2 and affect the climate indirectly. Forest can be removed by three major causes, forest fires, clearcut, and insect outbreak. These disturbances reduce forest vegetation in different severities and extents, resulting in different recovery patterns of the vegetation. The short-term recovery or initial recovery is the most important stage because it has the greatest vegetation growth rate of the ecosystem over the recovery period and determines the long-term ecosystem composition under stable dynamics.

Study Area

The study area is East Tenakihi Range in northern BC, Canada (Fig. 1). It has an area of 110,441 hectares. In 2006, forest fires, clearcut, and insect outbreak all occurred in this region, which makes it an ideal site to compare the post-disturbance recovery rates.

Methods

The overview of the method is shown in Figure 2. The analysis started with data acquisition. I used the Landsat Best Available Pixel composite from IRSS lab and analyzed the Landsat data of each year from 2005, one year before the disturbance, to 2011, five years after. The second step is to crop the Landsat images. I created the boundary by drawing a polygon in Google Earth Pro based on the terrain features and used the boundary to crop the Landsat data for each year. The third step is to extract the distributions of vegetation indices for each disturbance group. I used Normalized Burn Ratio (NBR) to evaluate the vegetation health in the areas affected by fires and clearcut and Normalized Difference Moisture Index (NDMI) for insect outbreak. NBR is designed to evaluate fire severity, and NDMI to measure the moisture content of the vegetation. I also used dNBR and dNDMI to compare the annual recovery. The images need to be classified into disturbance groups. I used two classification algorithm, maximum likelihood algorithm for fire and clearcut classification and decision tree for insect outbreak classification. The last step is to compare the differences in regrowth rates among the three groups. I performed time-series statistical analysis for the distributions in each year. I used ANOVA analysis to compare all the three groups and t-test for any two groups. Before the statistical analysis, I resampled the indices by bootstrapping to reduce computation complexity while not affecting the overall vegetation health parameters. This process calculated the mean of 500 randomly selected samples for 1000 times. I calculated the percentage of recovery by the ratio of the difference in the index value between the year of interest and 2005 over the difference in the index value between 2005 and 2006. The percentage of recovery was then used as the input of the statistical analysis. For each year from 2007 to 2011, the ANOVA analysis was performed with the input of all the three groups, and the t-test was performed for any two groups. The classification results are shown on the map.

Results

The classification results are shown on the map (Fig. 3). Fire and clearcut affected 1.3% and 0.3% of the total area respectively. However, the insect outbreak affected almost half of the study area, much more the other two types. The statistical analysis results indicate that there are significant differences in the recovery rates of the three disturbance groups in each year from 2007 to 2011 after the disturbances occurred.

The vegetation regrowth patterns are shown in the boxplots (Fig. 4). The regrowth in five years recovered vegetation health by 70% after fire and 50% after clearcut. However, the areas infested by beetles suffered 50% more vegetation loss in five years. Figure 5 and 6 imply how the vegetation recovered in each year by dNBR and dNDMI. The areas affected by fire and clearcut recovered the fastest in 2009 and 2011 respectively. But the infested areas experienced continuous disturbance after 2008.

 

Figure 5. The change of dNBR in burned and harvested areas over the study period (2006 – 2011).

 

Conclusion

The limitations of the study include the underestimation of the variance of NDMI during the evaluation of insect outbreak. And the classification results need to be verified by ground-truth data to examine and improve accuracy. To conclude, the vegetation recovered the fastest from forest fires, followed by clearcut harvesting and insect outbreak. Forest operations need to consider the differences in vegetation recovery management. NBR and NDMI may be applied to evaluate the vegetation recovery in other boreal forest areas, such as Yukon, Alaska, and eastern Russia.