Two Decades of Satellite Data Reveal How Brazil’s Caatinga Forest Defies Drought

Brazil’s Caatinga, the largest and most densely populated seasonally dry tropical forest on Earth, has long been dismissed as a barren, low-productivity wasteland. A new study published in Environmental Management turns that assumption on its head. By fusing twenty years of satellite observations with machine learning and spatial statistics, researchers Lucas Nascimento da Silva and Bartolomeu Israel de Souza of the Federal University of Paraíba found that this sprawling semiarid biome, covering roughly 844,453 square kilometers of northeastern Brazil, has maintained remarkably stable productivity and water-use efficiency from 2004 to 2024, even as regional temperatures climbed significantly. The findings suggest the Caatinga has not yet crossed the functional tipping points that have triggered abrupt vegetation collapse in drylands elsewhere, offering what the authors describe as a critical window of opportunity for conservation.

The research team built their assessment on the Carnegie-Ames-Stanford Approach, or CASA model, a process-based framework that estimates Net Primary Productivity, the net carbon an ecosystem gains after plants respire away their own energy. Monthly productivity was calculated as the product of absorbed photosynthetically active radiation and a light-use efficiency term, the latter dynamically downregulated by temperature and water stress scalars. To avoid signal saturation in dense vegetation and soil noise in degraded patches, the team averaged the fraction of absorbed radiation derived from two spectral indices, NDVI and the Simple Ratio. Because only two eddy covariance flux towers have ever operated in the Caatinga, and their footprints and overlap were far too limited for a biome-wide, two-decade analysis, the researchers cross-validated their estimates against the MOD17A3HGF global product, achieving a correlation of R-squared equals 0.79. They are candid that this constitutes a model-to-model comparison reflecting internal consistency rather than absolute accuracy, and they point to independent carbon-accounting datasets as future benchmarks.

Water Use Efficiency, the ratio of carbon assimilated to water lost through evapotranspiration, served as the study’s second headline metric. It captures the physiological trade-off every dryland plant faces: how much biomass can be manufactured per unit of scarce water. The team paired their productivity estimates with evapotranspiration data from the MOD16A2GF product and then subjected both time series to a battery of statistical tests. Theil-Sen slope estimation, chosen for its robustness to outliers, quantified the magnitude of change pixel by pixel, while the Mann-Kendall test assessed statistical significance. Standardized anomaly scores, or Z-scores, isolated the years when climate deviated most sharply from the historical mean, allowing the researchers to trace how extreme droughts ripple through the carbon and water cycles of the biome.

The temporal results are striking in their nuance. Neither NPP nor WUE showed a statistically significant long-term trend over the two decades, with p-values of 0.803 and 0.405 respectively, indicating that the balance between carbon fixation and water loss has remained resilient. Yet beneath that apparent calm lies extraordinary interannual volatility. The detrended anomalies revealed a strong, statistically significant inverse coupling between productivity and water-use efficiency, with a correlation coefficient of minus 0.741 and a p-value below 0.001. In wet years the Caatinga behaves as a highly efficient carbon sink, while in drought years productivity plunges and efficiency spikes in compensation. Precipitation itself remained statistically stable across the period, but mean annual temperature showed a significant and continuous warming trend, with a p-value of 0.014, underscoring the thermal pressure building on the system.

Perhaps the most consequential discovery concerns which environmental variable actually governs the biome’s behavior. Conventional wisdom holds that dryland productivity tracks annual rainfall almost linearly. The Caatinga does not obey that rule. Using a Random Forest ensemble of 500 decision trees, the researchers ranked the predictive power of five drivers: precipitation, temperature, the Land Surface Water Index (LSWI), the Temperature-Vegetation Dryness Index (TVDI), and land governance regime. LSWI, a spectral proxy for water held in vegetation and soil, dominated by a wide margin, accounting for roughly 0.85 of the relative importance for NPP trends and 0.70 for WUE trends. Precipitation, considered alone, explained far less. In other words, how much water the landscape retains matters more than how much water falls from the sky, a distinction with profound implications for how restoration and drought policy should be designed.

The mechanism makes ecological sense. In the semiarid Northeast, rainfall arrives in a handful of intense bursts concentrated into three to four months, and much of it runs off the surface before it can recharge the soil root zone. Preserved, structurally complex vegetation canopies dissipate the kinetic energy of raindrops, slow surface flow, and enhance infiltration, effectively banking moisture that sustains photosynthesis deep into the eight-month dry season. Deforested and degraded land does the opposite, reflecting more solar radiation, pumping more sensible heat into the atmosphere, and altering the local microclimate. This also distinguishes the Caatinga from Brazil’s Cerrado savanna, where deep-rooted trees tap groundwater and can sustain carbon exchange through prolonged droughts regardless of surface moisture. The Caatinga’s shrubby vegetation and shallow, rocky soils leave it far more dependent on what happens in the top layers of the earth.

While climate dictates the year-to-year rhythm, the study found that land governance shapes the long-term geography of degradation and recovery. Conservation Units, which cover only about 4.4 million hectares or 5.09 percent of the biome, emerged as its principal ecological refugia, concentrating the highest relative proportions of areas classified as early and late restoration. A buffer analysis probing distances of 0 to 2, 2 to 5, and 5 to 10 kilometers from territorial boundaries revealed that these units achieve peak stability within the first 2 kilometers inside their edges, a protective halo that progressively dissipates toward the unprotected exterior matrix. Indigenous Lands and Quilombola Territories, though much smaller at roughly 0.46 and 0.50 percent of the biome respectively, displayed a different and equally remarkable signature: nearly constant restoration levels from edge to core, suggesting that the traditional land-use practices of these communities buffer external degradation across their entire extent.

Not every governance story is a success. Indigenous Lands showed the highest internal proportion of degraded areas, and Quilombola Territories exhibited a stark spatial dichotomy, simultaneously ranking among the biome’s greatest conservation triumphs and its most severe degradation hotspots. The authors interpret this polarization through the lens of land tenure security. Where Afro-descendant communities hold formally recognized collective rights, customary institutions appear to promote restoration effectively, consistent with recent evidence that secure Indigenous and Quilombola lands in South America conserve irrecoverable carbon out of proportion to their area. Where title remains unresolved, the absence of enforceable collective property rights leaves territories exposed to chronic disturbances, continuous firewood extraction, uncontrolled fire, and overgrazing, that quietly erode biomass and biodiversity even without outright clear-cutting.

These findings arrive against a backdrop of rapid transformation. Between 1985 and 2019, forest and savanna formations within the Caatinga declined by approximately 8 and 11 percent respectively, while pastureland expanded by 62 percent and agricultural land by a staggering 284 percent. Climate projections point toward further aridification and warming through the end of the century, and global drylands are known to undergo abrupt regime shifts once critical aridity thresholds are breached. The current stability of the Caatinga’s carbon balance should therefore be read not as invulnerability but as evidence that the system is still operating within its adaptive capacity, a capacity that simplified, degraded landscapes lose first. National greenhouse gas inventories add weight to the stakes: in certain recent years, carbon removals in the Caatinga represented nearly half of Brazil’s total, and eddy covariance studies have found the biome’s carbon use efficiency exceeds that of other dry forests worldwide.

The policy prescriptions that flow from the study are as clear as they are demanding. Because formally protected areas cover only about 5 percent of the biome, regional resilience depends critically on the vast private matrix, and the authors argue that reactive emergency drought subsidies must give way to proactive landscape management that enhances soil water retention at scale. They call for robust financial mechanisms such as Payments for Ecosystem Services to reward private landowners who conserve native vegetation, urgent land tenure regularization for vulnerable traditional territories, and genuine inclusion of Indigenous Peoples and Afro-descendant communities in environmental decision-making. Continuous satellite monitoring of LSWI, they add, offers managers a real-time early-warning indicator capable of separating natural climate variability from human-induced degradation before vulnerable hotspots cross irreversible desertification thresholds. In the world’s most populous semiarid region, the fate of millions may hinge on whether that shift happens in time.

Subject of Research: Remote sensing assessment of ecosystem productivity, water-use efficiency, and land-use governance in the Caatinga tropical dry forest

Article Title: Ecosystem Resilience and Land-Use Governance: Remote Sensing Insights for Environmental Management in a Tropical Dry Forest

Article References: Ecosystem Resilience and Land-Use Governance: Remote Sensing Insights for Environmental Management in a Tropical Dry Forest. (n.d.). https://doi.org/10.1007/s00267-026-02629-4

Image Credits: AI Generated

DOI: 10.1007/s00267-026-02629-4

Keywords: Caatinga, tropical dry forest, Net Primary Productivity, Water Use Efficiency, remote sensing, LSWI, ecosystem resilience, land governance, Conservation Units, Indigenous Lands, Quilombola Territories, desertification