Increased areas of ITs/PAs for forest conservation
The number and area of ITs/PAs in the BLA increased from 1980 to 2018 with noticeable characteristics (Fig. 1a). The ITs area rose slowly from 0.2 × 106 ha in 1980 to 4 × 106 ha in 1988 but started to increase rapidly after 1988 and reached 74 × 106 ha in 2000 and 115 × 106 ha in 2016. This large and rapid increase in numbers and areas of ITs was driven by the 1988 Constitution’s requirement that the government demarcate all Indigenous lands within 5 years. Although the Constitution’s requirement has yet to be fulfilled, it stimulated substantial expansion of ITs, including 40 × 106 ha demarcated with financial support from the G7 Pilot Program to Conserve the Brazilian Rainforest. In 2000, the Brazil’s National Protected Areas System was officially established to expand PAs and better manage the PAs for forest conservation. The nPAs area increased from 9 × 106 ha in 1980 to 27 × 106 ha in 2000, 62 × 106 ha in 2008 and 66 × 106 ha in 2018. The sPAs area increased from less than 1 × 106 ha in 1980–1989 to 25 × 106 ha in 2000, 56 × 106 ha in 2007 and 60 × 106 ha in 2018. When lumped together, the total area of ITs/PAs substantially expanded from 10 × 106 ha in 1980 to 126 × 106 ha in 2000 and 241 × 106 ha in 2018. The numbers of ITs, nPAs and sPAs also had similar changes over time and substantially increased from 2, 11 and 2 in 1980 to 227, 70 and 95 in 2000 and 387, 146 and 191 in 2018, respectively.
Fig. 1: Cumulative areas and numbers of ITs and PAs and cumulative forest areas in ITs/PAs in the BLA. a, Cumulative areas and numbers of ITs, nPAs and sPAs from 1980 to 2018. ITs data in 2017 and 2018 are not available. b, Cumulative forest areas in ITs, nPAs and sPAs from 2000 to 2018. Full size image
As most ITs/PAs are covered by forests, the large increase of ITs/PAs strengthens forest conservation substantially in the BLA. Figure 1b shows the interannual change in forest area in ITs, nPAs and sPAs during 2000–2018. Forest area in ITs increased from 67 × 106 ha in 2000 to 105 × 106 ha in 2016, an increase of 55%. Forest area in nPAs increased from 25 × 106 ha in 2000 to 62 × 106 ha in 2018, an increase of 146%. Forest area in sPAs increased from 18 × 106 ha in 2000 to 50 × 106 ha in 2018, an increase of 169%. We overlaid the 2000 annual forest map (a total of 394 × 106 ha of forest) with the 2018 boundary maps of ITs/PAs and these ITs/PAs in 2018 covered 206 × 106 ha forest, accounting for 52% of the total forest area in 2000, which clearly shows the importance of the ITs/PAs for forest conservation.
Interannual change of forest area during 2000–2021
We used our annual forest maps to quantify the interannual changes in forest area during 2000–2021 (Fig. 2a). The total forest area in the BLA decreased substantially from 394 × 106 ha in 2000 to 366 × 106 ha in 2021, a loss of 28 × 106 ha (~7% of the forest area in 2000 or an area loss larger than Brazil’s state of Rondônia). The 2018 ITs/PAs boundary maps were overlaid on the annual forest maps to calculate the interannual change of forest area in ITs/PAs during 2000–2021. Total forest area in ITs, nPAs and sPAs decreased from 105.7 × 106 ha, 62.6 × 106 ha and 50.5 × 106 ha in 2000 to 104.9 × 106 ha, 62.2 × 106 ha and 50.0 × 106 ha in 2021, respectively, with the average loss rates of 0.04% yr−1, 0.03% yr−1 and 0.10% yr−1 over their forest areas in 2000 (Fig. 2b–d). As some of the ITs/PAs overlap each other, we combined the ITs/PAs together; the total forest area in the combined ITs/PAs decreased slightly from 206.4 × 106 ha in 2000 to 204.9 × 106 ha in 2021, an average annual loss rate of 0.08 × 106 ha yr−1 (0.04% yr−1). To our surprise, ITs/PAs together had a slightly larger forest area in 2018–2021 (204.6 × 106 ha) than in 2014–2017 (204.1 × 106 ha), which may be related to the recovery of forests after severe damage in 2015/2016 and, to a much lesser extent, from tree-planting projects. In comparison, the forest area in the non-PAs (Fig. 2e) decreased from 187.8 × 106 ha in 2000 to 161.3 × 106 ha in 2021, with an average annual loss rate of 1.3 × 106 ha yr−1 (0.7% yr−1), about 14 times more than the ITs/PAs. The loss of 27 × 106 ha forest area during 2000–2021 in the non-PAs accounted for ~95% of the total loss of forest area in the BLA over the same period.
Fig. 2: Interannual changes of forest areas in the BLA from 2000 to 2021. a–e, Forest areas in the BLA (a), ITs (b), nPAs (c), sPAs (d) and non-PAs (e). f, Forest area map in 2000. g–i, Forest area change rates in ITs (g), nPAs (h) and sPAs (i). Forest area change rates are calculated from annual forest areas between 2000 and 2021 on the basis of linear regression analysis at the 90% confidence level. Full size image
Geographically, the interannual change of forest area in ITs, nPAs and sPAs from 2000 to 2021 had noticeable spatiotemporal patterns. The trend analysis of forest areas (Fig. 2f–i) showed divergent dynamics among ITs/PAs. ITs/PAs in the southern and eastern portions of the BLA (the ‘arc of deforestation’) had the largest losses of forest area. The ITs/PAs in the northern portion of the BLA either had no significant change in forest area or had increased forest area. In total, 39.8% of nPAs (59 out of 146 nPAs), 40.4% of ITs (154 out of 387 ITs) and 44.0% of sPAs (84 out of 191 sPAs) had significant forest area loss from 2000 to 2021 (P < 0.1).
Spatiotemporal dynamics of primary forest area loss
We used the 2001 forest map as the reference map and identified the first year the forest pixels in 2001 were classified as non-forest pixels during 2002–2021 (Fig. 3a,b) and we counted the number of pixels with a change from forest to non-forest (forest loss) in a year as annual gross forest area loss over the BLA, ITs, PAs and non-PAs from 2002 to 2021 (Fig. 3c–g). The cumulative gross forest area losses during 2002–2021 were 49 × 106 ha for the BLA, including 2.1 × 106 ha for ITs, 1.2 × 106 ha for nPAs, 2.8 × 106 ha for sPAs and 43.1 × 106 ha for non-PAs. The combined ITs/PAs had a 5.9 × 106 ha gross forest area loss from 2002 to 2021, accounting for ~12% of total gross forest area loss in the BLA, which clearly indicates the critical role of the ITs/PAs in forest conservation.
Fig. 3: Annual gross forest area losses in the BLA from 2002 to 2021. a, Spatial distribution of the year a forest pixel in 2001 was classified from forest to non-forest (forest loss) in the ITs/PAs. b, Spatial distribution of the year a forest pixel in 2001 was classified from forest to non-forest (forest loss) outside ITs/PAs. c–g, Annual gross forest area loss in the BLA (c), ITs (d), nPAs (e), sPAs (f) and non-PAs (g). We used the 2001 forest map as the reference and identified the first year that forest pixels became non-forest pixels during 2002–2021. The dark grey pixels in a and b are forest in 2001. Full size image
The interannual change of gross forest area loss in the BLA (Fig. 3c) during 2002–2021 reveals three interesting results. First, annual gross forest area losses in 2005, 2007, 2010 and 2015 were substantially larger than those in previous and subsequent years. Years 2005, 2007, 2010 and 2015 were characterized by strong El Niño events (Extended Data Fig. 3), high air temperature or severe drought26. Second, annual gross forest area loss in 2013 was least during 2000–2021. Year 2013 was a year with high air temperature without an El Niño or Atlantic Multidecadal Oscillation event. Third, when those years (2005, 2007, 2010, 2013 and 2015) are taken out, the temporal dynamics of annual gross forest area loss show five noticeable phases and each of these phases lasted a few years: (1) increased forest area loss in 2002–2004, (2) reduced forest loss in 2006–2009, (3) increased forest area loss in 2011–2014, (4) decreased forest area loss in 2016–2018 and (5) increased forest area loss in 2019–2021. In 2018–2021, the annual gross forest area loss rates increased 3.6 times in ITs/PAs, larger than the increase in non-PAs (1.6 times), indicating increasing deforestation pressure and an alarming signal in ITs/PAs.
Varying effects of ITs/PAs on annual forest area loss
To investigate the effect of ITs/PAs on reducing forest area loss after they were established, we selected those ITs/PAs that were established after 2002 and analysed the average annual gross forest area loss rates before and after their establishment years (Fig. 4). The results show that 72 ITs, 32 nPAs and 38 sPAs had substantially reduced annual forest area loss rates but 49 ITs, 21 nPAs and 27 sPAs still had small to moderate increases in annual forest area loss rates; the other 24 ITs, 2 nPAs and 7 sPAs had no change in forest area loss rates as they had little or no forest area loss (Fig. 4a–c). There was no clear geographical cluster among those ITs/PAs with reduced or increased forest loss rates (Fig. 4d–f). Most of the sPAs in the northern and western BLA had small increases in annual forest area loss rates. In terms of institution and governance, the average annual gross forest area loss rates were reduced substantially for the nPAs (36%, 13.7 × 103 ha) and the ITs (30%, 10.7 × 103 ha) but only slightly for the sPAs (5%, 2.7 × 103 ha; Fig. 4g and Extended Data Fig. 4a). In terms of management objectives, the average annual gross forest area loss rates were reduced substantially for the strict-protection PAs (48%, 8.6 × 103 ha) and the ITs (30%, 10.7 × 103 ha) but only moderately for the sustainable-use PAs (11%, 7.7 × 103 ha; Fig. 4h and Extended Data Fig. 4b).
Fig. 4: The effects of ITs/PAs on annual forest area loss rates in the BLA (2001–2021). a–c, Comparison of average annual gross forest area loss rates before and after ITs/PAs establishment (after the year 2002) by ITs (a), nPAs (b) and sPAs (c). d–f, Spatial distribution of the changes in annual gross forest area loss rates before and after ITs/PAs establishment by ITs (d), nPAs (e) and sPAs (f). g,h, The changes in average annual gross forest loss rates before and after the ITs/PAs establishment by governance (nPAs, ITs and sPAs) (g) and management (strict protection and sustainable use) (h). All ITs/PAs established before and during 2002 are white polygons. Full size image
Different deforestation dynamics inside different PA types
Forest area in nPAs with the strict-protection objective (Fig. 5a) varied slightly during 2000–2013 but decreased moderately in 2014–2016, with a net loss of 0.14 × 106 ha (0.45%) from 2000 to 2021. Forest area in the nPAs with the sustainable-use objective (Fig. 5a) decreased continuously from 2000 (31.7 × 106 ha) to 2021 (31.5 × 106 ha), a net loss of 0.22 × 106 ha (0.7%). Forest area in sPAs with the strict-protection objective (Fig. 5b) decreased from 5.91 × 106 ha in 2000 to 5.88 × 106 ha in 2021, a net loss of 0.03 × 106 ha (0.51%). Forest area in sPAs with the sustainable-use objective (Fig. 5b) decreased from 44.65 × 106 ha in 2000 to 43.53 × 106 ha in 2016, a net loss of 1.12 × 106 ha (2.5%) but it had a modest recovery by 2021 (44.19 × 106 ha), thus a net loss of 0.46 × 106 ha (1.0%) from 2000 to 2021.
Fig. 5: Annual forest areas and gross forest area loss rates in PAs with different governance and management regimes from 2000 to 2021. a, Annual forest areas in nPAs (strict protection, left Y-axis, and sustainable use, right Y-axis). b, Annual forest areas in sPAs (strict protection, left Y-axis, and sustainable use, right Y-axis). c, Area proportion of annual gross forest area loss to forest area in 2001 in nPAs (strict protection and sustainable use). d, Area proportion of annual gross forest area loss to forest area in 2001 in sPAs (strict protection and sustainable use). Full size image
Differences in annual gross forest area loss rates were small and not significant between the PAs with different governance and management (Fig. 5c,d). The sPAs with the sustainable-use objective had the highest average gross forest area loss rate (0.31 ± 0.14% yr−1 of the forest area in 2001), followed by nPAs with the strict-protection objective (0.11 ± 0.05% yr−1) and nPAs with the sustainable-use objective (0.09 ± 0.04% yr−1). The sPAs with the strict-protection objective had the lowest average gross forest area loss rate (0.04 ± 0.04% yr−1) due to these PAs being far away from the ‘arc of deforestation’ and therefore were under little deforestation pressure (Fig. 4f). The gross forest area loss rates increased from 2018 to 2021; these losses were probably related to the loosened forest conservation policies during the Bolsonaro presidential administration that began in January 201917,18,27. Gross forest area loss from 2018 to 2021 increased 1.5 times in nPAs with the strict-protection objective and 5 times in nPAs with the sustainable-use objective. Gross forest area loss from 2018 to 2021 increased 12.4 times in the sPAs with the strict-protection objective and 4.3 times in sPAs with the sustainable-use objective.