Spatio-Temporal Evolution and Economic Assessment of Sediment and Nutrient Retention Service in Ifrane National Park Based on InVEST Models

2.1. Study Area

The study area encompasses the upper BEHT watershed; the most ecologically and hydrologically significant basin within Ifrane National Park, situated in the western part of the central Middle Atlas region in Morocco (Figure 1).

The INP aims to conserve the region’s unique biodiversity and ecosystems while promoting the sustainable management of natural resources. It also supports local development by fostering environmentally responsible practices and sustainable economic activities, such as ecotourism and agro ecology [21]. The park’s watersheds contribute over 80% of Morocco’s mobilizable water resources, with the upper BEHT watershed representing a particularly vital hydrological basin, which plays a critical role in water regulation and supply, serving as a primary source for downstream reservoirs and agricultural systems. Its significance lies in its capacity to capture precipitation, regulate seasonal flows, and mitigate sediment transport, making it essential for sustaining water availability in the region [21]. Three main soil types within the park based on substrate are: volcanic, calcareous, and dolomitic. The park is characterized by two distinct bioclimates: subhumid and humid. These climates exhibit cool conditions at mid-altitudes, cold temperatures across the plateau, and very cold extremes at the summits of the eastern mountain ranges [21]. The park hosts 22% of Morocco’s vascular plant species, with a high endemism rate of 25%, exemplified by the iconic Atlas cedar (Cedrus atlantica). Additionally, it supports 33% of the country’s mammalian species, including the endangered Barbary macaque (Macaca sylvanus) [22].

2.2. Mapping Historical LULCC

Satellite data for the years 1992, 2002, 2012, and 2022 were analyzed to assess LULCC within the INP over three decades. These datasets, obtained from the Google Earth Engine (GEE) platform, were derived from the Landsat 5, Landsat7, Landsat 8, and Landsat 9 satellite series. The analysis incorporated all available spectral bands, as well as the Normalized Difference Vegetation Index (NDVI), to enable a detailed examination of spatial and temporal LULCC.

Preprocessing steps, including radiometric and atmospheric corrections, were applied to ensure consistency and comparability across the different time periods [23,24]. LULC classification was performed using the Random Forest (RF) machine learning algorithm, selected for its reliability and high accuracy in remote sensing studies [25,26]. RF leverages ensemble learning by aggregating predictions from multiple decision trees, reducing overfitting and increasing model stability. Its inherent feature selection capabilities are particularly useful for identifying significant variables, such as spectral bands or vegetation indices, critical for accurate classification. RF has been shown to perform well in various LU/LC studies, including integration with advanced techniques like Markov chain models and multi-layer perceptrons for change prediction, achieving high accuracy and interpretability [27]. Combining RF with multi-sensor data, such as optical and SAR images, has further enhanced classification accuracy, achieving over 96% accuracy in some studies [28]. While RF generally excels, challenges remain in distinguishing spectrally similar classes and achieving consistent precision and recall measures, especially in heterogeneous landscapes [29]. Despite these limitations, RF’s balance of simplicity, accuracy, and computational efficiency makes it a preferred choice for large-scale LULC classification in GEE.

In this study, supervised classification categorized LULC into six classes: forests, shrubs, crops, built-up areas, water, and bare soil. The selection of study years, encompassing both pre- and post-establishment periods of the park (created in 2004), provided insights into the evolution of LULC before and after the park’s creation. Validation of classification results was conducted using confusion matrices and the Kappa index, demonstrating satisfactory accuracy levels for all time periods analyzed.

2.3. Quantifying Water Regulating Ecosystem Services Using InVEST

After analyzing historical LULCC, we quantify WRES using the InVEST models, in particular the Sediment Delivery Ratio (SDR) and Nutrient Delivery Ratio (NDR) models [17,18,19,20]. Annual soil loss and the sediment delivery ratio, defined as the proportion of soil loss reaching the stream, are computed by the SDR model, and it is assumed that sediment is transported from the source to the stream and then to the watershed outlet. The Revised Universal Soil Loss Equation (RUSLE) is employed by the model to estimate annual soil loss. The spatial movement of nutrient masses is explained by the NDR model through a simple mass balance methodology, in which LULC and loading rates are considered to determine nutrient loads [30]. The input rasters for the InVEST models are presented in Table 1.

2.4. Economic Assessment of Water Regulating Ecosystem Services

In this study, a revealed preference economic valuation method is employed, specifically, the damage costs avoided method, which quantifies expenses that would have been incurred in the absence of a specific environmental function [36,40]. Damages related to the absence of Water Regulating Ecosystem Services are primarily associated with potential degradation affecting agricultural lands and rangelands.

Loss in agricultural yields is obtained using the relationship of Den Biggelaar et al. (2004) [41] (Equation (1)).

r = EwP^1.224 * 0.0114

(1)

with:

• r: Relative decrease in yield due to erosion (%).
• EwP: Erosion rate (t/ha/year).

Then, the relative decrease in yield due to erosion was multiplied by the average crop yield to determine the decline in yield in t/ha (2).

∆R = r * Average crop yield

(2)

with:

• ∆R: Relative decrease in yield due to erosion (t/ha);
• r: Relative decrease in yield due to erosion (%).

Forage yield losses are determined using Table 2, which provides the percentage of forage productivity losses for each erosion class [42].

Once the damage quantities had been determined, their economic costs were estimated. For agricultural damages, the market prices of cereals were applied, as more than 58% of the cultivated land in the study area is dedicated to cereal crops [43]. For rangeland damages, the price of fodder barley was relied upon, since the value of one forage unit is considered equivalent to one kilogram of barley. The evaluation was carried out for multiple reference years (1992, 2002, 2012, and 2022). Accordingly, the market prices of cereals and barley corresponding to each year were used to ensure accurate cost estimation [44].

3. Results

3.1. Historical LULCC Within the INP Watershed

The analysis revealed a significant increase in forest cover between 2012 and 2022, the decade following the park establishment (Figure 2). Prior to the park’s creation, the region experienced a gradual decline in forested areas due to various anthropogenic activities, including deforestation and land conversion for agricultural purposes. However, the implementation of conservation measures and sustainable management practices within the park has led to a remarkable recovery and expansion of the forest ecosystem.

An analysis of conservation initiatives implemented within the park reveals significant achievements. During the period 2012/2022, more than 2,000 hectares were successfully reforested, and over 4,000 hectares were placed under protection or designated for restoration and reforestation. These areas were managed through a participatory approach involving local communities. The number of local associations engaged in park management increased from 5 at the park’s inception to 12 at present, reflecting a growing community commitment. Furthermore, there was a nearly 50% reduction in forest-related offenses during this time. These positive indicators align with the findings from remote sensing analysis, which demonstrated an increase in forest cover, thereby confirming the tangible impact of these conservation efforts [45].

Table 3 illustrates the evolution of each LULC class within the INP watershed.

For the different years analyzed, we achieved a Kappa coefficient value equal to or greater than 90% (Table 4), indicating a near-perfect agreement between the reference classification and the automated classification performed using Google Earth Engine. This high level of concordance signifies that the model accurately replicated the reference data.

The classification accuracy results for the years 1992, 2002, 2012, and 2022 show a clear trend of improvement in the model’s performance over time, as indicated by the increasing overall accuracy and the decreasing disagreement metrics. The overall accuracy increased progressively, starting from 87.5% in 1992 and reaching an impressive 98.96% in 2022. This steady improvement suggests significant advancements in the classification model’s ability to align with reference data. The high accuracy achieved in 2022 highlights the impact of improved input data quality and enhanced feature extraction or modeling techniques over the study period.

The quantity disagreement, which reflects errors related to mismatches in class proportions, exhibited an initial increase from 2.78% in 1992 to 6.45% in 2002. This could be attributed to limitations in the availability or representativeness of the training data during that period. However, this metric showed a consistent decline thereafter, reducing to 4.26% in 2012 and further to 1.04% in 2022. The observed reduction in quantity disagreement over time indicates that the classification model has progressively become better at predicting the correct number of pixels for each class, aligning more closely with the reference data distributions.

Similarly, allocation disagreement, which measures spatial misallocations of classes, was notably high in 1992 (9.72%) and 2002 (11.29%), indicating significant challenges in correctly assigning class locations during the earlier years. However, this metric saw a dramatic reduction to 2.13% in 2012 and was completely eliminated in 2022 (0.00%). The absence of allocation disagreement in 2022 demonstrates a perfect spatial alignment between the predicted and reference data, suggesting substantial advancements in the classification model’s spatial accuracy. These improvements may be attributed to better-distributed training data, higher-resolution satellite imagery, or the adoption of more robust classification algorithms.

Overall, the results indicate that the classification process has undergone considerable refinement over the study period. The observed improvements in both accuracy and disagreement metrics emphasize the critical role of evolving data quality and modeling techniques in achieving reliable LULC mapping. These trends highlight the importance of continuous innovation in remote sensing methodologies to ensure accurate and spatially coherent classification outcomes.

3.2. Water Regulating Ecosystem Services Evolution

Our results indicate a notable reduction in soil erosion within the INP watershed following the establishment of the park (2012/2022) (Figure 3). This decline can be attributed to the significant expansion of forest cover, which has enhanced soil stability and reduced erosion vulnerability. These findings underscore the critical role of forest restoration and protection within protected areas in mitigating soil degradation.

However, the observed reduction in erosion could have been more substantial if not for the simultaneous expansion of agricultural activities. The increase in agricultural lands has heightened soil sensitivity to erosion, partially offsetting the benefits provided by forest cover recovery. This highlights the importance of integrated land management strategies that balance agricultural development with ecosystem conservation to maximize the benefits of soil erosion control.

Figure 4 illustrates the spatial evolution of the biophysical quantity of sediment loss within the INP watershed, highlighting variations across the study period.

Our results also indicate a significant reduction in nutrient export, specifically nitrogen and phosphorus, within the INP watershed following the establishment of the park (2012/2022) (Figure 5). This decline can similarly be attributed to the expansion of forest cover, which demonstrates a higher efficiency in nutrient retention compared to other LULC types. These findings further highlight the pivotal role of forest restoration and protection in enhancing ecosystem services.

However, this reduction could have been significantly greater if not for the concurrent expansion of intensive agricultural activities and the substantial decrease in grasslands. These grasslands, which play a vital role in nutrient retention and erosion control, have either been degraded into bare soil due to overgrazing or converted into agricultural fields. This shift in land use has not only compromised the ecological benefits provided by grasslands but has also amplified nutrient runoff. Intensive agricultural practices often involve excessive use of fertilizers, which, combined with the loss of natural vegetation, exacerbate nutrient leaching into downstream systems. As a result, while reforestation efforts have improved certain ecosystem services, the negative impacts of agricultural expansion have partially offset these gains.

This duality highlights the critical need for integrated and sustainable land management strategies that balance forest conservation with the growing demands for agricultural development. Policies should prioritize sustainable practices such as agroforestry, soil conservation techniques, and the restoration of degraded grasslands to mitigate the environmental impacts of agriculture. Encouraging local communities to adopt such practices through education and incentives can play a key role in maintaining the functionality of nutrient retention services.

Figure 6 and Figure 7 presents the spatial evolution of the biophysical quantity of Nutrients export within the INP watershed. Highlighting variations from 1992 to 2022.

3.3. Economic Assessment of Water Regulating Ecosystem Services

The results show also an increase in agricultural yield losses cost (Table 5), driven by the expansion of agricultural areas, even after the park’s establishment, leading to greater erosion-related damages. Conversely, forage yield losses cost has declined (Table 6), primarily due to a significant reduction in grazing land, which has been converted to agriculture or built-up areas. These findings highlight that, despite efforts to restore forest ecosystems and enhance ecosystem services, agricultural expansion continues to undermine these gains, resulting in a low economic value, which would have been much higher. This underscores the need for interventions to modify agricultural practices and promote sustainable land use for grazing areas to improve environmental outcomes.

The economic value of the WRES is determined based on the potential on-site losses in agricultural and forage yields that would occur in the absence of this ecosystem service. The results show that the WRES value was negative before the park’s establishment. However, after the park’s creation (in 2004), the WRES value became positive, estimated at 10,000 USD/Year (Figure 8).

4. Discussion

The results of the spatio-temporal and economic assessment of the WRES in Ifrane National Park highlight significant trends at both spatial and economic levels. Prior to the establishment of the park in 2004, the WRES value was negative, estimated at -7,000 USD/year. This negative value was primarily attributed to agricultural expansion, which led to increased sediment and nutrient losses, negatively impacting both the environment and agricultural yields.

However, following the creation of the park, the WRES value became positive, reaching 10,000 USD/year. This positive shift can be attributed to the expansion of forest cover, which enhanced sediment and nutrient retention, thereby reducing erosion-related losses and contributing positively to the ecosystem. The improvement in soil quality and reduction in erosion were particularly notable in areas where forest cover was restored. Despite this positive trend, the WRES value could have been significantly higher had it not been for the continued agricultural expansion and land conversion to built-up areas after the park’s establishment.

The conversion of grazing lands into agricultural lands and built-up zones further negatively impacted nutrient retention. While agricultural yields increased, this growth was accompanied by a heightened vulnerability of soils to erosion, reducing the park’s ability to provide optimal sediment and nutrient retention services. The decrease in grazing land area, primarily converted into agricultural land considered more profitable, led to a loss of the land’s retention capacity [46,47,48,49,50,51].

These findings underscore the importance of integrated land management strategies that balance ecosystem protection with agricultural development. Despite efforts to restore forest ecosystems, the continued expansion of agriculture and urbanization continues to mitigate these environmental gains. The results suggest the need for sustainable agricultural practices and the preservation of grazing lands to optimize ecosystem services such as sediment and nutrient retention. Sustainable land management strategies should be implemented to maximize nutrient retention and limit negative environmental impacts while supporting local economic activities [52,53,54,55,56,57].

It is crucial for park managers to invest in advanced tools focused on the continuous mapping and assessment of the evolution of the park ecosystem services. This approach would support data-driven decision-making and more adaptive management strategies to address ecosystem changes effectively [58].

The findings from this study can serve as a powerful tool for advocating sustained conservation efforts among policymakers and relevant stakeholders. By showcasing the dynamics of ecosystem services, these insights can drive support for the implementation of more comprehensive and long-term conservation policies [59].

Awareness campaigns highlighting the value of ecosystem services should be prioritized to encourage land-use practices that align with the park’s conservation goals and enhance its ecological resilience [60].

The results of this study should be systematically integrated into the upcoming revisions of the Park Management and Development Plan (PAG). Considering the trends in ecosystem service evolution will inform more effective conservation strategies and contribute to the sustainable management of park resources [61].