Figure 2. Map of the 1114 municipalities of the Caatinga region showing variation in forest cover (a) and food security index (b) between the years of 2006 and 2017.
Also, the correlation between forest cover and food security was complex and spatially structured (all models performed better with spatial error) and, as expected, presented a boom-bust pattern described by a weak but significant quadratic function for both years (for 2006; z = -2.40, p=0.016 and for 2017; z = -3.13, p = 0.001). Briefly, food security tends to increase in mid-deforested and reaches its peak around 50% of deforestation when it comes to drop again to similar levels of food security found in highly forested municipalities. A deeper analysis suggests that this pattern is mostly driven by economic poverty that presents a u-shaped curve in response to forest cover for both years (Fig 4.) However, the second most important dimension (PC 2) of food security presented a direct relationship with forest cover suggesting that inequality increases in more forested areas. As a result, low levels of food security can be grouped in two types: 1) poor people living in forested areas with social inequality and; 2) poor people living in deforested areas with smaller social inequality. In our study, the tradeoffs (gain-lose or lose-gain) group the poorest municipalities, but the ones losing forest and gaining food security are more unequal (in terms of income) while the ones gaining forest and losing food security are less unequal.