Depopulation of rural landscapes exacerbates fire activity in the western Amazon.

Destructive fires in Amazonia have occurred in the past decade, leading to forest degradation, carbon emissions, impaired air quality, and property damage. Here, we couple climate, geospatial, and province-level census data, with farmer surveys to examine the climatic, demographic, and land use factors associated with fire frequency in the Peruvian Amazon from 2000 to 2010. Although our results corroborate previous findings elsewhere that drought and proximity to roads increase fire frequency, the province-scale analysis further identifies decreases in rural populations as an additional factor. Farmer survey data suggest that increased burn scar frequency and size reflect increased flammability of emptying rural landscapes and reduced capacity to control fire. With rural populations projected to decline, more frequent drought, and expansion of road infrastructure, fire risk is likely to increase in western Amazonia. Damage from fire can be reduced through warning systems that target high-risk locations, coordinated fire fighting efforts, and initiatives that provide options for people to remain in rural landscapes.

Land cover change interacts with drought severity to change fire regimes in Western Amazonia.

Fire is becoming a pervasive driver of environmental change in Amazonia and is expected to intensify, given projected reductions in precipitation and forest cover. Understanding of the influence of post-deforestation land cover change on fires in Amazonia is limited, even though fires in cleared lands constitute a threat for ecosystems, agriculture, and human health. We used MODIS satellite data to map burned areas annually between 2001 and 2010. We then combined these maps with land cover and climate information to understand the influence of land cover change in cleared lands and dry-season severity on fire occurrence and spread in a focus area in the Peruvian Amazon. Fire occurrence, quantified as the probability of burning of individual 232-m spatial resolution MODIS pixels, was modeled as a function of the area of land cover types within each pixel, drought severity, and distance to roads. Fire spread, quantified as the number of pixels burned in 3 × 3 pixel windows around each focal burned pixel, was modeled as a function of land cover configuration and area, dry-season severity, and distance to roads. We found that vegetation regrowth and oil palm expansion are significantly correlated with fire occurrence, but that the magnitude and sign of the correlation depend on drought severity, successional stage of regrowing vegetation, and oil palm age. Burning probability increased with the area of nondegraded pastures, fallow, and young oil palm and decreased with larger extents of degraded pastures, secondary forests, and adult oil palm plantations. Drought severity had the strongest influence on fire occurrence, overriding the effectiveness of secondary forests, but not of adult plantations, to reduce fire occurrence in severely dry years. Overall, irregular and scattered land cover patches reduced fire spread but irregular and dispersed fallows and secondary forests increased fire spread during dry years. Results underscore the importance of land cover management for reducing fire proliferation in this landscape. Incentives for promoting natural regeneration and perennial crops in cleared lands might help to reduce fire risk if those areas are protected against burning in early stages of development and during severely dry years.

Early-season warning of soybean rust regional epidemics using El Niño Southern/Oscillation information.

Soybean rust (SBR) is a disease of significant impact to Brazilian soybean production. Twenty-four locations in a major growing region in southern Brazil, where long-term (30 years) weather information was available, were selected to estimate the risk of SBR epidemics and identify potential predictors derived from El Niño 3.4 region. A rainfall-based model was used to predict SBR severity in an "epidemic development window" (the months of February and March for the studied region) in the time series. Twenty-eight daily simulations for each year-location (n = 720) were performed considering each day after 31 January as a hypothetical detection date (HDD) to estimate a severity index (SBRindex). The mean SBRindex in a single year was defined as the 'growing season severity index' (GSSI) for that year. A probabilistic risk assessment related GSSI and sea surface temperatures (SST) at the El Niño 3.4. region (here categorized as warm, cold or neutral phase) in October-November-December (OND) of the same growing season. Overall, the median GSSI across location-years was 34.5%. The risk of GSSI exceeding 60% was generally low and ranged from 0 to 20 percentage points, with the higher values found in the northern regions of the state when compared to the central-western. During a warm OND-SST phase, the probability of GSSI exceeding its overall mean (locations pooled) increased significantly by around 25 percentage points compared to neutral and cold SST phases, especially over the central western region. This study demonstrates the potential to use El Niño/Southern Oscillation information to anticipate the risk of SBR epidemics up to 1 month in advance at a regional scale.