Quantifying Deforestation

Remote sensing can be a powerful tool for assessing the effectiveness of protected areas. In this analysis, we use Madagascar as a case study to quantify and compare deforestation rates between strictly protected and less protected parks.

First, let’s specify what we mean by ‘deforestation’ or ‘strictly protected areas,’ since these definitions often vary between studies. Here, forests are defined as areas with at least 50% tree cover in the year 2000, while tree cover loss is defined as occurring in areas that lost this original forest in the year 2010. Strictly protected areas are those with a IUCN category of Ia, Ib, II, III, or IV, and less protected areas are any that are not strictly protected.

We can use the Hansen Global Forest Change dataset (Hansen et al. 2013) to map tree cover and loss relative to park boundaries; the results are visualized in Figure 1. To explore the map further, click the button to open an interactive version.

The results of this analysis suggest that greater levels of protection may indeed be effective in reducing deforestation: in 2010, strictly protected parks lost 0.55% of their tree cover, while less protected parks lost 0.81% of their tree cover. However, when comparing such small percentage values, it is difficult to say if this discrepancy is significant or meaningful. The difference could be negated by variation in the way the analysis is structured – the scale of the pixels being aggregated, for example, or the selected years of original forest and tree cover loss.

We can also compare 2010 deforestation rates between each of Madagascar’s strictly protected areas. As seen in Figure 2, there is a great deal of variation in forest loss, from 0% in several parks to 6.5% in Beza Mahafaly Reserve, located in the southwest region of the country. Eklund et al. (2019) arrived at the same conclusion that this park experienced the greatest level of deforestation, but also found that the park’s management had a relatively high impact in avoiding deforestation, indicating that forest loss in this area would have been much greater in the absence of its protected area status.

There are certain limitations to our assumptions of park protection levels. For instance, parks with an IUCN status of “Not Applicable” or “Not Reported” were categorized as less protected areas, even though they may not necessarily have a lower level of protection. Out of 126 parks in Madagascar, 63 do not have an IUCN category, representing a large proportion of the target study areas. On the flip side of the coin, an IUCN designation of Ia, Ib, II, III, or IV does not guarantee that the strict level of protection is actually effective; the success of enforcement likely varies from park to park. Eklund et al. (2019) used protected area management effectiveness (PAME) rather than IUCN categorization as a proxy for protection level, which takes into account many factors to quantify actual management effectiveness, although they note that PAME data still may not be a good indicator of protected area management reality. These researchers found no association between protection level and deforestation rates within parks, suggesting that the percentage difference in our analysis is negligible.

Moreover, the definitions of ‘tree cover,’ ‘forest,’ and ‘forest loss’ used in this analysis may introduce certain biases into the conclusions. For example, excluding areas with less than 50% tree cover may ignore critical damage to more sparsely forested but still biologically important ecosystems. Setting the baseline level of forest at tree cover that was present in 2000 may also fail to account for areas of land that were rapidly reforested and then deforested again; Eklund et al. (2019) notably used a baseline year of 2005, which was different from our analysis. Finally, it is difficult to compare and aggregate data from protected areas representing very different ecosystems. As seen in Figures 2 and 3, some of Madagascar’s protected areas in the west contain very little tree cover, while others in the east are entirely composed of forest, and we have applied the same standardized forest cover analysis to each of them.

Overall, it is clear that while remote sensing can offer valuable insight into land cover changes across areas of differing protection statuses, it is critical to examine the structures and assumptions underlying these analyses to evaluate the validity and significance of our conclusions.

References

Eklund et al. (2019). What constitutes a useful measure of protected area effectiveness? A case study of management inputs and protected area impacts in Madagascar. Conservation Science and Practice, 1(10): 1-12. https://doi.org/10.1111/csp2.107

Hansen, M., et al. (2013). High-resolution global maps of 21st-century forest cover change. Science, 342(6160): 850-853. https://doi.org/10.1126/science.1244693