Dynamics of Wetlands in Ifrane National Park, Morocco: An Approach Using Satellite Imagery and Spectral Indices

Abstract

Our study aims to analyze the spatiotemporal dynamics of six lakes in Ifrane National Park (Morocco) using remote sensing and satellite imagery over the period 2000–2024. Spectral indices such as NDWI, MNDWI, EWI, AWEI, and ANDWI were employed to extract water bodies from Landsat images, while the NDVI index was used to identify irrigated agricultural lands. Additionally, the SPEI and RDI indices were applied to assess the impact of climate fluctuations on the hydrological evolution of the lakes. The results reveal an alarming reduction in lake surface areas, with some lakes having completely dried up. This decline is correlated with decreased precipitation and the expansion of irrigated agricultural lands, highlighting the impact of human activities. The analysis of hydrological correlations between lakes demonstrates significant interactions, although some indices show disparities. The rapid expansion of agricultural land, particularly arboriculture, increases pressure on water resources. These changes threaten local biodiversity and heighten the socio-economic vulnerability of surrounding populations.

1. Introduction

Water bodies play a fundamental role in environmental balance and human well-being. These ecosystems, considered true natural treasures, provide crucial services to both ecosystems and society [1,2,3]. Indeed, they serve as freshwater reservoirs, essential for consumption, agriculture, and industry [4,5]. Furthermore, water bodies contribute to local climate regulation by storing and releasing heat while supporting exceptional biodiversity by hosting numerous animal and plant species [6]. As a result, most of these water bodies attract tourism and provide resources for fishing [7,8].

However, these exceptional natural environments have become particularly vulnerable to the effects of climate change and anthropogenic pressures, including pollution, overexploitation of water resources, infrastructure development, and construction [9,10]. In this context, monitoring the dynamics of water bodies is essential to understanding their hydrological functioning and planning management interventions when necessary.

In situ measurements are the most reliable and recommended method for tracking the dynamics of water bodies over time and assessing their ecological status [11]. However, this approach is not always feasible, particularly in developing countries, due to the high costs of fieldwork, necessary equipment, and the vast extent of water bodies [12]. To overcome these challenges, remote sensing and satellite imagery analysis have become key tools. These methods enable the monitoring of water level variations, changes in surface area, and water quality over large regions using satellites such as Sentinel-2, Landsat, or MODIS [13,14,15].

Remote sensing provides extensive spatial coverage, making it ideal for monitoring large surfaces or hard-to-reach areas [16]. It allows for continuous temporal monitoring thanks to the regular frequency of satellite data, facilitating the detection of seasonal variations, long-term trends, and abrupt changes [17]. This method also reduces costs and effort compared to in situ measurements by minimizing human intervention in the field while increasing data acquisition speed. Moreover, remote sensing can be easily integrated with other technologies, such as hydrological models or Geographic Information Systems (GIS), to provide a comprehensive view and decision-making tools for sustainable management [18,19].

Over the past four decades, the application of remote sensing techniques for hydrological monitoring and landscape analysis has significantly increased [11,14,20]. Spectral water indices, which are computationally inexpensive, easy to implement, and transferable across different platforms, have become an attractive approach for delineating water bodies using remote sensing data [21].

Several spectral indices have been proposed to extract water bodies from raw satellite images. The Normalized Difference Water Index (NDWI), introduced by McFEETERS (1996) [22], is one of the most commonly used. However, Xu (2006) [23] pointed out that the NDWI could lead to significant errors due to built-up areas (asphalt, roofs), vegetation, and soil. He thus proposed a modified version, the Modified Normalized Difference Water Index (MNDWI), which uses the Shortwave Infrared (SWIR) band instead of the Near-Infrared (NIR).

Other indices have also been developed, such as the Normalized Difference Pond Index (NDPI) proposed by Lacaux et al. (2007) [24] or the Automatic Water Extraction Index (AWEI) by Feyisa et al. (2014) [15], designed to reduce errors caused by shadows and dark surfaces. However, these indices remain subject to classification errors in certain scenarios, particularly in areas containing low-albedo surfaces (dark vegetation, asphalt, shadows) or turbid waters.

Nevertheless, a thorough understanding of the field remains essential to confirming the accuracy of results obtained through satellite products, particularly in specific conditions such as dust storms over water bodies, muddy waters, hydrothermal waters, ice, and frozen water bodies [25]. These situations can significantly reduce water classification accuracy by increasing commission errors (misclassification of non-aquatic surfaces as water) and omission errors (failure to detect certain water areas).

This study aims to analyze, using remote sensing and satellite imagery analysis, the spatiotemporal dynamics of six emblematic water bodies within the Ifrane National Park and its peripheral zone in Morocco. These lakes represent rich ecosystems that host exceptional biodiversity, characteristic of the region’s wetlands [26]. The ecological and hydrological significance of the first four lakes has earned them the designation of RAMSAR sites, an international recognition highlighting their value as wetlands of global importance for conservation.

These ecosystems play a key role not only for local biodiversity, providing habitats for numerous species, particularly migratory birds, but also for local communities by offering essential ecosystem services and significant tourism and educational opportunities [27]. However, these lakes currently face increasing pressures due to environmental and anthropogenic factors, including declining precipitation and the overexploitation of water resources for irrigation. Water stress, exacerbated by climate change and the expansion of irrigated agricultural land, has led to a decrease in the piezometric level of the hydrogeological reservoir and an alarming regression of wetland surface areas. This situation poses a serious threat to the ecological balance of these lakes, endangering their ability to support the ecosystems that depend on them.

The study of these hydrosystems aims to assess their hydrological dynamics, identify the factors and their interaction responsible for their degradation, and propose solutions for sustainable management. This approach aligns with a strategy to preserve this natural heritage for future generations while raising awareness of their importance as biodiversity reservoirs and indicators of environmental changes in the region.

2. Methodology and Study Area

2.1. Study Area

The selected water bodies in this study are located within the Ifrane National Park and its peripheral zone. This territory is a protected area due to its significant ecological and hydrological value. It plays a key role in regulating water resources upstream of the Sebou and Oum Er-Rbia watersheds, acting as a powerful water reservoir that ensures the sustainability of surface water flow [28]. It also contributes to the conservation of numerous endangered species, including the Atlas cedar, an endemic species and a unique symbol of the Mediterranean region.

Due to their major ecological importance and to promote their conservation along with their associated flora and fauna, some areas of the park have been designated as RAMSAR sites [29]. The park’s ecosystems also include oak forests, wetlands, and grasslands, supporting a diverse range of wildlife, including iconic species such as the Barbary macaque, the Barbary sheep, the Barbary deer, and the porcupine, as well as threatened birds such as the ruddy shelduck, the red-knobbed coot, the golden eagle, and the Egyptian vulture [30].

Officially established in October 2004, the park initially covered an area of 51,800 hectares before being expanded to 125,000 hectares in 2006. Its main objectives focus on biodiversity conservation, forest ecosystem protection, habitat preservation, and ecosystem restoration. The adopted strategy emphasizes environmental education and public awareness, the promotion of ecotourism as a sustainable local development solution, and the rational management of natural resources [26].

Ifrane National Park, located in the central Middle Atlas of Morocco (Figure 1 and Figure 2), is renowned for its distinct mountainous Mediterranean climate and contrasting landscapes. With altitudes ranging between 1200 and 2400 m, the region’s climate is influenced by its elevation and unique local conditions [31]. The mountains and dense forests play a crucial role in shaping climatic conditions, leading to significant spatial variations in temperature and precipitation. During winter, temperatures often drop below freezing, with averages ranging between −1 °C and 10 °C, resulting in frequent snowfall over the plateaus and mountainous areas [32]. In contrast, summers remain mild and dry, with average temperatures fluctuating between 15 °C and 28 °C.

Regarding precipitation, due to the foehn effect caused by the Middle Atlas relief, the plateau zone receives the highest amount of rainfall, with an annual average exceeding 800 mm in Ifrane, compared to the foothills, where annual averages are only approximately 550 mm [31].

Relative humidity is higher, especially in winter, and the dominant winds from the north and west bring cool and humid air from the mountains and the Atlantic Ocean. Despite its inland location, Ifrane benefits from the moderating influence of the cleanliness, contributing to its unique microclimate.

This climate favors the growth of lush forests, particularly cedars, oaks, and pines, and supports a diverse fauna, including (Barbary macaques) and various bird species. The snowy winters and mild summers of Ifrane make it an exceptional natural destination, attracting both locals and visitors.

The studied lakes include Dayet Iffer, Dayet Afourgaa, Dayet Aoua, Dayet Ifrah, Dayet Hachlaf, and Afennourir. Except for Lake Afnourir, which is located in the center of the PNI, the other lakes are close to each other with an average distance of 13 km as the crow flies and are located in the northern part of the study area. These wetlands host a rich diversity of fauna and flora, including more than 50 bird species, some of which are threatened, such as the common pochard (Aythya ferina) and the white-headed duck (Oxyura leucocephala). Additionally, these lakes play a crucial role as wintering habitats for migratory species such as the ruddy shelduck (Tadorna ferruginea), especially when other breeding sites in the Middle Atlas are insufficiently flooded.

Despite being part of Ifrane National Park, Dayet Aoua (512 hectares), Dayet Hachlaf (70 hectares), and Dayet Ifrah (250 hectares) (

Table 1. Geographic coordinates, altitudes, and maximum surface area of the lakes studied.

Lake Name	Latitude	Longitude	Altitude (m)	Area (ha)
Dayet Aoua	33.654795	−5.040634	1468	512
Dayet Ifrah	33.559808	−4.927445	1614	250
Dayet Hachlaf	33.547118	−5.002192	1668	70
Dayet Afennourir	33.280131	−5.252495	1800	800
Dayet Yffer	33.606774	−4.907936	1508	7
Dayet Afourgaa	33.614352	−4.877976	1415	12

) face major challenges linked to the overexploitation of groundwater for irrigation and declining precipitation. These factors led to the complete drying up of Dayet Aoua and Dayet Hachlaf in 2019 and Dayet Ifrah in 2024. This situation highlights the environmental challenges facing these wetlands, which rely heavily on variations in the piezometric level of the karstic hydrogeological reservoir. Restoration efforts have begun, and a rainwater channeling project has been launched to revitalize the ecological, landscape, cultural, and socio-economic potential of Dayet Aoua, with an initial budget of 6.1 million dirhams, or 600,000 euros.

Lake Afennourir, located on the Ain Leuh plateau at an altitude of 1800 m, covers approximately 800 hectares (Table 1) and is surrounded by cedar and holm oak forests. It serves as a vital refuge for both local and migratory wildlife, particularly aquatic birds that use it as a wintering, breeding, and migration site. Recognizing its ecological significance in surface water regulation and groundwater table stabilization, it was the first Moroccan site to be listed as a Ramsar site in 1980. However, like many mountain lakes, Afennourir faces challenges exacerbated by climate change, as its water supply heavily depends on surface runoff. Efforts are underway to expand the drainage area, protect and restore natural vegetation, and promote the sustainable management of human activities around the lake.

Dayet Iffer (7 hectares) and Dayet Afourgaa (12 hectares) (Table 1) are two small natural lakes located in the Middle Atlas, approximately 5 and 8 km from Dayet Ifrah. Their water levels depend on the groundwater table. Dayet Iffer is surrounded by a forest of cedars, holm oaks, and maritime pines, while Dayet Afourgaa is bordered by private and communal lands, where agricultural activities are rapidly expanding. These two wetlands provide a picturesque setting, attracting visitors from Fez and Sefrou for activities such as hiking, picnicking, and birdwatching.

2.2. Methodology

2.2.1. Satellite Image Analysis

This study aims to understand the evolution dynamics of the wetlands in Ifrane National Park and its peripheral zone by analyzing satellite images from Landsat 4, 5, 8, and 9, covering the period from 2000 to 2023. In selecting the images, we focused on those from April, which correspond to the peak recharge period of hydrogeological reservoirs and coincides in the Mediterranean region with the end of winter, rising temperatures, and increased evapotranspiration.

The monitoring was conducted by analyzing the changes in the wet surface area of the study wetlands over the selected periods. The satellite images used in this study were carefully chosen to minimize cloud cover, which could interfere with the results. Atmospheric and radiometric corrections were applied to enhance image quality.

Unfortunately, data is unavailable for certain years (2002, 2003, 2006, and 2012), creating gaps in the time series. These missing data were completed using an interpolation method based on temporal linear regression, based on the trends observed over the available years. To ensure the consistency of the resulting time series, a trend analysis was performed using the non-parametric Mann-Kendall test, both before and after interpolation. This approach ensures that the direction and significance of the trends are not altered by the data completion process.

To assess the evolution of irrigated agricultural areas surrounding the studied wetlands—a factor that may impact the hydrological functioning of these water bodies—we analyzed the Normalized Difference Vegetation Index (NDVI) from satellite images corresponding to the years 2001, 2014, and 2024. The images from 2001 and 2014 were obtained from Landsat 5 and Landsat 8, respectively, while the 2024 image was derived from Sentinel-2 data. To avoid the influence of natural vegetation, which mainly grows in spring, and to minimize cloud interference, which is more frequent in winter, we selected images taken during the summer period.

A literature review on remote sensing applied to hydrology indicates that most researchers use a set of indices considered to be more reliable and robust for extracting water bodies from satellite images. Each of these indices has specific advantages and limitations as well as a particular field of application. In this study, we selected the five most recommended indices, which are NDWI, MNDWI, EWI, AWEI, and ANDWI. The main objective of choosing these five indices is to compare and validate the results obtained against each other.

• The NDWI (Normalized Difference Water Index):

It is a widely known index in remote sensing for identifying and monitoring water surfaces, such as lakes, rivers, reservoirs, and wetlands [22]. It is calculated using the spectral bands of satellite images in the near-infrared (NIR) and green (Green) bands, according to the following formula:

$$NDWI={\displaystyle \frac{Green-NIR}{Green+NIR}}$$

(1)

• The MNDWI Index (Modified Normalized Difference Water Index):

It is a modified version of the NDWI, used to improve the detection of water bodies, especially in environments where artificial surfaces, such as buildings or bare soil, could affect the results of the classic NDWI [17]. The MNDWI replaces the near-infrared (NIR) band with the shortwave infrared (SWIR) band and is calculated using the following formula:

$$MNDWI={\displaystyle \frac{Green-SWIR}{Green+SWIR}}$$

(2)

• The EWI Index (Enhanced Water Index):

This is an index primarily used to improve the detection and quantification of water surfaces. Unlike other indices such as the NDWI or MNDWI, the EWI incorporates additional bands and applies specific adjustments to better discriminate water areas under complex conditions, such as wet soils, urban areas, or dense vegetation [12]. There is no universal formula for the EWI, as it can vary depending on the sensors and adjustments required for the specific application [33]. However, its calculation often relies on an optimized combination of spectral bands from the visible spectrum, particularly green and red (Green; RED), near-infrared (NIR), and shortwave infrared (SWIR) [1]. In this study, we opted to calculate this index using the following formula:

$$EWI={\displaystyle \frac{Green-RED}{NIR+RED+SWIR}}$$

(3)

• The AWEI Index (Automated Water Extraction Index):

Feyisa et al. (2014) [15] introduced two versions of this index (AWEInsh and AWEIsh). The first uses the weighted difference in green (Green) and the sum of the near-infrared (NIR) and shortwave infrared (SWIR 1–2) bands, and it has been shown to effectively eliminate commission errors related to shadows and roofs in urban areas, while the second takes into account the weighted difference between the sum of blue and green (Blue; Green) and the sum of the near-infrared (NIR) and shortwave infrared (SWIR 1–2) bands. In this study, since our study area is in a forest environment, far from the impacts of human developments and infrastructures that could disrupt the results of water body extraction, we chose to use only AWEInsh. The latter was calculated using the following formula:

$$AWEInsh=4\times \left(Green-SWIR1\right)-(0.25\times NIR+2.75\times SWIR2)$$

(4)

• The ANDWI Index (Augmented Normalized Difference Water Index):

This is a robust water index that surpasses its predecessors by successfully separating water pixels from non-aquatic environments. Compared to other indices, even though the near-infrared (NIR) band has the highest reflectance in muddy and hydrothermal waters, the inclusion of blue and red bands in the ANDWI results in the correct classification of these types of water, while AWEInsh may misclassify them [34]. Additionally, the ANDWI is expected to conceptually outperform the MNDWI by eliminating noise generated by dark vegetation. The difference between the blue, green, and red bands and the near-infrared (NIR) and shortwave infrared (SWIR 1–2) bands of dark vegetation is negative, and therefore, they are classified as non-aquatic by the ANDWI index [2]. In contrast, the difference between the green bands and the shortwave infrared (SWIR1) bands in the MNDWI results in the misclassification of dark vegetation as water. The ANDWI allows for the identification of water bodies with higher reflectance in RGB compared to that of the NIR and SWIR1-2 combined, which cancels out the very high reflectance of the NIR in the case of hydrothermal and muddy waters (unlike NDWI and AWEInsh). Similarly, the ANDWI can correctly classify waters obscured by dust storms and can suppress noise generated by roofs in urban areas where the reflectance in the SWIR bands is higher than in clear water conditions (but not higher than the reflectance in RGB). This index is calculated using the following formula:

$$ANDWI={\displaystyle \frac{Blue+Green-Red-NIR-SWIR1-SWIR2}{Blue+Green+Red+NIR+SWIR1+SWIR2}}$$

(5)

For the classification of the results obtained from the indices used, we set the value 0 as the threshold to distinguish aquatic pixels from non-aquatic pixels. This classification was validated using a comparison with the color composition of the images, historical data from Google Earth, as well as our thorough knowledge of the field.

These indices used in this study have certain limitations. Indeed, the NDWI has certain limitations, notably its sensitivity to shadows from trees and buildings, which can lead to false positives in the detection of water surfaces. Furthermore, in areas of dense vegetation, it can confuse wet vegetation with water bodies, thus reducing its reliability in certain specific contexts. The MNDWI, on the other hand, is highly dependent on the quality of SWIR data, which can limit its effectiveness in regions where these data are degraded. It also remains sensitive to artificial surfaces, which can influence the results and distort the identification of water areas. The EWI, for its part, suffers from a lack of standardization of the formulas used, which makes comparison between different studies difficult. Its complexity lies in the need to master the specific adjustments required for each application, which can limit its use to experts with the appropriate knowledge. AWEI, while effective in certain conditions, presents significant algorithmic complexity, requiring precise adjustments and advanced calculations. It can also face difficulties in dense urban environments, where reflective surfaces often disrupt analyses. Finally, ANDWI, while outperforming other indices in certain situations, shows limitations in the detection of highly turbid waters, where its performance can be compromised. Furthermore, its effectiveness varies depending on environmental conditions and the characteristics of the sensors used, highlighting a variability that can affect its widespread use.

2.2.2. Trend and Breakout Tests

To assess the existence of a significant trend in the time series of water body surfaces, we applied the non-parametric Mann-Kendall test. Furthermore, to identify possible breakpoints in the time series, i.e., sudden changes in the behavior or mean of the data, we used the Pettitt test.

• Mann-Kendall Trend Test

To analyze the trends in the time series of the water surface areas of the lakes studied, we applied the classic non-parametric Mann-Kendall test [35,36]. Statistical trend tests identify and/or estimate the existence or not of a trend in a time series according to the desired degree of significance. Mann-Kendall is a non-parametric test requiring only that the data be serially independent, without assuming the normality of the distribution [37,38]. The Mann-Kendall statistic S is calculated as follows:

$$S=\sum _{i=1}^{n-1}\sum _{j=i+1}^{n}sgn\left(x_{j-}{x}_i\right)$$

(6)

where n is the length of the data series and xi and xj are the data values in time series i and j (j > i), respectively, with sgn (xj − xi) responding to the following sign function:

$$sgn\left(x_j-x_i\right)=\left\{\begin{array}{c}+1, if x_j-x_i>0\\ 0, if x_{j-x_i=0}-1, if x_j-x_i<0\end{array}\right\}$$

(7)

The variance (V) of S is calculated as:

$$V\left(s\right)={\displaystyle \frac{n\left(n-1\right)\left(2n+5\right)-{\sum }_{k=1}^m{t}_k\left(t_k-1\right)\left(2t_k+5\right)}{18}}$$

(8)

In the equation number 3, n is the number of data points, m is the number of linked groups, and tk is the number of links of extent k. A related group is a set of data samples with the same value. In cases where the sample size n is greater than 10, the standard normal test statistic ZS is calculated using the following equation:

$$Z=\left\{\begin{array}{c}{\displaystyle \frac{S-1}{\sqrt{var\left(s\right)}}}, if S>0\\ 0, if S=0\\ {\displaystyle \frac{S+1}{\sqrt{var\left(s\right)}}}, if S<0\end{array}\right\}$$

(9)

Positive Z values indicate increasing trends, while negative values indicate decreasing trends. The trend test is performed at a specific α significance level. When |Z| > Z1 − α/2, the null hypothesis is rejected and a significant trend exists in the time series. Z1 − α/2 is obtained from the standard normal distribution table [39]. In this study, the significance levels chosen are: α = 0.01 (or 99% confidence intervals), α = 0.05 (or 95% confidence intervals), and α = 0.1 (or 90% confidence intervals). At the level of significance of 1%, 5%, and 10%, the null hypothesis “absence of trend” is rejected if |Z|> 2.57, 1.96 ≤ |Z| <2.57, and 1.64 ≤ |Z| 1.96 [40], respectively. Kisi and Ay [41] stated that in the case of long-term rainfall trend analysis, the MK test performs better than parametric tests.

• Pettitt’s Test

Pettitt’s Test. This is a powerful method for highlighting the stationary or nonstationary nature of time series. The test examines the existence of a break at an unknown time (t) in the series, using a formula derived from Mann–Whitney. According to the Pettitt test, $x_i$, $x_2$, $x_3$,…,$x_n$ is an observed data series that has a change point at t such that $x_i$, $x_2$,…,$x_t$ has a distribution function $F_1\left(x\right)$ that is different from the distribution function $F_2\left(x\right)$ of the second part of the series $x_{t+1}$, $x_{t+2}$, $x_{t+3}$,…, $x_n$ [42]. The Ut non-parametric test statistic for this test can be described as follows:

$$U_{t }=\sum _{i=1}^t\sum _{j=t+1}^{n}sing (x_t-x_j),$$

(10)

$$Sign \left(x_i-x_j\right)=\left[\begin{array}{c}1, if \left(x_i-x_j\right)>0,\\ 0, if \left(x_i-x_j\right)=0,\\ -1, if \left(x_{i-}{x}_j\right)<0\end{array}\right]$$

(11)

The test statistic K and the associated confidence level ($\rho$) for the sample length (n) can be described as follows:

$$K=max\left|U_t\right|$$

(12)

$$\rho =exp\left({\displaystyle \frac{-K}{n^2+n^3}}\right)$$

(13)

When $\rho$ is below the specific confidence level, the null hypothesis is rejected. The approximate probability of significance ($P$) for a point of change is defined as follows:

$$P=1-\rho$$

(14)

Clearly, when there is a significant change point, the series is segmented at the change point into two subseries. The K test statistic can also be compared with standard values at different confidence levels for the detection of a change point in a series.

The Pettitt test is applied in our study at the 5% significance threshold (α = 0.05).

2.2.3. Climate Indices

To highlight the role of climate fluctuations in the change of the surface area of the studied wetlands, we opted for two drought indices: the SPEI (Standardized Precipitation Evapotranspiration Index) and the RDI (Reconnaissance Drought Index). These indices are particularly effective for visualizing dry and wet periods in a climate series. The data used includes the monthly totals of liquid precipitation as well as the monthly averages of maximum and minimum temperatures recorded at the Ifrane meteorological station over the period from 1991 to 2024.

• Reconnaissance Drought Index (RDI):

The Reconnaissance Drought Index (RDI), designed by Tsakiris and Vangelis in 2005 [35], was created to address the limitations of some indices such as the Standardized Precipitation Index (SPI) and the Palmer Drought Severity Index (PDSI). Indeed, the SPI focuses solely on precipitation to assess water deficit, and the PDSI lacks sensitivity in drought monitoring [43]. The RDI is based on the ratio between cumulative precipitation and potential evapotranspiration (PET). The initial value of the index, called a^k, is calculated by dividing the sum of precipitation over a given period by the sum of potential evapotranspiration over the same period. This relationship is expressed by the following equation:

$$a_k{\displaystyle \frac{{\sum }_{j=1}^{j=k}{P}_{ij}}{{\sum }_{j=1}^{j=k}{PET}_{ij}}}$$

(15)

where $P_{ij}$ and ${PET}_{ij}$ represent precipitation and potential evapotranspiration for the $j^{th}$ month of the $i^{th}$ year, respectively.

This ratio is then standardized using equations similar to those used for the SPI to obtain the RDI values. The drought severity classifications for the RDI are identical to those of the SPI, and this index can also be defined at multiple scales.

• The Standardized Precipitation Evapotranspiration Index (SPEI)

The Standardized Precipitation Evapotranspiration Index (SPEI), proposed by Vicente-Serrano et al. in 2010 [44], was developed to integrate the sensitivity of the PDSI to variations in evaporative demand while retaining the multi-scale characteristic of the SPI. This index is particularly effective for detecting and monitoring the impact of climate change on drought conditions [45]. The SPEI is based on a simple climatic water balance, calculated as the weekly or monthly difference between cumulative precipitation and potential evapotranspiration. This relationship is mathematically expressed as follows:

$$D_i=P_i-{PET}_i$$

(16)

The calculated values of ‘D’ are then aggregated at different time scales as follows:

$$D_n^k={\sum }_{i=0}^{k-1}(P_{n-i}-{PET}_{n-i})$$

(17)

where k (in months) represents the time scale of aggregation and n is the month of calculation.

2.2.4. Mapping the Land Evolution of the Park

Mapping agricultural lands is based on the NDVI (Normalized Difference Vegetation Index), which is commonly used to monitor vegetation cover changes. This index allows for quantifying the density and health of vegetation, making it a valuable tool for monitoring agricultural areas. In addition, a land area classification was carried out based on our in-depth knowledge of the terrain, acquired through regular visits and direct observations. This approach enabled the creation of more accurate land cover maps, tailored to local specifics and reinforced by our field expertise.

3. Results

First, after calculating the selected water indices, we extracted the pixels corresponding exclusively to water for each year and for each of the studied lakes. This allowed us to generate annual rasters representing the lake surfaces, thus facilitating the analysis of their spatiotemporal evolution. A preliminary analysis of the reflectance of each land cover element was carried out to better control the classification and extraction of all pixels representing water, as shown in Figure 3.

Three main types of land cover are distinguished: water (lakes and rivers), vegetation (forests and agricultural lands), and bare soils, taking into account the specific environmental characteristics of the studied region. Figure 3 shows low reflectance of water in all bands, particularly in the NIR (near-infrared) and SWIR (shortwave infrared) bands, 0.086 and 0.082, respectively. This corresponds to calm and low-turbidity water bodies, such as those found in the dayets of the Middle Atlas, notably Dayet Aoua, Dayet Ifrah, and Dayet Afennourir. These water bodies, often fed by inputs from hydrogeological reservoirs, rainfall, or snowmelt, exhibit reflectance that varies with the seasons. However, for the period chosen in our study, which coincides with the end of April, the reflectance in all spectral bands is optimal due to the absence of precipitation, floods, or dense aquatic vegetation, elements that could disrupt reflectance results. Bare soils, common in the mountainous areas and rocky terrains of the Middle Atlas, show high reflectance in the NIR and SWIR bands, 0.44 and 0.52, respectively, reflecting their dry nature and often calcareous or clayey composition. Additionally, forest and agricultural vegetation cover is characterized by a peak reflectance in the NIR, 0.257. High reflectance in this band indicates dense and healthy foliage, while a decrease in the SWIR reflects the water content of the plants.

Figure 4 addresses periods of total drought for Dayet Aoua Lake in 2001, 2008, and 2014. After 2016, a general downward trend in the lake surface area was clearly observed. Since 2020, the lake has been permanently dry until today. The curves for Dayet Ifrah Lake show a steady growth in its area from 2000 to 2016, reaching a peak of approximately 120 hectares in 2013 before gradually decreasing after 2018. Unlike Dayet Aoua, only one period of total drying is observed in this lake in 2001. In 2024, this lake also became completely dry. Dayet Hachlaf shows a period of low areas, close to 0, between 2000 and 2008. After a strong growth dynamic between 2010 and 2016, reaching a peak of approximately 66 hectares in 2013, the lake experienced a rapid decline after this date and became dry from 2020 onwards. Regarding the annual hydrological dynamics of Dayet Afennourir Lake, Figure 4 shows significant fluctuations, with peaks reaching 120 hectares in 2018, followed by a gradual decrease until 2022. A partial recovery is observed in 2023, indicating some resilience of the lake in the face of recent hydrological fluctuations.

Dayet Iffer shows a relatively small area, being a small, closed lake, and remains stable throughout the study period, with peaks not exceeding 3 hectares. The observed stability, with moderate fluctuations, is well captured by all hydrological indices, suggesting that this lake is less affected by major variations but remains vulnerable due to its small size. The hydrological functioning of Dayet Afourgaa Lake shows significant fluctuations, similar to those observed at Dayet Aoua, particularly between 2000 and 2015. The areas reach peaks of more than 6 hectares in 2013. After 2018, a stable downward trend is observed, and this lake becomes dry from 2022 onwards.

In order to verify the divergences as well as the similarities that may link the different lakes studied, correlations between the index results were carried out. Depending on the level of correlation, values close to 1 indicate the presence of hydrological similarities between the two compared hydrosystems, while values close to 0 reflect a divergence in their functioning.

The results obtained (Figure 5) show that the AWEI index reveals a strong homogeneity in correlations, with values often ranging between 0.7 and 0.8, indicating positive and stable interactions between the areas. Similarly, the NDWI highlights generally strong correlations, often exceeding 0.7 between certain lakes, such as Dayet Ifrah and Dayet Hachlaf. In contrast, the EWI and MNDWI indices show greater variability, with some very weak correlations, possibly reflecting differences specific to the indices used to measure the interactions.

Analysis of these results indicates that the hydrosystems of Yffer, Ifrah, and Afourgaa exhibit relatively similar hydrological behavior. This is likely linked to their location within the same dolomitic hydrogeological block (Guigou causse), as well as their perfect alignment on the eastern side with the Tizi-n-Tghten fault field, which separates the Ifrane and Imouzzer causses to the northwest from that of Guigou to the southeast.

Lake Hachlaf, although belonging to the Ifrane causse, shows strong hydrological similarity with Ifrah and Afourgaa. This can be explained by the combined effect of its exposure to similar climatic and lithological conditions, as well as its northwestern alignment with the same Tizi-n-Tghten fault field.

On the other hand, Dayet Aoua, despite evolving under climatic and lithological conditions similar to those of Hachlaf, Ifrah, Yffer, and Afourgaa, does not show similarities in functioning and hydrological response with these hydrosystems. This divergence is likely related to the impact of excessive water withdrawals for irrigation purposes. It suggests that the hydrogeological reservoir feeding the lake is probably under high pressure due to the expansion of irrigated agriculture.

As for Lake Afennourir, despite its distance from the other lakes and its location within a different hydrogeological block—the Aïn Leuh causse—most indices reveal a similarity with Ifrah. However, these results should be interpreted with caution, as the two hydrosystems exhibit completely different behaviors. Indeed, the evolution of Lake Ifrah’s size is closely linked to fluctuations in the piezometric level of the Guigou causse’s hydrogeological reservoir, whereas Afennourir, due to the impermeability of its substratum, is mainly fed by surface runoff; its size therefore depends on the amount and nature of precipitation.

Regarding the temporal variability of lake surfaces (

Table 2. Coefficient of variation of the series of lake surface areas studied for each index between 2000 and 2024.

	Dayet Aoua	Dayet Ifrah	Dayet Yffer	Dayet Afourgaa	Dayet Hachlaf	Dayet Afennourir
NDWI	118.8	77.1	65.1	99.4	227.9	76.5
MNDWI	99.5	56.1	51.8	89.1	134.4	46.7
ANDWI	120.7	61.0	63.0	101.3	174.8	55.8
EWI	99.5	56.0	51.8	89.1	134.4	46.7
AWEI	93.1	56.1	52.5	75.5	143.2	48.1

), Dayet Aoua Lake shows the highest values of the coefficient of variation (CV), with peaks for the ANDWI (120.7%) and NDWI (118.8%) indices. This suggests that the water surfaces of Dayet Aoua vary significantly over time when measured with these indices. The other indices, although having slightly lower values, also show significant variability, confirming that the surface of this lake is unstable. For Dayet Ifrah Lake, the coefficients of variation are lower than those of Dayet Aoua, with a maximum CV of NDWI (77.1%). This reflects a moderate variability of the calculated surfaces, meaning that Dayet Ifrah undergoes fewer fluctuations compared to some of the other lakes. The values of MNDWI (56.1%) and EWI (56.0%) are almost identical, showing low variation according to these indices. Thus, Dayet Ifrah has a water surface that is slightly sensitive to hydrological variations.

Dayet Hachlaf shows the highest coefficients of variation, 227.9% for the NDWI, which is by far the highest value among all lakes and indices. This high variability indicates that the calculated water surfaces for Dayet Hachlaf fluctuate greatly. The other indices, although having lower CV values than the NDWI, remain high, with an ANDWI (174.8%) and an AWEI (143.2%), confirming that Dayet Hachlaf is a particularly unstable lake in terms of water surface. In contrast, Dayet Afennourir shows the lowest CV values among all the lakes. The MNDWI (46.7%) and the EWI (46.7%) display the lowest coefficients, meaning that the water surfaces measured by these indices are relatively stable over time. Even the NDWI (76.5%), which is higher than the other indices, remains well below the values observed for Dayet Hachlaf, indicating that Dayet Afennourir experiences much fewer fluctuations. This suggests that the surface of Dayet Afennourir is much more stable and less subject to environmental or seasonal variations.

Dayet Iffer shows coefficients of variation generally similar to those of Dayet Ifrah, with slightly lower values for AWEI (52.5%) and EWI (51.8%). The NDWI (65.1%) is the index that presents the greatest variability, suggesting that measurements taken with this index are more sensitive to fluctuations in the lake’s water levels. Overall, the calculated surfaces for Dayet Iffer are moderately variable but less subject to major changes compared to lakes like Dayet Aoua. For Dayet Afourgaa, the coefficients of variation also fall within the high value range, with the ANDWI (101.3%) showing the greatest variability. Other indices, such as the MNDWI (89.1%) and the EWI (89.1%), also show notable variability, highlighting that Dayet Afourgaa undergoes significant fluctuations in its water surface area. This may indicate that the lake’s surfaces, as measured by these indices, vary greatly, potentially depending on external conditions.

The analysis of trends using the Mann–Kendall test reveals contrasting dynamics among the studied water bodies (

Table 3. Results of the Mann–Kendall test applied to the time series of lake water surface areas studied between 2000 and 2024.

Water Body	Index	Z-Score	p-Value	Trend Type	Significant? (α = 0.05)
Dayet Aoua	AWEI	−0.71	0.476	No trend	No
	NDWI	−1.24	0.216	Slight decrease	No
	MNDWI	−0.58	0.561	No trend	No
	EWI	−0.53	0.595	No trend	No
	ANDWI	−1.14	0.255	Slight decrease	No
Dayet Ifrah	AWEI	−1.73	0.0832	Decreasing trend	No
	NDWI	−1.77	0.0768	Decreasing trend	No (borderline)
	MNDWI	−1.68	0.0931	Decreasing trend	No
	EWI	−1.77	0.0768	Decreasing trend	No (borderline)
	ANDWI	−1.80	0.0716	Decreasing trend	No (borderline)
Dayet Yffer	AWEI	2.45	0.0142	Increasing	Yes
	NDWI	1.91	0.056	Slight increase	No (borderline)
	MNDWI	2.64	0.0083	Increasing	Yes
	EWI	2.41	0.0159	Increasing	Yes
	ANDWI	2.72	0.0065	Increasing	Yes
Dayet Afourgaa	AWEI	1.49	0.136	Slight increase	No
	NDWI	1.47	0.140	Slight increase	No
	MNDWI	0.99	0.323	No trend	No
	EWI	1.21	0.225	Slight increase	No
	ANDWI	0.94	0.348	No trend	No
Dayet Hachlaf	AWEI	2.01	0.044	Increasing	Yes
	NDWI	1.91	0.056	Slight increase	No (borderline)
	MNDWI	2.43	0.015	Increasing	Yes
	EWI	2.48	0.013	Increasing	Yes
	ANDWI	2.19	0.028	Increasing	Yes
Dayet Afennourir	AWEI	4.20	0.000026	Increasing	Yes
	NDWI	4.26	0.000020	Increasing	Yes
	MNDWI	4.29	0.000018	Increasing	Yes
	EWI	4.26	0.000020	Increasing	Yes
	ANDWI	4.16	0.000031	Increasing	Yes

). Afennourir stands out with statistically significant increasing trends across all tested indices, indicating a sustained expansion of its surface area over the past two decades. This behavior is particularly noteworthy in the regional context, as the lake is exclusively fed by rainfall and exhibits an almost immediate response to precipitation events.

Hachlaf and Yffer also exhibit significant upward trends for four out of the five indices, although these trends are less pronounced compared to Afennourir.

In contrast, Afourgaa shows no statistically significant trend, despite the Z-scores being mostly positive across the indices. Indeed, the associated p-values, all exceeding 0.05, suggest the absence of a robust statistical signal.

For Dayet Aoua, all trend signals are slightly negative, indicating no statistically significant trend for any of the indices. This situation suggests a slow degradation process that remains undetectable at the statistical level.

Finally, Dayet Ifrah displays moderate downward trends, with p-values ranging between 0.0716 and 0.0931 across the five indices. Although these values do not reach the conventional 5% significance threshold, they lie close to it, hinting at a possible latent decline.

The analysis of hydrological breakpoints using the Pettitt test (

Table 4. Results of the Pettitt test applied to the time series of lake water surface areas studied between 2000 and 2024.

Water Body	Index	Year of Change	p-Value	Significance Level
Dayet Aoua	AWEI	2016	0.016	Significant
	NDWI	2016	0.019	Significant
	MNDWI	2016	0.019	Significant
	EWI	2016	0.019	Significant
	ANDWI	2016	0.034	Significant
Dayet Ifrah	AWEI	2009	0.058	Weakly significant
	NDWI	2009	0.012	Very significant
	MNDWI	2009	0.058	Weakly significant
	EWI	2008	0.054	Weakly significant
	ANDWI	2009	0.034	Significant
Dayet Yffer	AWEI	2010	0.012	Very significant
	NDWI	2010	0.006	Very significant
	MNDWI	2009	0.015	Significant
	EWI	2009	0.012	Very significant
	ANDWI	2006	0.027	Significant
Dayet Afourgaa	AWEI	2013	0.010	Very significant
	NDWI	2011	0.030	Significant
	MNDWI	2012	0.020	Significant
	EWI	…	…	…
	ANDWI	…	…	…
Dayet Hachlaf	AWEI	2016	0.07658	Not significant
	NDWI	2009	0.18826	Not significant
	MNDWI	2016	0.07658	Not significant
	EWI	2016	0.07658	Not significant
	ANDWI	2016	0.21156	Not significant
Dayet Afennourir	AWEI	2009	0.0316	Significant
	NDWI	2009	0.0117	Very significant
	MNDWI	2009	0.0270	Significant
	EWI	2009	0.0316	Significant
	ANDWI	2009	0.0098	Very significant

) highlights major discontinuities in the time series of surface water extent, revealing distinct dynamics from one lake to another. Dayet Aoua exhibits a clear rupture in 2016, consistently detected by all indices used, with p-values ranging between 0.016 and 0.034.

Similarly, Dayet Afennourir shows a statistically significant hydrological shift in 2009, with extremely low p-values across all indices (as low as 0.0098), indicating a sudden change in its hydrological behavior.

Despite its small size, Dayet Yffer displays a series of significant breakpoints between 2006 and 2010, particularly for the NDWI in 2010 (p = 0.006), suggesting notable hydrological variability.

In the case of Dayet Afourgaa, the identified breakpoints are spread between 2011 and 2013, with significance levels ranging from p = 0.010 (AWEI, 2013) to p = 0.030 (NDWI, 2011). This temporal distribution suggests a gradual transition rather than a single abrupt shift.

Dayet Ifrah also shows breakpoints at approximately 2009, though with varying levels of statistical significance. While some indices, such as NDWI (p = 0.012), indicate highly significant changes, others (AWEI, MNDWI, EWI) present p-values close to the significance threshold (p ≈ 0.054–0.058), indicating weaker evidence of change.

Finally, Dayet Hachlaf does not exhibit any statistically significant breakpoints according to the indices used (p > 0.07), suggesting relative stability in its surface dynamics over the study period.

The graphs of the hydrological indices calculated for all the lakes present several common aspects. Indeed, all the curves reveal significant fluctuations over the years, reflecting periodic and cyclical variations. In this context, very low values of water surface areas were observed in 2001 and 2005 in all the lakes. This result coincides with marked drought episodes: the first extends from 1998 to 2001, while the second shortest lasts only the year 2005. These periods are characterized by a remarkable rainfall deficit (Figure 6 and Figure 7).

Furthermore, the majority of lakes recorded significant peaks in occupied surface area between 2009 and 2010, which coincides with a wet phase extending from 2009 until the end of 2011. These peaks are followed by a marked decline after 2016, explained by negative values of the RDI and SPEI indices between 2016 and 2018, leading to a progressive decrease in water availability (Figure 6 and Figure 7). Towards the end of the study period, between 2020 and 2024, the indices show lower values, indicating a reduction in the surface areas of water bodies. The graphs also show that a period of severe drought began in 2020 and will continue until 2024.

The different curves representing the indices follow a similar temporal evolution, demonstrating strong coherence in detecting variations in lake surface areas.

The maps in Figure 8 illustrate the evolution of the surface areas of different types of land use (bare soil, vegetable farming, and arboriculture) around the lakes between 2001, 2014, and 2024.

The analysis of the evolution of surface areas according to land use types between 2001 and 2024 highlights a significant transformation linked to agricultural expansion, with notable repercussions on water resources. During this period, the area of bare soil continuously decreased, from 14,646 hectares in 2001 to 12,547 hectares in 2014, and then to 11,718 hectares in 2024, representing an overall decline of 20%. This reduction reflects a conversion of previously unused lands to agricultural uses.

At the same time, vegetable farming experienced significant expansion until 2014, with a 122.01% increase in its area, from 1145 hectares in 2001 to 2542 hectares. However, this trend reversed over the following decade, with a slight decrease of 5.51%, bringing the area down to 2402 hectares in 2024. Over the entire period, the overall growth remains substantial (+109.78%). This type of crop, particularly water-intensive, exerts significant pressure on water resources, especially due to the intensive irrigation practices often associated with vegetable farming. The reduction observed after 2013 suggests a shift in agricultural practices towards arboriculture, particularly towards rosaceous crops.

Arboriculture, on the other hand, shows continuous and spectacular growth. Its area increased from 393 hectares in 2001 to 1057 hectares in 2014 (+168.96%), then doubled to reach 2117 hectares in 2023 (+100.28%), representing an overall increase of 438.68% over the entire period. This rapid conversion to often irrigated arboricultural crops increases pressure on water resources, particularly due to the high-water needs of these crops, especially during the summer. The intensification of arboriculture also contributes to a modification of the local hydrological regime, with an increase in evapotranspiration and a reduction in infiltration.

These transformations indicate increased exploitation of water resources, marked by a rise in withdrawals for irrigation and a decrease in the natural recharge of aquifers. Agricultural expansion, particularly arboriculture and vegetable farming, intensifies pressure on water resources, with an increased risk of deficits, especially during drought periods. These changes can also generate conflicts of use, opposing agricultural needs to those of drinking water or ecosystem preservation.

The apple tree, irrigated using the drip irrigation technique, is the main cultivated species in the study area. Measurements carried out at the Dayet Aoua basin show that the requirements of one hectare of apple trees vary between 6397 m³/ha/year for the density of 833 trees/ha and 12,795 m³/ha/year for the density of 1666 trees/ha, which must be provided from May to October [46].

4. Discussion

Our work shows that the use of satellite remote sensing for detecting and monitoring surface water resources is a robust technique. It provides consistent and continuous data, allowing for the monitoring of water body surface variations year after year. These findings are also shared by several researchers who have emphasized the importance of satellite remote sensing techniques in monitoring surface water resources [47,48,49,50].

The choice of the most robust index among those used in this study is complex, as each index has its advantages and limitations. However, it seems that the results of two indices overlap: the MNDWI and the EWI. Thus, from the perspective of temporal variations, the results of these two indices are the most stable. Several researchers have highlighted the effectiveness of the MNDWI in extracting surface water from satellite images, particularly in complex territories where there is a great diversity of land uses [51,52,53].

The main result of this study shows that today, the reduction in the surface area of lake water bodies leads to a decrease in water resources, a concerning reality. Several of them are on the verge of total desiccation, with significant environmental and socio-economic repercussions. A recent study on lake evolution in Spain showed a marked decrease in their levels in recent years, with a weak correlation with precipitation, which can be explained by their primarily underground water supply [54]. Furthermore, an analysis of wetlands at the global scale revealed that those located in arid and semi-arid zones suffer both from a decrease in the amount of water and a degradation of water quality, directly affecting aquatic ecosystems [55].

In general, the evolution of water body surface area depends on the climatic conditions of the year, whether wet or dry. This also reflects the response of the hydrogeological reservoir to the climatic events of the same year. In other words, any change or fluctuation in precipitation or temperature patterns has a direct impact on the hydrological functioning of these lakes.

In the same context, numerous studies have demonstrated that changes in seasonal and interannual precipitation regimes directly impact the amount of water available in rivers, lakes, and reservoirs [56,57,58,59]. These phenomena collectively influence water systems and increase their vulnerability. Moreover, climate warming, which results in temperatures exceeding the global average, leads to increased evaporation and the drying up of lakes, thus influencing land use changes [60,61,62].

In countries located at low latitudes, including Morocco, evaporation plays a major role in the water and thermal balance of lakes, affecting their stratification and mixing. Studies show a significant increase in lake evaporation in recent decades, a trend that is expected to continue with climate warming, leading to a reduction in freshwater reserves [61,62,63,64]. By the end of the century, a 16% increase in global lake evaporation is expected, especially in arid and semi-arid regions [65,66].

Currently, 971 million people live in high-risk climate zones, particularly in the least developed countries, where vulnerable populations (women, youth, the elderly, and poor populations) are the most affected [67].

Unfortunately, our study area—like much of the Moroccan territory—is covered by only a single meteorological station. This limited coverage significantly constrains the ability to accurately analyze the impact of climatic factors on the dynamics of surface water extent in the lakes of Ifrane National Park.

Moreover, a significant expansion of irrigated agricultural land has been observed around the studied lakes. These agricultural areas rely primarily on groundwater resources for irrigation, which undoubtedly plays a crucial role in the regression of surface water extent in the lakes.

In this context, the impact of land use changes and human activities, particularly agriculture, on water resources is well documented in the hydrological literature and in the field of water resource management. Several studies have shown that the expansion of human activities leads to a reduction in the quantity and quality of freshwater resources. The study of land use changes intensified in the 20th century, revealing significant impacts on water resources and ecosystems [68]. Agriculture, mining, and urbanization strongly contribute to these transformations [69], affecting hydrological processes and ecosystem services, and increasing environmental vulnerability [70,71,72].

Precisely defining the boundary between the impact of climate change and human activities on water resources is a major challenge in our study [73,74,75]. This concern is also shared by several researchers who focus on studying the combined effects of these two factors [56,76,77,78].

5. Conclusions

This work is based on remote sensing techniques, particularly the use of spectral indices, as a relevant monitoring tool for water resource management, conservation planning, and decision-making support.

The results obtained do not indicate any significant overall trend across all the lakes studied. However, a decrease in lake surface areas has been observed since 2016, highlighted by certain indices in specific cases. Moreover, the complete drying up of four lakes has been noted since 2020. This aspect, however, is not confirmed by the Mann-Kendall trend test, whose results do not always accurately reflect on-the-ground realities. Indeed, the high coefficient of variation for all lakes indicates strong interannual variability in water surfaces, which complicates the interpretation of observed trends.

The analysis of the lake surface evolution curves reveals a period of maximum filling, generally occurring between 2013 and 2016 for the majority of water bodies. It is likely that this exceptional episode influenced both the overall trends extracted from the time series and the identified change points, artificially accentuating certain hydrological dynamics.

Finally, the correlation between wetland surface areas and drought indices is particularly pronounced: wet years correspond to the maximum extent of wetlands, while dry years are characterized by a significant reduction in these areas.

Furthermore, an expansion of irrigated agricultural land has been observed around the lakes, especially after 2013. Indeed, that year marked the fifth year since the launch of Morocco’s largest agricultural development program, the Green Morocco Plan. The increasing pressure from irrigated agriculture, particularly apple orchards, is intensifying water demand, thereby threatening the hydrosystemic balance of wetlands.

These changes in wetland functioning undoubtedly have major environmental and socio-economic repercussions. In the face of these challenges, it is imperative to implement sustainable water resource management strategies, including the restoration of aquatic ecosystems and the promotion of less water-intensive agricultural practices. Raising awareness among local communities about the importance of these ecosystems, as well as fostering cooperation among various stakeholders, is essential to preserve this natural heritage for future generations.