Multitemporal Analysis of the Dynamics of High-Andean Wetlands in the Metropolitan Region of Chile Using Sentinel-2 Images and ERA5-Land Climate Data ^†

Abstract

High-Andean wetlands are critical ecosystems for water regulation and carbon storage. This study analyzes the impact of climate variability on vegetation dynamics in eight wetlands located in the Estero Ortiga sub-basin, within the Los Nogales Nature Sanctuary (Metropolitan Region, Chile), between 2017 and 2024. We used time series of NDVI and NDCI indices derived from Sentinel-2 imagery (January 2017 to September 2024), along with monthly temperature and precipitation data from ERA5-Land (January 2016 to September 2024). Trends were assessed through linear regression, and vegetation–climate relationships were analyzed using Pearson correlations with a one-year lag. Results show a progressive decline in vegetation cover (slope: −2.04 × 10⁻⁵ NDVI) and chlorophyll content (slope: −1.15 × 10⁻⁵ NDCI), with strong positive correlations between annual precipitation and vegetation indices in the subsequent summer (R = 0.83–0.88 for NDVI; R = 0.84–0.90 for NDCI). Annual NDVI reclassification highlighted a reduction in healthy vegetation cover from 2020 onward. This research provides novel evidence linking climate trends and vegetation health in high-Andean wetlands, reinforcing the utility of satellite-based indicators for conservation monitoring.

1. Introduction

The increase in greenhouse gas emissions has elevated global temperatures by 1.55 ± 0.13 °C above pre-industrial levels, intensifying climatic events and impacting vulnerable ecosystems such as high-Andean wetlands [1,2]. These wetlands provide essential ecosystem services, including water regulation, carbon storage, and habitats for biodiversity [3,4,5]. In Chile, high-Andean wetlands are located above 3000 m above sea level, distributed across the mountainous areas of the northern, central, and southern regions of the country, and include peatlands, lagoons, and salt flats [3,4].

Due to their exposure to climate change and increasing anthropogenic pressures, monitoring these ecosystems is critical for assessing their status and designing effective management strategies. Remote sensing enables multi-temporal and spatial evaluations of vegetation conditions using spectral indicators such as NDVI (Normalized Difference Vegetation Index) and NDCI (Normalized Difference Chlorophyll Index), which are widely used to estimate plant biomass and productivity [6,7]. These tools support the development of evidence-based conservation measures [8,9].

Although several studies have applied satellite-based vegetation indices to Andean wetlands, many focus on short timeframes, specific phenological events, or localized regions. Several works have analyzed the relationship between vegetation health and climatic variability in semi-arid and temperate environments. Studies in semi-arid grasslands have shown that both the amount and timing of precipitation can significantly influence vegetation cover, with higher sensitivity observed in upland and slope areas compared to riparian zones. Long-term analyses in various regions of East Asia have also revealed strong relationships between vegetation dynamics and climate variability over multiple decades. However, such interactions have been less frequently explored in high-Andean wetlands, particularly in central Chile, where long-term datasets remain limited. Additionally, while the NDVI is widely used to estimate vegetation greenness, the combined use of NDVI and NDCI to simultaneously assess vegetation cover and chlorophyll content remains uncommon [10,11,12,13,14].

In this context, the present study evaluates vegetation behavior in response to climatic variables in eight high-Andean wetlands located in the upper, middle, and lower sections of the Estero Ortiga sub-basin, within the Los Nogales Nature Sanctuary (SNLN), Metropolitan Region of Chile, during the period 2017–2024. We analyzed trends in NDVI and NDCI time series derived from Sentinel-2 imagery and examined their relationship with average temperature and annual accumulated precipitation obtained from the ERA5-Land monthly data product. By incorporating both vegetation indices and climate variables over an eight-year period, this study provides new evidence on how interannual climate variability (particularly changes in precipitation) affects vegetation dynamics and ecological health in high-altitude wetlands. The results aim to support the development of adaptive monitoring tools and conservation strategies to address the impacts of ongoing climatic stress in these vulnerable ecosystems.

2. Materials and Methods

2.1. Study Area

Los Nogales Nature Sanctuary is located in the Metropolitan Region of Chile, between coordinates 33°19′46.441″ S; 70°26′52.526″ W and 33°6′19.015″ S; 70°20′52.037″ W. The high-Andean wetland system, situated in the western sector, is composed of eight wetlands, accounting for 1.7% of the total evaluated area (1914 ha).

Figure 1 illustrates the location of the SNLN and the distribution of the wetlands under study, classified according to their altitudinal position: Upper (US), Middle (MS), and Lower (LS), with surface areas ranging from 0.95 to 11.22 hectares (

Table 1. Area of Andean wetland analyzed.

Surface Wetlands (ha)
2US	4US	5US	1MS	8MS	3LS	6LS	7LS
11.22	1.31	0.92	8.19	1.05	3.23	2.83	3.92

).

2.2. Methodology

Figure 2 summarizes the methodological workflow used to assess vegetation–climate relationships in high-Andean wetlands within the Los Nogales Nature Sanctuary. Two types of remote sensing data were utilized: climate variables (temperature and precipitation) obtained from the ERA5-Land dataset (2016–2024), with a spatial resolution of 11.1 km and monthly temporal resolution [15]; and vegetation indices (NDVI and NDCI) derived from Sentinel-2 imagery (2017–2024) [16], which has a spatial resolution of 10 m and a revisit time of 5 days. The Sentinel-2 images were accessed through the EO Browser platform [17].

Sentinel-2 imagery was selected based on the availability of Level-2A surface reflectance products, which include atmospheric correction using the Sen2Cor processor (European Space Agency, ESA, Frascati, Italy). The images were accessed and processed via the EO Browser platform (Sinergise Ltd., Ljubljana, Slovenia). To minimize the influence of clouds, only images with less than 10% cloud cover were used. A cloud-masking algorithm based on the Scene Classification Layer (SCL) was applied in Google Earth Engine to exclude pixels classified as cloud, cloud shadows, and cirrus. For each month, a single image was selected using the median pixel value from the available cloud-free scenes, generating a consistent monthly composite time series for each wetland.

NDVI and NDCI were calculated using band combinations in Google Earth Engine, and monthly time series were generated for each wetland. Additionally, NDVI values corresponding to the month of maximum vegetation activity each year were reclassified into vegetation condition categories. This reclassification followed the thresholds proposed by López et al. (2015) [17], which define five classes: (i) clouds, snow, and water (NDVI < 0.01); (ii) bare soil (0.01–0.1); (iii) light vegetation (0.1–0.2); (iv) medium vegetation (0.2–0.4); and (v) dense or healthy vegetation (NDVI > 0.4). This classification enabled the evaluation and comparison of vegetation cover dynamics across the studied wetlands.

Finally, Pearson correlation analysis was applied to evaluate the relationship between annual accumulated precipitation and average vegetation index values during the following summer season. Although Pearson’s R assumes normality and the absence of autocorrelation, no formal statistical tests were conducted to verify these assumptions, which represents a methodological limitation of this study. Nonetheless, the approach was considered appropriate for exploring first-order linear relationships between climatic and vegetation variables at the interannual scale.

2.3. Climatic Data

To describe climatic conditions for the period 2017–2024, we developed time series for average air temperature (measured at 2 m above ground) and total accumulated precipitation (including rain and snow) using data from January 2016 to September 2024. The inclusion of 2016 data allowed us to assess the influence of prior-year climatic conditions on vegetation development during the subsequent summer (2017). Temperature data were used for descriptive purposes only and were not included in statistical analyses.

2.4. NDVI and NDCI Indices

To characterize vegetation dynamics, we calculated the NDVI and NDCI indices using Equations (1) and (2). NDVI and NDCI values were computed for each wetland to obtain monthly averages and to generate monthly time series from January 2017 to September 2024 for each wetland [16,18].

$$NDVI={\displaystyle \frac{NIR \left(B08\right)-Red\left(B04\right)}{NIR \left(B08\right)+Red\left(B04\right)}}$$

(1)

Equation (1). NDVI calculation [7].

$$NDCI={\displaystyle \frac{RedEdge \left(B05\right)-Red\left(B04\right)}{RedEdge \left(B05\right)+Red \left(B04\right)}}$$

(2)

Equation (2). NDCI calculation [19].

2.5. Statistical Analysis

To assess the relationship between annual accumulated precipitation and average summer (January–March) NDVI and NDCI values across altitudinal sections (high, middle, and low), we applied the Pearson Correlation Coefficient [20]. We calculated R, R², RMSE, MAE, and p-values to analyze differences between index values, interannual variability in precipitation, and the topographical position of the wetlands.

3. Results

3.1. Climatic Description

Figure 3 displays the monthly accumulated precipitation, annual cumulative precipitation, and average air temperature from December 2016 to February 2025. A thermal seasonality is observed, with summer peaks (e.g., up to 19.3 °C in January 2017) and winter lows (e.g., down to −1.8 °C in July 2022). Precipitation events are concentrated in the winter months and exhibit high interannual variability, ranging from 291.7 mm (June 2024) to 1.6 mm (March 2020). Between 2016 and 2022, precipitation showed a declining trend, which corresponds to the megadrought period in central Chile [21].

3.2. NDVI and NDCI Time Series

The evolution of NDVI and NDCI indices from 2016 to 2024 is illustrated in Figure 4 and Figure 5. Both series reveal differentiated patterns depending on wetland type and topographic position. NDVI shows a general decreasing trend (slope = −2.62 × 10⁻⁵), represented by the dotted line in Figure 4. This line corresponds to a linear regression fitted across the entire NDVI time series, serving as a visual indicator of the long-term decline in vegetation cover. Although no statistical significance test was conducted on this slope, it is interpreted as a descriptive and exploratory indication of the temporal evolution of vegetation cover. Wetlands in the middle (1MS, 8MS) and upper (2US) zones maintained consistently higher values throughout the period, while those located in the lower section (3LS, 6LS, 7LS) presented intermediate values with regular seasonal fluctuations. Similarly, NDCI also displays a negative trend (slope = −1.09 × 10⁻⁵). As with NDVI, no statistical test was applied to determine the significance of this trend; however, it serves as a reference for describing general variations in chlorophyll content over time. The largest wetland (1MS) consistently showed the highest NDCI values, while wetlands in the upper section exhibited the lowest levels.

Multitemporal variation in the indices for each wetland is reflected in vegetation cover proportions. For instance, strong contrasts were found in the persistence of vegetation between 2017 and 2024 in the largest wetlands (2US and 1MS). In 2US, high vegetation cover decreased drastically in 2020 (32.3%), accompanied by an increase in medium (45.9%) and light vegetation (19.7%). Although progressive recovery occurred, it did not exceed 60% until 2024. In contrast, 1MS maintained high vegetation above 85% for most of the study period, except in 2020 (57.0%), with recovery reaching 88.8% in 2024. These trends are summarized in

Table 2. Vegetation cover classification for wetlands 1MS and 2US (2017–2024 period).

Wetland	Date	Bare Soil		Light Vegetation		Median Vegetation		High Vegetation
		Surface (ha)	Percent (%)	Surface (ha)	Percent (%)	Surface (ha)	Percent (%)	Surface (ha)	Percent (%)
2US	19 January 2017	0.06	0.41	0.49	3.36	2.95	20.22	11.09	76.01
	24 January 2018	0.07	0.48	0.66	4.52	3.18	21.8	10.68	73.2
	04 January 2019	0.07	0.48	1.17	8.02	4.17	28.58	9.18	62.92
	24 December 2020	0.31	2.12	2.87	19.67	6.69	45.85	4.72	32.35
	04 March 2021	0.34	2.33	2.19	15.01	5.35	36.67	6.71	45.99
	03 January 2022	0.25	1.71	2.61	17.89	6.13	42.02	5.6	38.38
	13 January 2023	0.22	1.51	1.87	12.82	5.58	38.25	6.92	47.43
	27 February 2024	0.19	1.3	1.25	8.57	4.36	29.88	8.79	60.25
1MS	19 January 2017	-	-	0.01	0.1	0.42	4.19	9.6	95.71
	24 January 2018	-	-	0.02	0.2	0.69	6.88	9.32	92.92
	04 January 2019	0.01	0.1	0.19	1.89	1.23	12.26	8.6	85.74
	24 December 2020	0.02	0.2	0.14	1.4	4.15	41.38	5.72	57.03
	04 March 2021	-	-	0.01	0.1	2.03	20.24	7.99	79.66
	03 January 2022	-	-	0.12	1.2	4.47	44.57	5.44	54.24
	13 January 2023	-	-	0.08	0.8	2.39	23.83	7.56	75.37
	27 February 2024	-	-	0.02	0.2	1.1	10.97	8.91	88.83

, which presents the annual NDVI-based classification of vegetation cover for both wetlands. In both cases, NDVI-based classified mosaics reveal a decline in high vegetation in 2020 and a subsequent increase (Figure 6).

3.3. Correlation Analysis

The analysis revealed a statistically significant (p < 0.05) and positive relationship between annual accumulated precipitation and average summer values of NDVI and NDCI indices across the three altitudinal zones of the study area. A stronger fit was observed in the upper section, while greater dispersion was noted in the middle and lower sections. This pattern suggests that wetlands in the upper zone exhibit greater dependency on precipitation than those in lower zones for both NDVI (Figure 7) and NDCI (Figure 8). In these figures, each blue dot represents a single year′s observation. The blue line shows the linear regression fitted between annual precipitation and vegetation indices, and the shaded band around the line represents the confidence interval of the model, illustrating the range of uncertainty associated with the estimated relationship.

4. Discussion

The analysis of vegetation indices and their correlation with accumulated annual precipitation between 2017 and 2024 reveals a progressive decline in vegetation cover and productivity in the studied high-Andean wetlands of the Los Nogales Nature Sanctuary. These results are consistent with the prolonged drought documented for central Chile in the 2016–2022 period [22].

The significant correlations found between annual precipitation and summer average values of NDVI and NDCI demonstrate the high sensitivity of these ecosystems to hydrological cycles [23]. Additionally, the response appears to be influenced by wetland size. For example, the contrasting behavior observed in wetlands 1MS and 2US (selected for their differing surface areas) illustrates their varied responses to dry periods. These differences reflect the distinctive seasonal response patterns of each wetland [24] and highlight the importance of adaptive monitoring tailored to the specific dynamics of each ecosystem.

However, the observed variability between sites cannot be fully explained by climatic drivers alone. Local anthropogenic pressures (such as livestock grazing, trail formation, and proximity to infrastructure) may alter vegetation structure and soil compaction, affecting wetland resilience. Moreover, differences in topographic configuration, groundwater input, and surrounding land use can modulate hydrological connectivity and vegetation response. These elements should be considered in future analyses to better understand the spatial heterogeneity observed across the wetlands.

Finally, the use of satellite imagery (Sentinel-2) combined with updated climate data (ERA5-Land) facilitates the detection of cumulative changes in ecosystem conditions [25,26]. This integrated approach supports the design of long-term, multitemporal monitoring systems for conservation and restoration planning. As such, remote sensing proves to be a fundamental tool for managing large-scale ecological information and supporting long-term decision-making in the face of climate change.

5. Conclusions

The multitemporal analysis using Sentinel-2 satellite imagery and ERA5-Land climate data allowed the characterization of both seasonal and interannual vegetation dynamics in eight high-Andean wetlands between 2017 and 2024, confirming the dependence of these ecosystems on interannual precipitation variability. This behavior was more pronounced in wetlands located in the upper section, highlighting the need to prioritize their monitoring and conservation.

To advance the characterization of high-Andean wetlands, long-term studies are needed that integrate the relationship between annual accumulated precipitation, summer values of spectral indices, and other influencing factors such as wetland size or anthropogenic pressures.

Although there is a difference in spatial resolution between Sentinel-2 imagery (10 m) and ERA5-Land data (11.1 km), this was not considered a relevant limitation in this study, since the analysis was conducted at the sub-basin scale and climatic conditions in the study area are relatively homogeneous. Likewise, although significant correlations were found, these do not imply causality. Future studies incorporating field measurements and hydroecological variables are needed to deepen the analysis.

Remote sensing stands out as a valuable tool for the systematic monitoring of vulnerable ecosystems. Its ability to generate functional and temporal indicators enables the early detection of degradation processes, the identification of critical areas, and supports informed decision-making for the development of adaptive management strategies in the context of climate change.