The answer, it turns out, depends on when you look and where you measure. Mean annual air temperature across the mangrove domain increased by approximately 0.02 degrees Celsius per year between 1979 and 2019, equivalent to about 0.20 degrees Celsius per decade, a rate consistent with published warming trends for southeastern Brazil. The fastest warming occurred during the coolest and relatively driest part of the year. The July-to-September quarter warmed at roughly 0.03 degrees Celsius per year, and September alone showed the strongest monthly warming at approximately 0.04 degrees Celsius per year, with statistically significant warming covering about 39 percent of the mapped mangrove area. This seasonal concentration of warming matters because September is a transition month in the estuary, when rainfall begins to recover from its winter minimum while solar radiation and atmospheric water demand climb.

The study area itself is ecologically significant. The Cananéia-Iguape Coastal System spans roughly 3,414 square kilometers across the municipalities of Cananéia, Iguape, and Ilha Comprida, including approximately 125 square kilometers of mapped mangrove. Marine terraces, sandy barrier islands, tidal flats, and estuarine channels create pronounced environmental gradients over short distances, meaning that mangrove stands just kilometers apart can experience very different conditions of salinity, tidal inundation, and freshwater delivery. Much of the mangrove lies within state parks, extractive reserves, and the Juréia-Itatins Ecological Station, which limits direct clearing but does not shield the estuary from hydrological and sedimentary changes originating in the surrounding catchments. At Cananéia, the mean annual air temperature is about 22.1 degrees Celsius and annual precipitation approximately 2,109 millimeters, with rainfall exceeding 200 millimeters per month from January through March and dropping to its lowest values during June through August.

Methodologically, the research harnessed the strengths of two very different data streams. Thermal conditions were represented exclusively by the ERA5 reanalysis of daily mean two-meter air temperature produced by the European Centre for Medium-Range Weather Forecasts, which combines a forecast model with assimilated observations on an atmospheric grid of roughly 0.25 degrees. Vegetation was tracked through three MODIS products: the normalized difference vegetation index (NDVI) and the enhanced vegetation index (EVI) at 250-meter resolution, and the leaf area index (LAI) at 500-meter resolution, all from Collection 6.1. These three metrics are not interchangeable. NDVI summarizes the contrast between red and near-infrared reflectance but can saturate in dense vegetation; EVI corrects for some atmospheric and background effects and remains more responsive at high biomass; LAI is a retrieved structural variable estimating one-sided leaf area per unit of ground area. The common analysis window ran from January 2003 through December 2019, and all processing, from quality filtering to mangrove masking and grid alignment, was performed in Google Earth Engine.

Trend detection relied on the nonparametric Mann-Kendall test paired with Sen’s median slope estimator, robust choices for climate series that may contain outliers and non-normal distributions. Temperature-vegetation associations were quantified with Pearson correlation coefficients and ordinary least-squares regression at three temporal scales: annual, month-specific, and fixed calendar quarters (January-March, April-June, July-September, and October-December). The authors deliberately used calendar quarters rather than standard austral meteorological seasons because these groupings capture four distinct phases of the local annual cycle: peak summer warmth and rainfall, autumn cooling, the coolest and driest interval, and the spring recovery of both temperature and rainfall. Crucially, the team analyzed associations at two spatial scales, averaging across all valid mangrove pixels and also computing cell-by-cell correlations for each 250-meter pixel.

The headline annual result is striking: warmer years were associated with greener mangroves. Annual temperature correlated positively with NDVI (r = 0.61, R² = 0.38, p = 0.009) and with EVI (r = 0.55, R² = 0.30, p = 0.022), suggesting that as the atmosphere warmed, the canopy became more photosynthetically active or denser. Yet the LAI relationship was not statistically significant (r = 0.27, R² = 0.07, p = 0.304), a divergence indicating that greenness and canopy structure did not respond in lockstep. This split between spectral indices and a structural variable may reflect genuine ecological differences in how leaf chemistry and leaf area respond to climate, but it may also stem from the products themselves, which differ in compositing interval, spatial resolution, saturation behavior, and retrieval algorithms.

Within the year, the sign of the association flipped in ways that carry real ecological meaning. During the July-September quarter, NDVI and EVI correlated negatively with temperature (r = -0.31 and -0.28, respectively), while in October-December, EVI and LAI showed positive associations (r = 0.44 and 0.48). In January-March, LAI was strongly and positively associated with temperature (r = 0.61), and in April-June all three indices showed modest positive coefficients. The authors propose a hydroclimatic explanation: during the cool, dry winter interval, relatively warm conditions can raise atmospheric evaporative demand without a matching increase in rainfall, potentially stressing the trees, whereas warmer conditions in summer and spring coincide with recovering freshwater inputs. These mechanisms remain hypotheses, however, because precipitation, salinity, tidal inundation, radiation, and soil moisture were not included in the models.

The spatial maps add another layer of complexity. Significant grid-cell relationships occupied limited and discontinuous areas rather than sweeping across the estuary. Moderate positive EVI coefficients covered about 32 percent of the domain annually, with strong positive coefficients confined to roughly 2.3 percent, concentrated south of Iguape and in central Cananéia. Negative coefficients appeared in parts of Cananéia and southern Ilha do Cardoso, and no single mechanism fully explains them; differences in stand structure, salinity exposure, geomorphology, disturbance history, and the interpolation of the atmospheric field are all candidate factors. A June 2019 disturbance near Papagaios Island, which reduced both NDVI and LAI, illustrates how episodic events can distort relationships summarized by quarter-mean temperature. The authors are equally candid about statistical limitations: thousands of spatial tests were performed without a false discovery rate correction, so isolated significant cells may be false positives, and the coarse ERA5 grid, smoothed by bicubic interpolation onto the 250-meter MODIS grid, cannot resolve tidal-channel microclimates or canopy temperature.

Perhaps the most valuable aspect of the study is its disciplined interpretation. Because the design is observational and bivariate, the authors explicitly caution that the reported correlations should not be read as causal or as responses to temperature alone. The area-mean series provides stronger evidence of warming than the cell-level tests, a contrast consistent with short pixel records, the coarse reanalysis grid, and the absence of corrections for temporal and spatial dependence. The mapped associations describe relationships with a regional atmospheric field, not measurements of stand-level mangrove microclimate. This restraint matters for coastal managers and policymakers who might otherwise overinterpret a positive annual correlation as proof that warming will uniformly benefit mangroves at their subtropical edge.

Looking forward, the research team identifies clear priorities. Temperature should remain part of long-term mangrove monitoring in the Cananéia-Iguape system, but future attribution studies must pair it with precipitation, freshwater discharge, salinity, tidal inundation, and radiation, supported by local sensors and permanent field plots. Models incorporating lagged responses, a four-month rainfall lag has already been reported in this estuary, would provide a firmer foundation. The September transition, where the strongest warming coincides with recovering rainfall and rising atmospheric demand, emerges as the critical window for focused investigation. As mangroves worldwide face rising seas, shifting climates, and mounting human pressure, studies like this one demonstrate both the power of open satellite archives to illuminate remote coastal ecosystems and the care required to translate correlation into understanding.

Subject of Research: Satellite-based assessment of air temperature trends and their associations with mangrove vegetation indices in the Cananéia-Iguape Coastal System, Brazil

Article Title: Assessing mangrove sensitivity to thermal variations through a satellite-based analysis on the southern São Paulo coast, Brazil

Article References: Baratto, J., de Bodas Terassi, P. M., de Lima, N. G. B., & Galvani, E. (2026). Assessing mangrove sensitivity to thermal variations through a satellite-based analysis on the southern São Paulo coast, Brazil. Environmental Monitoring and Assessment, 198(11), Article 1169. https://doi.org/10.1007/s10661-026-16005-6

Image Credits: AI Generated

DOI: 10.1007/s10661-026-16005-6

Keywords: mangroves, climate change, remote sensing, MODIS, ERA5 reanalysis, NDVI, EVI, leaf area index, Mann-Kendall test, São Paulo coast, Cananéia-Iguape estuary, subtropical ecology

The team harnessed more than two decades of satellite observations from NASA’s MODIS instruments, computing three complementary vegetation indices monthly from 2001 to 2025: the Normalized Difference Vegetation Index (NDVI), the Enhanced Vegetation Index (EVI), and the Modified Soil Adjusted Vegetation Index (MSAVI). Each index captures a different facet of canopy condition. NDVI, the workhorse of vegetation remote sensing, responds strongly to chlorophyll content but can saturate in dense canopies. EVI incorporates blue-band reflectance and improved atmospheric correction, making it better suited to the high-biomass tropical forests where NDVI loses sensitivity. MSAVI, designed to suppress the signal of bare soil, proves invaluable in the intertidal zones where tides repeatedly expose mudflats between the trees.

Processing this volume of data required cloud computing. The researchers used Google Earth Engine to aggregate quality-filtered MODIS scenes into monthly composites, then clipped every raster to the mangrove extent defined by the Global Mangrove Watch Version 3.0 dataset, a product built from L-band radar data from the ALOS PALSAR satellite fused with Landsat optical imagery, achieving an overall classification accuracy of 94 percent. In total, 720 monthly raster layers, covering 240 months across three indices, were processed and then collapsed into national monthly time series by averaging all mangrove-masked pixels across India’s twelve mangrove-bearing states and union territories.

The forecasting engine at the heart of the study is the Seasonal Autoregressive Integrated Moving Average model, or SARIMA. Unlike simpler trend analyses, SARIMA explicitly models both the slowly evolving trend of a time series and its recurring seasonal cycles, making it well matched to ecosystems whose phenology is locked to monsoon rhythms. Before fitting any model, the team tested each vegetation index series for stationarity using the Augmented Dickey-Fuller and KPSS tests. Both tests rejected their respective null hypotheses for all three indices, pointing to seasonal non-stationarity, so the researchers applied one non-seasonal and one seasonal difference to stabilize each series before model selection.

Seasonal decomposition using the STL method revealed just how rhythmically these forests breathe. Seasonal strength coefficients ranged from 0.813 for MSAVI to 0.855 for EVI, far above the 0.3 threshold commonly used to flag strong seasonality. Intriguingly, all three indices peaked during the dry season and bottomed out during the wet season, a counterintuitive pattern the authors attribute to the unique physiology of mangroves, which are adapted to tidal and salinity stress rather than limited by freshwater availability. The trend components told a quieter story: relatively flat through the 2000s, then climbing steadily from around 2010 onward, consistent with the documented 1.2 percent annual growth in India’s mangrove area driven by conservation policies and restoration programs.

Automated model selection using R’s auto.arima function, guided by the Akaike Information Criterion, produced a distinct model for each index. NDVI required a first-order autoregressive structure, EVI relied on a fourth-order moving average with no autoregressive term, and MSAVI demanded a third-order autoregressive specification, likely reflecting the complex interplay of tidal cycles, soil moisture, and vegetation phenology that soil-adjusted indices register. Residual diagnostics were clean across the board: Ljung-Box tests found no remaining autocorrelation in any model, although formal normality tests revealed slight heavy-tailedness, a familiar feature of vegetation index series where post-monsoon greening surges occasionally exceed what a Gaussian distribution would predict.

The critical test came from data the models had never seen. The final five years of observations, from 2021 to 2025, were withheld as an out-of-sample validation set. Against this holdout, the SARIMA models achieved mean absolute percentage errors between 2.5 and 4.9 percent, well inside the 10 percent threshold conventionally considered high-accuracy forecasting. NDVI emerged as the most dependable index, with a MAPE of just 2.50 percent and the only statistically significant improvement over a seasonal-naive benchmark in Diebold-Mariano comparisons. EVI and MSAVI performed respectably in absolute terms but were statistically indistinguishable from simpler benchmarks, and their prediction intervals covered fewer observations than intended, signaling that forecast uncertainty for those indices was underestimated.

With the models validated, the team refit each specification to the full 2001 to 2025 record and generated operational forecasts for 2026 through 2030. All three indices are projected to continue their gentle ascent: NDVI is expected to average about 6.4 percent higher than the 2001 to 2020 training-period mean, while EVI and MSAVI are projected to rise by roughly 9.8 and 10.7 percent respectively. The larger relative gains in MSAVI hint at new vegetation colonizing tidally exposed substrates, where a soil-adjusted index is most responsive, while EVI’s rise is consistent with canopy structural recovery in previously degraded high-biomass zones. NDVI’s more modest climb likely reflects saturation in already dense, mature stands rather than an absence of underlying progress. Crucially, all projections remain within the bounds of historical variability, suggesting stable to improving conditions under current human and climatic pressures.

The authors are careful to frame these results as a statistical baseline rather than a guarantee. Because the analysis aggregated all mangrove pixels nationally, the forecasts cannot resolve state-level or site-specific trajectories, and the study itself notes that localized studies have documented declines, such as falling NDVI in Bhitarkanika and widespread EVI-based browning across much of the Sundarbans. The prediction intervals capture only parameter and stochastic uncertainty, not structural risks like major cyclones or land-use change. Moreover, vegetation indices are optical proxies for canopy greenness, not direct measures of biomass, carbon stocks, or physiological health, and MODIS’s 500-meter pixels can blur the narrow fringing mangrove belts of Gujarat and Kerala with surrounding water and aquaculture ponds.

Even with those caveats, the framework offers conservation managers something they have lacked: a quantitative, forward-looking early warning system. Deviations from forecast trajectories could flag ecosystem stress before it becomes visible on the ground, while the strong, predictable seasonality identified in the data points to optimal windows for restoration planting, roughly November through January, when mangroves show peak physiological resilience. Because the entire pipeline runs on open satellite data, Google Earth Engine, and open-source R and QGIS tools, the approach is readily transferable to mangrove regions across the tropical and subtropical world, offering a scalable template for predictive ecology at a moment when coastlines everywhere face mounting climate pressure.

Subject of Research: SARIMA time-series forecasting of satellite-derived mangrove vegetation indices across India

Article Title: SARIMA-based forecasting of mangrove vegetation indices across India

Article References: Suhail S., M., Sharma, L. K., & Bhaveshkumar, K. I. (2026). SARIMA-based forecasting of mangrove vegetation indices across India. Discover Oceans, 3(1), Article 64. https://doi.org/10.1007/s44289-026-00179-5

Image Credits: AI Generated

DOI: 10.1007/s44289-026-00179-5

Keywords: mangroves, SARIMA, time series forecasting, remote sensing, vegetation indices, NDVI, EVI, MSAVI, MODIS, Google Earth Engine, India, coastal conservation

The challenge the researchers set out to address is deceptively complex. Vegetation water stress expresses itself through coupled changes in canopy greenness, leaf area, surface temperature, and evaporative behavior, all of which can be observed from space. Satellite-derived indicators such as the Normalized Difference Vegetation Index (NDVI), the Enhanced Vegetation Index (EVI), Leaf Area Index (LAI), Fraction of Photosynthetically Active Radiation (Fpar), the Evaporative Stress Index (ESI), and Land Surface Temperature (LST) provide temporally consistent, spatially exhaustive information on crop condition. Yet translating these multiple, partly redundant signals into a reliable single estimate of vegetation stress has remained difficult, particularly in irrigated deltas where human water-management decisions strongly modulate how plants respond to climate forcing.

The study uses MODIS satellite products covering 2018 to 2025, with NDVI and EVI drawn from the MOD13Q1 product at 250-meter resolution, LAI and Fpar from MOD15A2H, LST from MOD11A2, and actual and potential evapotranspiration from MOD16A2, from which ESI was calculated as the ratio of actual to potential evapotranspiration. Quality-assurance layers were used to exclude pixels contaminated by clouds, shadows, aerosols, or low-confidence retrievals, and the data were aggregated into monthly governorate-level values. Observations from 2018 to 2023 served for model training, while 2024 and 2025 were reserved as an independent testing period, supplemented by 5-fold cross-validation to assess temporal robustness.

A central methodological innovation lies in the coupling of Best Subset Regression (BSR) with machine learning. Rather than feeding every available indicator into the models, BSR systematically evaluated all possible combinations of LAI, Fpar, ESI, NDVI, and LST, selecting the subset that best balanced explanatory power and parsimony using criteria such as adjusted R², the Akaike and Bayesian Information Criteria, and Mallows’ Cp. The Variance Inflation Factor was used to screen out multicollinearity. This physically grounded variable reduction was then paired with four machine-learning architectures: a multilayer perceptron (MLP) neural network, a Random Subspace (RS) ensemble, Regression by Discretization (RD), and Random Forest (RF).

The feature-selection analysis produced a strikingly clear hierarchy of controls. NDVI emerged as the overwhelmingly dominant predictor of EVI, with a correlation of 0.934 and a standardized regression coefficient of 0.935, supported by an extraordinarily large t-statistic of nearly 199 and a narrow, strictly positive confidence interval. LAI contributed a smaller but statistically significant refinement, capturing canopy structural attributes such as layering and leaf clumping that a two-band greenness index cannot fully represent. By contrast, Fpar, ESI, and LST showed no detectable linear contribution once NDVI and LAI were accounted for, suggesting that in this irrigated setting their influence is either mediated through greenness signals or expressed through nonlinear interactions, and that managed water supply buffers canopy condition from short-lived thermal or evaporative stress.

The consequences of feature optimization for model performance were substantial, though architecture-dependent. Before selection, Regression by Discretization led the independent testing period, while Random Forest lagged with a correlation of only 0.9571. After BSR, Random Forest transformed into the clear best performer, achieving a correlation of 0.9943, a mean absolute error of just 0.0063, and a root mean square error of 0.0161 during the 2024–2025 testing period, with relative absolute error falling to under 5 percent. Random Subspace also improved markedly, its testing correlation rising from 0.9414 to 0.9738 and its slope moving from 0.4768 to 0.9666, while the multilayer perceptron failed to benefit and deteriorated slightly across all metrics.

Cross-validation reinforced these findings. Across five folds, the optimized Random Forest achieved a correlation of 0.9793 with relative errors dropping to roughly 9.6 and 20.2 percent, demonstrating low bias and low variance across resampled data partitions. The authors attribute this behavior to the underlying data structure: because EVI is governed by a near one-dimensional NDVI–EVI relationship with a modest incremental contribution from LAI, tree-based ensembles can exploit repeated splits on these dominant predictors to form many shallow, low-bias trees, while bagging averages out noise. The neural network, with its larger parameterization relative to the effective information content, remained more sensitive to sample variation.

To confirm that the optimized Random Forest was not simply outperforming weak baselines, the team benchmarked it against univariate and multiple linear regression, XGBoost, and CatBoost. The linear baselines produced testing RMSE values of approximately 0.053, while the boosting algorithms reduced this to around 0.045 and 0.046. The feature-optimized Random Forest, however, retained a substantially lower RMSE of 0.0161, roughly a third of the boosting models’ error, confirming a genuine methodological advantage rather than an artifact of comparison. Time-series analysis further showed that after feature selection the model reproduced the timing and amplitude of seasonal greening and senescence cycles with markedly tighter agreement.

The practical implications extend to operational water management in semi-arid regions. Because NDVI and LAI are routinely available from satellite platforms at frequent intervals, the framework could be updated as new observations arrive, supporting governorate-scale vegetation-condition monitoring. Estimated EVI could then be converted into crop- and season-specific anomalies and combined with crop-water requirements and irrigation schedules to inform allocation decisions. The authors are careful to note limits: the framework estimates contemporaneous vegetation condition rather than forecasting future stress, and its findings are specific to Dakahliyah Governorate and the 2018–2025 period. Validation across longer records, contrasting irrigation regimes, and additional regions, along with the incorporation of microwave soil moisture and field-scale data, remains a priority before broader transferability or early-warning use can be claimed.

Nevertheless, the study offers a compelling demonstration that less can be more in agricultural machine learning. By systematically identifying which satellite indicators genuinely carry information about crop water stress, and by discarding the rest, the researchers built a model that is simultaneously more accurate, more interpretable, and cheaper to operate, requiring only a small set of widely available inputs. As water scarcity intensifies across semi-arid agricultural regions worldwide, such parsimonious, physically grounded AI frameworks may become essential instruments for safeguarding food production under increasingly uncertain climatic conditions.

Subject of Research: Estimation of agricultural vegetation water stress using feature-optimized remote sensing and ensemble machine learning in the Nile Delta

Article Title: Estimation of agricultural vegetation water stress using feature-optimized remote sensing and ensemble machine learning

Article References: Elbeltagi, A., Srivastava, A., & Alsumaiei, A. A. (2026). Estimation of agricultural vegetation water stress using feature-optimized remote sensing and ensemble machine learning. Smart Agricultural Technology, 15, Article 102536. https://doi.org/10.1016/j.atech.2026.102536

Image Credits: AI Generated

DOI: 10.1016/j.atech.2026.102536

Keywords: vegetation water stress, remote sensing, machine learning, random forest, Nile Delta, MODIS, EVI, NDVI, best subset regression, agricultural drought, irrigation management, Egypt