
Spatial Concentration of NDVI and Fire Activity in South America (2000-2022)
Jun 24
18 min read
Abstract
This study investigates the spatial relationship between vegetation decline and fire activity across South America from 2000 to 2022. It is hypothesized that vegetation decline and fire activity have both become more spatially concentrated over time, and that areas experiencing persistent vegetation loss are likely to coincide spatially with regions of high fire incidence. Utilizing MODIS-derived Normalized Difference Vegetation Index (NDVI) data and MODIS/VIIRS fire detection records, Global Moran’s I, Average Nearest Neighbor Analysis (NNA), and Local Moran’s I were applied to assess patterns of vegetation loss and fire occurrence. The findings reveal that NDVI decline has become increasingly spatially clustered over time, with the highest clustering observed between 2011 and 2015. Similarly, fire incidents have shown a trend toward greater spatial concentration, particularly in the 2016–2022 period. Local Moran’s I analysis identified significant cold spots of NDVI change that coincide with high-density fire regions, notably in Brazil, Bolivia, and Paraguay. These results indicate a spatial correlation between vegetation degradation and fire activity, possibly associated with land-use practices such as slash-and-burn agriculture and pasture expansion. Understanding these spatial dynamics is crucial for developing targeted land management and fire mitigation strategies to preserve ecosystem resilience and carbon storage in South America.
1. Introduction
Land cover change in South America has accelerated because of agricultural expansion, deforestation, and climate variability, much of which is linked to global demand for animal-based foods. These pressures are transforming ecosystems, disrupting vegetation health, and reshaping natural fire regimes. As forests are removed to create pastures and grow feed crops, fire activity becomes more frequent and recovery of the land becomes more difficult. NDVI (Normalized Difference Vegetation Index) helps show these shifts by tracking vegetation greenness and stress across large regions, while satellite fire detections offer clear records of when and where fires occur. I approach this research from a perspective rooted in sustainability and concern for ecosystem resilience, with the goal of generating knowledge that supports a shift toward plant-based food systems that ease the burden on valuable and vulnerable environments like the Amazon.
1.1. What is the Problem and Why is it Important?
Fire and vegetation loss are not isolated events. They can alter the structure and function of biogeochemical cycles. Wildfires rapidly transfer carbon from terrestrial biomass to the atmosphere, releasing large volumes of CO₂ and reducing long-term carbon storage in forests (Anderegg et al. 2020). This weakens the role of terrestrial ecosystems as carbon sinks.
Nitrogen cycling is also disrupted. Fires volatilize nitrogen stored in vegetation and soil organic matter, leading to a net loss of nitrogen from ecosystems (Ribeiro et al. 2021). These losses are not offset by post-fire biological nitrogen fixation, which remains low for years after disturbance (Brais et al. 2020).
Soil microbial activity, which drives decomposition and nutrient turnover, is also impacted. Fire exposure alters microbial communities and suppresses gross nitrogen transformation rates in forest soils (Castro and Steudler 2021). This slows nutrient recovery and affects plant regrowth.
In pasture regions created after deforestation, fire recurrence and grazing jointly reduce soil carbon and nitrogen stocks over time (Neill et al. 1997). These losses contribute to long-term fertility decline and constrain ecosystem recovery.
At regional scales, fire use in land management contributes to increased greenhouse gas emissions and nutrient export. In Brazil, nitrogen losses due to land use and fire are significant enough to influence the national nitrogen budget (Bustamante et al. 2006).
The cumulative effect of these changes is reduced ecosystem resilience and enhanced climate forcing. Understanding how fire patterns relate to vegetation loss is essential for managing these disruptions to the Earth’s major biogeochemical cycles.
1.2. Research Question
Understanding the spatial and temporal relationship between vegetation loss and fire activity is essential for identifying the ecological consequences of deforestation. While it is well known that land-use change increases fire susceptibility, less is known about how the decline in vegetation, measured through remote sensing indices like NDVI, correlates with fire occurrence patterns over time. This is particularly relevant in South America, where widespread forest conversion and recurring fires contribute to long-term environmental degradation. To investigate this relationship, the following research questions guide the present analysis:
Have the percentage of changes in NDVI become more spatially concentrated since the year 2000 in South America?
Has fire activity become more spatially concentrated since the year 2000 in South America?
Do cold spots of NDVI change spatially coincide with concentrations in fire activity? (i.e., is there evidence of spatial correlation between areas of NDVI decline and fire activity)?
Akyürek (2023) conducted a spatial and temporal analysis of vegetation fires across Europe from 2000 to 2020, employing spatial statistical methods such as Global Moran’s I and Getis-Ord Gi* Hot Spot Analysis. The study identified regions like northern Portugal, the Balkans, southern Italy, and the Sicilian peninsula as areas with high wildfire intensity and risk. This research underscores the effectiveness of spatial statistical techniques in identifying fire-prone areas, providing a methodological framework that informs our analysis of fire activity and vegetation loss in South America.
Hypothesis
It is hypothesized that NDVI decline and fire activity have both become more spatially concentrated across South America from 2000 to 2022, and that areas experiencing vegetation decline, as indicated by NDVI cold spots, are spatially associated with zones of high fire activity.
1.3. What do we know?
Recent studies have established a clear link between land-use change and increasing fire activity across the Amazon Basin. Fires in this region are overwhelmingly anthropogenic in origin, with the majority ignited to facilitate deforestation for agricultural purposes, particularly cattle ranching and crop cultivation. This form of land conversion has been identified as a primary driver of the rising frequency and spatial concentration of fires, rather than climatic factors alone (Silveira et al., 2022).
Brazil accounts for the largest share of fire detections and burned areas in the Amazon, with most fires concentrated in deforested areas that have been converted to pasture (Silveira et al., 2022). While droughts and climate variability contribute to fire risk, the temporal alignment of fire outbreaks with peaks in deforestation rates suggests that human land use is the dominant explanatory factor. The spatial overlap of fire incidents with areas cleared for animal agriculture further reinforces this conclusion (Cochrane and Barber, 2009; Morton et al., 2008).
Projections indicate that the continuation of current land-use trajectories, particularly under scenarios of agricultural intensification, will substantially increase fire probability in the coming decades. Fonseca et al. (2019) found that under combined climate and land-use change scenarios, fire risk in the Brazilian Amazon could more than double. This trend is expected to be exacerbated by feedback mechanisms in which fire-induced forest degradation leads to canopy thinning, increased solar radiation, and drier microclimates, which in turn increase forest flammability and susceptibility to recurrent burning (Nepstad et al., 2008).
The global demand for animal protein is a central factor in this dynamic. Expansion of pasture for cattle and soybean production for animal feed are among the primary causes of deforestation in the Amazon. These activities not only lead to direct vegetation loss but also elevate ignition sources, enabling fires to escape into adjacent forested areas (Le Page et al., 2017). As forests degrade, their capacity to recover diminishes, increasing the likelihood of reaching a biome-level tipping point by which tropical rainforests transition to degraded savanna (Nepstad et al., 2008).
The cumulative evidence indicates that understanding and mitigating fire activity in the Amazon requires direct engagement with the drivers of deforestation, particularly those associated with animal agriculture. Addressing these pressures is needed for reducing fire risk, preserving forest integrity, and maintaining the stability of regional and global biogeochemical cycles.
1.3. What don't we know?
While the relationship between deforestation and fire activity in the Amazon is well documented, significant knowledge gaps remain regarding the spatial and temporal dynamics of these processes. Much of the existing literature emphasizes either the biophysical drivers of fire risk (drought, El Niño events) or the policy and governance dimensions of deforestation, but relatively few studies have quantitatively examined how vegetation degradation, measured through indices such as NDVI, correlates with fire patterns across both space and time.
The literature could use more research about how the intensity and spatial clustering of vegetation decline influences the probability and spread of fires in distinct biogeographic regions of the Amazon. Although some studies have used satellite-derived NDVI data to assess vegetation health post-fire, fewer have employed NDVI percentage change as a predictor of fire occurrence, particularly using spatial statistical tools such as Moran’s I or Nearest Neighbor Analysis.
Another gap lies in the temporal resolution of fire and vegetation interaction studies. Many analyses are based on decadal change or isolated extreme events. As a result, trends in the spatial concentration or dispersion of fire activity over intermediate timescales, spanning several years, are often overlooked. This makes it difficult to assess whether fire activity is becoming more spatially concentrated over time and if that concentration corresponds with persistent vegetation loss.
Moreover, most studies remain focused on aggregate regional or national trends, with less emphasis on sub-national variation or localized hotspots of vulnerability. This limits the ability to develop place-specific interventions and hinders the understanding of fire dynamics in less-studied regions of the Amazon outside of Brazil.
Finally, there is a scarcity of research that directly connects land use transitions associated with global dietary trends, such as rising demand for animal products, with the biogeochemical implications of fire-driven land degradation. Although the role of animal agriculture in deforestation is widely acknowledged, its specific contribution to long-term fire patterns and nutrient cycling disruptions remains underexplored.
1.4. How does this research fill these gaps?
This research addresses limitations identified in the current literature by integrating spatial statistical analysis with remote sensing data to examine the relationship between vegetation decline and fire activity across South America over a multi-decade period. Unlike many prior studies that focus on deforestation or fire risk independently, this project investigates whether percentage change in NDVI, a proxy for vegetation health and biomass, correlates spatially and temporally with fire occurrence, offering a more nuanced understanding of landscape degradation dynamics.
This study moves beyond simple trend analysis and explores whether these ecological changes are becoming more spatially concentrated, by calculating NDVI percentage change over four time intervals (2000-2005, 2006–2010, 2011–2015, and 2016–2022) and applying spatial statistics (i.e., Moran’s I and Nearest Neighbor Analysis).
2. Methodology
2.1. Methodological Approach
This study uses spatial statistical methods and remote sensing data to analyze whether vegetation loss, as indicated by NDVI decline, correlates with increased fire activity in South America over time. The research is structured around three spatial analysis questions: (1) whether NDVI percentage change values exhibit spatial clustering; (2) whether fire activity has become more spatially concentrated; and (3) whether cold spots of NDVI change coincide with regions of high fire density. Similar spatial statistical methods have been effectively applied in other regions, such as Europe, to analyze vegetation fire patterns (Akyürek, 2023).
2.2. Have percentage changes in NDVI become more spatially concentrated?
To assess spatial concentration of vegetation decline, this study calculates percentage change in NDVI for four time intervals: 2000-2005, 2006–2010, 2011–2015, and 2016–2022. MODIS-derived NDVI data for July of each year were used, and all raster files were pre-processed to floating point format to support arithmetic operations. NDVI percent change was calculated using Raster Calculator in ArcGIS Pro, applying the formula: ((NDVI_later – NDVI_earlier) / |NDVI_earlier|) × 100
To prepare the NDVI percentage change rasters for spatial statistical analysis in R, each raster was first converted to polygons in ArcGIS Pro using the Raster to Polygon tool. The resulting polygon shapefiles were then imported into R, where Global Moran’s I was calculated to determine whether vegetation decline values were spatially clustered or randomly distributed. Global Moran’s I was then computed for each time interval to determine whether values of NDVI percentage change were spatially clustered. A positive and significant Moran’s I value indicates that similar values (high or low NDVI change) are spatially clustered rather than randomly distributed. This method is appropriate for addressing the first research question because it quantifies spatial autocorrelation across the landscape and determines whether patterns of vegetation decline are becoming more localized or dispersed over time.
2.3. Has fire activity become more spatially concentrated?
To investigate whether fire activity has become more geographically concentrated over time, fire detection data from NASA’s MODIS and VIIRS satellite platforms were used. These data include daily fire locations from 2000 to 2022. Fire points were filtered to include only those occurring in South America and were grouped into three time periods to match the NDVI analysis intervals.
Each subset of fire points was analyzed using Average Nearest Neighbor Analysis (NNA) in ArcGIS Pro. NNA calculates the mean distance between each fire event and its closest neighboring event, then compares this value to the expected mean distance under a hypothetical random distribution. A nearest neighbor ratio less than 1 indicates clustering, while a ratio greater than 1 indicates dispersion.
This method directly addresses the second research question by providing statistical evidence for changes in the spatial structure of fire activity over time. Identifying increasing clustering in fire events would suggest an escalation in the intensity or concentration of fire-prone areas, potentially linked to land-use practices such as slash-and-burn agriculture or pasture expansion.
2.4. Do cold spots of NDVI change spatially coincide with concentrations in fire activity?
To explore the spatial coincidence between areas of vegetation loss and fire activity, Local Moran’s I was applied to each NDVI percentage change raster. This method identifies statistically significant spatial clusters of high or low values and classifies them as hot spots (high-high), cold spots (low-low), or spatial outliers. The analysis was conducted in R using the spdep package.
3. Data and Study Area
3.1. Study Area
This study focuses on the continent of South America, a region that contains approximately 5.5 million square miles of land and is home to the Amazon rainforest, the largest tropical forest on Earth. The Amazon Basin spans nine countries (Brazil, Peru, Colombia, Bolivia, Venezuela, Ecuador, Guyana, Suriname, and French Guiana) but Brazil contains more than 60% of its total area. Much of the fire activity and deforestation pressure is concentrated in this region, especially in southern and eastern Brazil, northern Bolivia, and parts of Paraguay and Colombia.
While the Amazon is the primary area of concern, this study includes the full South American landmass to ensure comprehensive coverage of fire incidents and vegetation change. The analysis does not rely on administrative boundaries but uses the entire continental landmass as a continuous study area. Raster-based NDVI and fire detection datasets were clipped to the South America boundary using a region-specific shapefile in ArcGIS Pro. The processing time to
3.2. Geographic Units and Temporal Scope
The analysis is conducted at the native resolution of the NDVI (~8 km). Raster cells are used as the geographic unit of analysis and converted to polygons for spatial statistical testing. For fire data, each MODIS detection is treated as a point event.
Four time intervals, each consisting of 5 years, were selected based on NDVI data availability and to support temporal comparison:
2000-2005
2006–2010
2011–2015
2016–2022
Breaking the study into intervals allows for the detection of when spatial clustering intensified, offering a timeline of change rather than a static view. Second, analyzing multiple time windows avoids masking short-term fluctuations or nonlinear patterns that could be hidden in a single 22-year comparison. Third, this approach better aligns the analysis with known fire activity cycles and land-use changes, such as those associated with El Niño years or shifts in agricultural expansion.
3.3. Data Sources and Variables
The study uses two main datasets: NDVI raster layers and MODIS/VIIRS fire detection points. The specifications for each one are provided in Table 1.
Table 1. Data Sources and Variables
Category | NDVI (Normalized Difference Vegetation Index) | MODIS/VIIRS Active Fire Products |
Source | NASA MEaSUREs Global Inventory Modeling and Mapping Studies (GIMMS) NDVI3g v1.1 | NASA Fire Information for Resource Management System (FIRMS) |
Access | ||
Temporal Coverage | Monthly composites, July 2000–2022 | 2000–2022 |
Variables | NDVI (unitless index, scaled and converted to floating point) | Latitude, longitude, acquisition date, brightness temperature, confidence level |
Processing Steps | Percent change calculated over 4 periods | Fire points filtered to South America; grouped into the same 4 periods; spatial clustering assessed using Average Nearest Neighbor |
4. Results
4.1. Global Moran's I for NDVI Percent Change (2000–2022)
To assess whether NDVI percent change values became more spatially clustered over time, Global Moran’s I was calculated for four sequential time periods: 2000–2005, 2006–2010, 2011–2015, and 2016–2022. Moran’s I is a measure of spatial autocorrelation that quantifies the degree to which similar values occur near one another. A positive Moran’s I value indicates spatial clustering, while a value near zero suggests a random distribution. The test was applied to polygon shapefiles derived from NDVI percent change rasters, and spatial weights were computed using Queen’s contiguity.
The results from the Global Moran’s I test are provided in Appendix 1 and compiled in Table 2. All time periods exhibited statistically significant positive spatial autocorrelation. The magnitude of Moran’s I increased from 2000–2005 to 2011–2015, suggesting that NDVI decline became increasingly spatially concentrated during that time. Although the Moran’s I value decreased slightly in the final interval (2016–2022), it remained highly significant, indicating continues clustering of vegetation loss. These findings are consistent with the hypothesis that NDVI decline across South America has become more geographically concentrated over the past two decades.
Table 2. Spatial Autocorrelation of NDVI Percent Change
Time | Moran’s I | Z-score | p-value | Interpretation |
2000–2005 | 0.00224 | 2.48 | 0.0065 | Weak but statistically significant clustering |
2006–2010 | 0.01357 | 16.68 | < 2.2e-16 | Strong clustering of similar NDVI change values |
2011–2015 | 0.03412 | 31.48 | < 2.2e-16 | Very strong clustering – highest observed |
2016–2022 | 0.01283 | 11.87 | < 2.2e-16 | Strong clustering, though slightly reduced |
Figure 1. Global Moran’s I for NDVI Percent Change (2000 -2022)

The Global Moran’s I analysis indicates that changes in NDVI across South America are not randomly distributed but exhibit spatial clustering. This indicates that areas with similar levels of vegetation change, particularly declines in NDVI, are geographically proximate. An increase in Moran’s I values over time implies that vegetation changes are becoming more spatially concentrated. However, a decline in Moran’s I values between 2016 and 2022 suggests a reduction in spatial clustering, implying that vegetation changes during this period became more dispersed across the continent. This shift may reflect a broader distribution of factors affecting vegetation, resulting in a more heterogeneous pattern of NDVI changes.
4.2. Spatial Concentration of Fire Activity (2000–2022)
Average Nearest Neighbor (ANN) was employed to analyze point pattern data derived from MODIS fire detections. ANN analysis is an established spatial statistical method for measuring the degree of spatial clustering or dispersion among point features. Fire incident data were grouped into four time periods: 2000–2005, 2006–2010, 2011–2015, and 2016–2022. For each period, ANN analysis was conducted by calculating the mean distance (in decimal degrees) between each fire incident and its nearest neighbor. Lower average distances indicate tighter clustering of fire events, while higher values suggest a more dispersed pattern. The results of this analysis are presented in Table 3. The output from the R script is contained in Appendix 2.
Table 3. Average nearest neighbor distance for fire incidents in South America
Time Period | Fire Incidents | ANN Distance (decimal degrees) |
2000–2005 | 3,919,881 | 0.00508 |
2006–2010 | 3,912,881 | 0.00507 |
2011–2015 | 2,807,944 | 0.00566 |
2016–2022 | 5,261,126 | 0.00438 |
The average nearest neighbor distance has declined over time (as shown in Figure 2), especially during the most recent period (2016–2022), which also had the highest number of fire detections. Despite a dip in fire counts during the 2011–2015 interval, the subsequent surge in both incident counts and spatial clustering suggests intensification of fire activity in certain hotspot regions.
Figure 2. Average Nearest Neighbor Distance of Fire Incidents (2000–2022)

The decreasing ANN values indicate that fires are occurring in closer proximity to one another, meaning there is a rising degree of spatial concentration. This supports the hypothesis that anthropogenic drivers (such as slash-and-burn agriculture, pasture expansion, and land clearing) may be associated with denser clusters of fire activity across South America.
The findings from the ANN analysis reinforce the broader hypothesis that fire activity in South America appears to be increasing in frequency and becoming more geographically concentrated, a pattern that poses serious implications for forest degradation, carbon emissions, and long-term ecological resilience in the region.
4.3. Coincidence of NDVI-change Cold Spots and Fire Concentrations
The Local Moran’s I results are shown in Table 4. During the initial test, the majority of polygons were classified as "Not Significant," likely due to the large scale of the study area. After applying a snap distance of 0.01 to better connect adjacent polygons and reducing the influence of extreme values, the number of significant High-High and Low-Low clusters increased. For example, in the 2000–2005 time period, 21,485 High-High and 23,645 Low-Low clusters were identified, compared to just 148 and 180 in the original unadjusted analysis. This trend was consistent across all four time intervals, with a notable increase in Low-Low clusters during the most recent period (2016–2022), suggesting growing spatial concentration of NDVI decline in certain regions. These improvements enhanced the ability of the Local Moran’s I method to detect localized patterns of vegetation degradation that may coincide with fire activity.
Table 4. Summary of LISA Cluster Results
Time Period | High-High | High-Low | Low-High | Low-Low | Not Significant |
2000-2005 | 21485 | 1143 | 2400 | 23645 | 140502 |
2006-2010 | 18928 | 1758 | 2432 | 19296 | 141166 |
2011-2015 | 24174 | 952 | 2875 | 20946 | 139533 |
2016-2022 | 18770 | 1417 | 1172 | 22445 | 132206 |
4.3.1. Spatial Distribution of NDVI Cold Spots (LISA Low-Low Clusters)
The spatial distribution of statistically significant NDVI cold spots, areas with clustered vegetation decline, was visualized in ArcGIS (Figure 3). In these maps, bright yellow indicates high-density clustering of NDVI loss, while blue-gray areas represent sparse or isolated clustering. A definition query was applied to each of the four map layers shown, so that only Low-Low clusters, representing NDVI decline, are mapped.
In the 2000–2005 period, cold spot clustering was most pronounced in Argentina and northeastern countries on the continent. The results indicate that these areas are where vegetation decline formed contiguous patterns across neighboring cells.
Between 2006–2010, dense clustering emerged in eastern Brazil, as well as in Bolivia and Columbia.
The 2011–2015 interval showed spatially extensive clustering, with bright yellow zones expanding across Venezuela and Columbia. The region that was prominent in Brazil during the 2006-2010 time period grew larger during 2011-2015, indicating broader-scale land cover degradation.
From 2016–2022, high-density cold spot clustering shifted to the southern part of the continent, in Bolivia, Paraguay, Argentina, and Chile.
These mapped patterns are consistent with the hypothesis that vegetation decline is becoming more spatially concentrated and widespread over time, offering a strong visual basis for analyzing its geographic alignment with fire activity patterns.
Figure 3. Spatial Density of NDVI Cold Spots (Low-Low Clusters) from 2000 to 2022

4.3.2. Spatial Density of Fire Incidents
Figure 4 displays kernel density estimates of fire incidents across South America for each of the four time periods analyzed. These maps were generated using point location data from MODIS-detected fires and visualize the concentration of fire activity using a color gradient from sparse (blue) to dense (yellow).
In the 2000–2005 period, high-density fire activity is observed across much of central Brazil, extending into eastern Bolivia, northern Paraguay, and northeastern Argentina.
During 2006–2010, fire activity remains elevated in central South America, particularly in southern and eastern Brazil, with dense fire clusters also visible in Paraguay and parts of the southern Amazon basin.
In the 2011–2015 interval, fire density is less widespread in Brazil and Bolivia as it was in 2006-2010. By 2016–2022, fire activity becomes the most spatially extensive of all four time periods, with high-density areas spreading across a wider swath of central Brazil, and persisting in the Brazil-Paraguay-Bolivia tri-border area. Additional clusters appear in eastern Peru, Colombia, and Venezuela, indicating a geographic shift and expansion in fire-prone zones.
These fire maps visually corroborate trends associated with land-use conversion and agricultural intensification (Morton et al, 2008). Although not all areas of fire activity align directly with NDVI cold spots, several regions in the countries of Paraguay, Bolivia, and southern Brazil, show recurring spatial overlap between zones of concentrated fire activity and statistically significant NDVI decline.
Figure 4. Spatial Density of Fire Incidents (2000-2022)

When viewed collectively, the results from Global Moran’s I, Average Nearest Neighbor, and Local Moran’s I analyses provide strong support for the hypothesis that NDVI decline and fire activity have become more spatially concentrated over time. The visual and statistical evidence of overlap between NDVI cold spots and fire hotspots supports the hypothesis that persistent vegetation loss coincides spatially with areas of dense fire activity in key regions.
5. Limitations
Although this study provides valuable insights into the spatial patterns of vegetation decline and fire activity, it has several limitations. First, the analysis relies solely on NDVI percent change and satellite-detected fire data. While these datasets are useful for large-scale spatial assessments, these datasets do not capture important underlying environmental drivers such as precipitation, temperature anomalies, or specific land use transitions.
Additionally, the study does not incorporate socio-economic or political variables that may strongly influence land clearing and fire dynamics. For example, during recent political administrations in Brazil, enforcement of environmental regulations was significantly weakened and resources for environmental monitoring and fire prevention were reduced. This reduction in institutional capacity likely contributed to increased deforestation and fire activity in ways that satellite data alone cannot fully capture (Abessa, Famá, & Buruaem, 2019; Ferrante & Fearnside, 2020). Spatial statistical methods such as Moran’s I and Average Nearest Neighbor detect clustering but do not explain the causes behind these patterns. While this analysis identifies where vegetation decline and fire activity tend to co-occur, it does not establish direct causal relationships between the two variables.
6. Concluding Remarks
This study examined the spatial distribution of vegetation decline and fire activity in South America from 2000 to 2022 using multiple spatial statistical methods. The analyses identified clear patterns of clustering in both NDVI percentage change and fire detections. Global Moran’s I showed increasing spatial autocorrelation of NDVI decline through 2015, followed by a slight decrease in the final interval. The Average Nearest Neighbor analysis results showed a reduced distances between fire incidents over time, indicating more spatially concentrated fire activity. Local Moran’s I identified statistically significant clusters of vegetation loss (Low-Low NDVI percent change) and confirmed that these clusters were often located near or within areas of dense fire activity.
Overall, the findings support the hypothesis that NDVI decline and fire activity have become more spatially concentrated across South America, and that areas experiencing persistent vegetation loss frequently coincide with zones of high fire activity. These patterns, while indicative of spatial association rather than direct causation, reveal important geographic hotspots where land degradation processes interact and intensify. Understanding these linkages is essential as pressures on these landscapes continue to grow.
Given that much of this change is tied to the expansion of animal agriculture, the study carries implications beyond land management. Reducing reliance on animal-based food systems could lessen the demand driving deforestation and fire use, helping to slow vegetation loss and restore balance to carbon and hydrologic cycles. As someone committed to supporting plant-based and climate-friendly solutions, I view these results as not only a scientific contribution, but as evidence that transforming food systems is essential for safeguarding ecosystem resilience. Moving forward, further research that takes a deep dive into a specific country(ies) would provide interesting insights, especially ones like Bolivia and Paraguay where the statistical results were most pronounced. Also, incorporating higher-resolution data and socio-economic variables will be vital to deepen our understanding of these spatial patterns and guide more effective and sustainable land stewardship.
7. References
Abessa, D., Famá, A., & Buruaem, L. (2019). The systematic dismantling of Brazilian environmental laws risks losses on all fronts. Nature Ecology & Evolution, 3(4), 510–511. https://doi.org/10.1038/s41559-019-0855-9
Anderegg, W. R. L., Trugman, A. T., Badgley, G., Anderson, C. M., Bartuska, A., Ciais, P., ... & Pacala, S. (2020). Droughts, wildfires, and forest carbon cycling: A pantropical synthesis. Global Change Biology, 26(9), 5041–5051. https://doi.org/10.1111/gcb.15228
Akyürek, B. (2023). Spatial and temporal analysis of vegetation fires in Europe. Fire, 6(1), Article 9. https://doi.org/10.3390/fire6010009
Brais, S., Harvey, B. D., & Bergeron, Y. (2020). Biological nitrogen fixation does not replace nitrogen losses after forest fires. Forest Ecology and Management, 474, Article 118371. https://doi.org/10.1016/j.foreco.2020.118371
Bustamante, M. M. C., Martinelli, L. A., Silva, D. A., Camargo, P. B., Klink, C. A., Domingues, T. F., & Santilli, M. (2006). Human activities changing the nitrogen cycle in Brazil. Biogeochemistry, 79(1–2), 95–122. https://doi.org/10.1007/s10533-006-9000-1
Castro, M. S., & Steudler, P. A. (2021). Impact of fire on soil gross nitrogen transformations in forest ecosystems. Forest Ecology and Management, 496, Article 119428. https://doi.org/10.1016/j.foreco.2021.119428
Ferrante, L., & Fearnside, P. M. (2020). Protect Brazil’s indigenous peoples from COVID-19. Science, 368(6488), 251. https://doi.org/10.1126/science.abc0073
Morton, D. C., DeFries, R. S., Shimabukuro, Y. E., Anderson, L. O., Arai, E., Del Bon Espirito-Santo, F., Freitas, R., & Morisette, J. (2008). Agricultural intensification increases deforestation fire activity in Amazonia. Global Change Biology, 14(10), 2262–2275. https://doi.org/10.1111/j.1365-2486.2008.01652.x
Neill, C., Melillo, J. M., Steudler, P. A., Cerri, C. C., de Moraes, J. F. L., Piccolo, M. C., & Brito, M. (1997). Fire in the Brazilian Amazon 2. Biomass, nutrient pools and losses in cattle pastures. Ecological Applications, 7(1), 91–108. https://doi.org/10.1890/1051-0761(1997)007[0091:FITBA]2.0.CO;2
Ribeiro, T. K. P., Resende, A. S., Pereira, M. G., Xavier, F. A. S., Campello, E. F. C., Urquiaga, S., ... & Boddey, R. M. (2021). Long-term impacts of livestock introduction on carbon and nitrogen cycling in the Brazilian Cerrado. Science of the Total Environment, 781, Article 146676. https://doi.org/10.1016/j.scitotenv.2021.146676