|
|
|
|
|
|
|
|
|
Sentinel‑2 NDVI Time‑Series Assessment of Vegetation Response to Floodwater Spreading in the Soh Aquifer
|
Mahdi Hashmi *   , Ali Dastranj |
| Isfahan |
|
|
Abstract: (9 Views) |
Introduction: Vegetation is a cornerstone of hydrological regulation and soil conservation in arid and semi-arid ecosystems. In Iran's central plateau, the confluence of decreasing precipitation, rising evapotranspiration, and excessive groundwater abstraction has critically disturbed the water-soil-vegetation balance (Noori et al., 2021). Annual groundwater depletion averages 49 cm, with non-renewable extraction exceeding 5.4 km³/year. In response, floodwater spreading has been implemented as a Managed Aquifer Recharge (MAR) strategy to enhance deep percolation and soil moisture. Despite its application, quantitative evidence of its long-term ecological impacts on vegetation remains scarce. While some studies report significant NDVI increases, others find negligible effects, indicating a clear research gap. This gap is characterized by the lack of homogeneous control areas, short monitoring periods (<5 years), and insufficient multi-scale trend analyses. This study introduces a novel approach by integrating a 9-year Sentinel-2 time series (2015–2024) with a homogeneous control site, robust Theil–Sen trend analysis, and non-parametric statistics to assess vegetation response in the Soh aquifer (Meymeh, Isfahan). The objectives were: (1) comparing mean NDVI between treated and control areas; (2) analyzing monthly, seasonal, and annual NDVI trends; and (3) statistically testing distribution differences.
Methodology: The study was conducted in the Soh aquifer basin (51°20'–51°22'E, 33°17'–33°20'N), encompassing an 800-hectare floodwater spreading site and an adjacent control area (<1 km apart). The control was selected based on rigorous homogeneity in topography, soil properties (loamy-sand), geology (Quaternary alluvium), climate, and vegetation (predominantly shallow-rooted rangeland and planted Prunus dulcis). Due to the absence of continuous field data, Sentinel-2 L2A imagery (10 m spatial resolution, <5% cloud cover) from 2015–2024 was acquired via Google Earth Engine. Monthly median composites were generated to minimize optical noise, producing 110 monthly observations per area, alongside seasonal and annual NDVI time series. NDVI was calculated using the standard formula (NIR – Red)/(NIR + Red). Trends were evaluated using Ordinary Least Squares (OLS) regression and the robust Theil–Sen slope estimator to mitigate outlier effects. Given the non-normal distribution of the data (confirmed by Shapiro–Wilk, p < 0.05), the non-parametric Mann–Whitney U test was applied to statistically compare NDVI distributions between the two areas. All statistical computations were performed using Python (SciPy, NumPy) in the Visual Studio Code environment.
Results and Discussion: Descriptive statistics revealed that the long-term mean NDVI in the floodwater spreading area (0.078) was nearly identical to the control (0.079). Critically, the Mann–Whitney U test confirmed no statistically significant difference (U = 7959.5, p = 0.057). However, the spreading area demonstrated notably lower temporal variability (SD: 0.011 vs. 0.012; CV: 0.141 vs. 0.152) and a higher maximum NDVI (0.119 vs. 0.115). Annual means peaked in 2020 for both areas (0.091 vs. 0.090) and hit their lowest in 2017 (0.068). Trend analyses at all scales revealed no significant directional changes. Annual OLS slopes were nearly zero (spreading: -0.000333, p=0.755; control: -0.000350, p=0.750). Seasonal trends were non-significant (all p>0.05), with the control's winter slope being the largest negative at -0.001433 (p=0.346). Monthly analyses showed highest positive slopes in June/July and the most negative in April, yet all p-values remained above 0.05 (smallest being April in control, p=0.115). Collectively, no significant upward or downward trends were detected in either area. The absence of a significant mean NDVI increase is scientifically attributed to several interacting factors:
Hydro-Ecological Mismatch: The shallow root systems (<30–50 cm) of the dominant rangeland species cannot access the deep-stored water (>1 m) recharged by spreading operations.
Initial Site Disturbance: Mechanical operations (excavation, leveling) during project establishment disturbed soil structure and depleted the native seed bank, requiring a multi-year recovery period.
Fine Sediment Accumulation: Repeated flood events deposit silt and clay, forming a denser surface layer that, while preserving permeability, physically impedes seed germination of annual and perennial herbaceous species.
Immature Tree Growth: Many planted almond trees are still in their vegetative phase, requiring up to 10 years to reach full canopy cover and significantly contribute to the mean NDVI.
Ecological Time Lag: Systematic evidence confirms that MAR’s positive effects on vegetation often manifest after a multi-year lag, frequently exceeding a decade. The marginal p-value (0.057) supports this, suggesting the effect is on the threshold of emergence.
Despite the unchanged mean, the reduced variability and higher maximum NDVI indicate a significant ecological achievement: enhanced ecosystem resilience. The spreading area shows a "buffering effect," preventing severe NDVI declines in dry years while allowing greater greenness peaks in wet years (e.g., 2020).
Conclusion: This study concludes that over the 9-year period, floodwater spreading in the Soh aquifer did not significantly increase mean NDVI compared to a homogeneous control area (p = 0.057). However, it significantly reduced temporal fluctuations and increased maximum greenness, enhancing ecosystem stability and resilience. No significant trends were observed, confirming long-term stability. The lack of mean increase is explainable by root-depth mismatch, initial disturbances, sedimentation, and ecological inertia. Critically, this study demonstrates that evaluating such projects solely on mean index changes is insufficient.
Management Recommendations: (1) Gradually replace shallow-rooted rangelands with deep-rooted species like Amygdalus scoparia and Nitraria schoberi; (2) Extend monitoring to at least 15 years; (3) Integrate soil moisture sensors and piezometers for direct hydrological assessment; (4) Manage surface sediment accumulation.
Limitations: Lack of field validation data, absence of detailed operational records (water volume/frequency), and reliance on a single spectral index (NDVI). Future studies should incorporate complementary indices (EVI, SAVI, NDMI). This study provides a replicable, scientifically robust framework for assessing floodwater spreading impacts across Iran’s arid aquifers, emphasizing that resilience improvement is a primary and measurable success indicator.
|
Article number: 5 |
| Keywords: Floodwater spreading, NDVI, Soh aquifer, Remote sensing, Sentinel-2, Mann –Whitney U test |
|
|
|
Type of Study: Research |
Subject:
Special
Received: 2026/05/16 | Revised: 2026/06/29 | Accepted: 2026/06/29 | ePublished ahead of print: 2026/07/5
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
|
|
|
|
|
|
|
|
