Vegetation cover plays a crucial role in maintaining ecosystem services, such as biodiversity preservation, water regulation, and carbon sequestration (Yapp et al., 2010). Changes in vegetation cover, driven by climatic events or human activities, can significantly alter these ecosystem functions and directly impact local communities (Dale, 1997; Aber et al., 2001). In arid and semi-arid regions like the Piura River Basin in northern Peru, changes in vegetation cover are especially critical due to the region’s vulnerability to both environmental stressors and socioeconomic pressures. Understanding these changes is essential for effective landscape management and conservation efforts (Barboza et al., 2024; Otivo Meza, 2015).

The sixth report of the Intergovernmental Panel for Climate Change (IPCC) highlighted the increasing frequency and intensity of extreme climatic events, such as droughts, floods, and heat waves, as a result of climate change (IPCC, 2021). These events continue to have substantial impacts in terms of hazard exposure, vulnerability, and risk management challenges, despite recent advances in climate adaptation strategies (Kreibich et al., 2022). In South America, the El Niño Southern Oscillation (ENSO) is a key driver of extreme events, bringing heavy rainfall to northwestern regions and causing floods in coastal areas and droughts in the Andes and the Amazon (Cai et al., 2020).

The Piura River Basin, located in northern Peru, is particularly affected by ENSO events. Previous events, such as the 1997–1998 El Niño, caused widespread flooding, disrupting infrastructure and local livelihoods, particularly in agriculture (Atarama and Rashid, 2019; Rubiños and Anderies, 2020). Agriculture, the main economic activity in the Piura Basin, mostly dependent on irrigation (Mills-Novoa, 2020), is highly sensitive to changes in temperature and precipitation (CONAM, 2006; Torres Ruiz de Castilla, 2010). The region’s vulnerability to climatic events (Rau et al., 2017), combined with the influence of human activities, makes it essential to study long-term trends in vegetation cover and their relationship with hydrometeorological variables.

Studies have extensively examined the impacts of ENSO and land-use changes on vegetation dynamics in Peru, particularly in tropical dry forest ecosystems (e.g., Móstiga et al., 2024a; Muenchow et al., 2013). However, the Piura River Basin remains underexplored in terms of how these dynamics vary across elevational gradients. Land cover plays a fundamental role in regulating multiple ecological processes, such as the hydrological cycle and the radiative balance (Song et al., 2018). As both a cause and consequence of environmental changes, land cover can influence climatic variability and hydrological processes, and these interactions are particularly relevant in arid regions like Piura (Eltahir and Bras, 1996; Rau et al., 2019; Rau et al., 2018). Nonetheless, there is still limited understanding of how these interactions vary across highland and lowland areas in the basin (Otivo Meza, 2015; Salazar et al., 2018).

This study aims to investigate the spatial and temporal dynamics of vegetation cover change in the Piura River Basin over the past 2 decades, focusing on comparing highland and lowland areas. This research will provide insights into the patterns of land cover change and their correlation with both climatic drivers, such as ENSO (Bourrel et al., 2015; Córdova et al., 2023), and human activities (Barboza et al., 2024; Yglesias-González et al., 2023). The findings will contribute to the broader understanding of how arid and semi-arid ecosystems in Peru respond to natural and anthropogenic pressures.

The Piura River Basin is located in the northwestern part of Peru, between latitudes 4.71°S and 6.01°S and longitudes 79.45°W and 81.03°W, with a maximum elevation of 3470 m.a.s.l (see Figure 1). The climate of the Piura Basin is characterized by a well-marked seasonality, with a wet season from December to May and a dry season from June to November. During the period from 1981 to 2020, the average monthly temperature in the basin ranged between 19.3°C and 31.2°C, with an annual mean precipitation of approximately 24 mm/month. The hydrology of the Piura Basin is influenced by various water bodies, including rivers, lagoons, and irrigation cana ls that are part of the region’s hydraulic infrastructure. The diversion dam and associated channels enable water management for agricultural purposes, which is the main economic activity in the basin. Between February and April, the average streamflow peaks at Puente Sánchez Cerro hydrological station due to heavy rainfall typical of the wet season. In contrast, minimum streamflow levels are observed between September and November, reflecting the natural cycle of water availability in the basin (see Figure 2).

The upper part of the Piura basin is located at higher altitudes and is predominantly characterized by mountain ecosystems, including relict montane forests and páramo. In contrast, the middle and lower parts, which are closer to the coast, feature seasonally dry forests and agricultural areas, representative of the arid and semi-arid ecosystems of northern Peru, also have a diverse range of vegetation types, including dry hill and mountain forests, riparian forests, and even mangrove areas near the coast (MINAM, 2019). (see Supplementary Figure S1). The montane forests are critical in stabilizing soils and reducing runoff, whereas dry forests in the lower-middle basin are highly susceptible to desertification, which is exacerbated by El Niño-related climatic anomalies.

In the upper Piura Basin, agricultural production is diverse, mostly consisting of rice (3,819 ha), corn (2,721 ha), beans (1,168 ha) for local and regional markets, but also mango (3,352 ha), banana (1,399 ha), cacao (811 ha) for export and currently increasing (Autoridad Nacional del Agua (ANA), 2014). It relies on traditional irrigation practices applied by small-holder farmers (Mills-Novoa, 2020). In the upstream part of the middle Piura Basin, the San Lorenzo district benefits from its own reservoir built by the Peruvian government in 1950s with funds from the World Bank in the framework of the Peruvian Agrarian Reform (Lynch, 2019). It allows the production of mangos (16,634 ha) and limes (5,176 ha) destined to export by small-holder farmers (Autoridad Nacional del Agua (ANA), 2014; Lynch, 2019). The middle and lower parts of the Piura Basin are experiencing large agricultural expansion (Mills-Novoa, 2020). Irrigated land increased from 19,140 ha to 28,210 ha (47%) between 2006 and 2013 (Junta de Usuarios: Medio-Bajo Piura, 2011; Junta de Usuarios: Medio-Bajo Piura, 2014), using both surface and poorly regulated groundwater resources, driven by Peruvian and international agro-business investments in large-scale table grape vineyards (Autoridad Nacional de Agua, 2012; Autoridad Nacional del Agua (ANA), 2014; Autoridad Nacional del Agua (ANA), 2015; Autoridad Nacional del Agua (ANA), 2016). In parallel, agricultural productions from small-holder farmers also shifted toward banana, lime, and mango destined to export (Mills-Novoa, 2020). The lower Piura shifted from its historically cotton production to small-scale rice production (8,011 ha) in rotation with corn (5,703 ha) and beans (1,574 ha) for the domestic market (Autoridad Nacional del Agua (ANA), 2014). This area is characterized by extensive drainage and irrigation infrastructures needed for historical cotton production and nowadays intensive land use with two harvests per year (Mills-Novoa, 2020).

The Piura River Basin also experiences strong socioeconomic pressures due to the expansion of agricultural and urban activities, which have led to significant transformations in vegetation cover over recent decades. These pressures, combined with the influence of climatic phenomena such as El Niño, contribute to soil degradation and vegetation loss, especially in vulnerable areas like the seasonally dry forests (Móstiga et al., 2024a; Muenchow et al., 2013). Studies have shown that the expansion of agricultural lands and urbanization are major drivers of deforestation and land cover change in the region, leading to the fragmentation of habitats, decline in biodiversity and increasing vulnerability to flooding (Rau et al., 2022; Rubiños and Anderies, 2020; Torres Ruiz de Castilla, 2010; Móstiga et al., 2024b). Changes in vegetation cover in this region reflect how ecosystems respond to both natural disturbances and human activities, highlighting the need for integrated land management to ensure the conservation of natural resources and the resilience of local ecosystems (Otivo Meza, 2015; Song et al., 2018).

The results highlight substantial differences between the MODIS and ESA-CCI datasets, primarily due to their classification methodologies, temporal resolutions, and detection sensitivities. MODIS exhibits greater sensitivity to interannual spectral variations, capturing more frequent land cover transitions, ranging from actual thematic class changes to more likely conditional changes in land cover that fall short of full conversion. In contrast, ESA-CCI provides a more stable representation of longer-term land cover transformations. These discrepancies align with previous studies emphasizing the impact of classification algorithms and temporal resolution in land cover mapping (Song et al., 2018; Friedl and Sulla-Menashe, 2022). While MODIS detects a higher number of land cover changes, these transitions may include both thematic class shifts and spectral condition variations (Wulder et al., 2018). ESA-CCI, by applying a 2-year temporal filter, prioritizes persistent land cover changes, reducing short-term fluctuations. This distinction is particularly relevant when interpreting land cover change percentages, as MODIS may overestimate short-term changes that do not necessarily reflect persistent transitions (Hansen et al., 2013).

The results also reveal clear differences in the stability of land cover changes detected by MODIS and ESA-CCI. MODIS captures more dynamic transitions, whereas ESA-CCI presents a more stable representation of land cover trends. This is evident in the magnitude of changes observed, where 67.2% of the basin remained unchanged in MODIS, compared to 93.2% stability in ESA-CCI. Additionally, MODIS identified more frequent transitions, with 29.2% of the basin undergoing up to five changes, versus only 6.8% in ESA-CCI. Notably, only MODIS detected areas with more than five transitions (3.6% of the basin), concentrated in the upper watershed near regions with complex topography. These differences highlight the impact of classification methodology on change detection, where MODIS, with its annual classification, is more sensitive to transient changes, capturing short-term fluctuations that likely exceed typical seasonal variability, such as atypical agricultural shifts or unexpected vegetation loss events. In contrast, ESA-CCI applies a temporal smoothing approach, filtering out short-term changes and emphasizing longer-term trends.

The quality of the classification approaches can also be affected by the cloud cover conditions over the study area. Northern Peru is characterized by the presence of mesoscale convective systems responsible for a dense cloud cover (Horel and Cornejo-Garrido, 1986; Goldberg et al., 1987; Jaramillo et al., 2017), especially during the rainy season from January to April (Figure 2A). As both MODIS and ESA CCI LULC are derived from multispectral images, the estimate of land cover classes is strongly affected by the unavailability of valid data during a large part of the year, increasing the risk of misclassification.

Importantly, the intensity analysis (Aldwaik and Pontius, 2012) helps discern between systematic and potentially permanent changes versus transient fluctuations or thematic classification errors. This analytical approach effectively distinguishes real thematic transitions from temporary variations in land cover conditions, particularly relevant in semi-arid regions like the Piura River Basin, where climatic variability and seasonal cycles can strongly influence vegetation dynamics.

The spatial heterogeneity of the Piura River Basin also plays a key role in land cover dynamics. The upper basin, characterized by steeper slopes, lower temperatures, and more stable vegetation cover, exhibits fewer detected changes, particularly in forested and non-forested vegetation categories. In contrast, the lower basin, where cropland expansion and urban growth are more prominent, shows greater variability, especially in cropland and bare soil classes. These patterns align with the distinct precipitation regimes across the basin, where the upper basin receives orographic rainfall, whereas the lower basin is more sensitive to ENSO-driven precipitation variability (Bourrel et al., 2015).

The relationship between ENSO events and land cover change is particularly relevant given the increased frequency of warm events after 2014 (ENFEN, 2024). The division of the study period into 2001–2011 and 2011–2022 effectively captures this shift, with the second period including intense El Niño events, such as the 2015–2016 strong El Niño and the 2017 moderate El Niño. During the 2015–2016 event, there was a notable expansion of bare soil, particularly in the lower basin, where prolonged dry conditions and extreme hydroclimatic stress likely led to vegetation loss and soil degradation (Cai et al., 2020; Rodríguez-Morata et al., 2018). Similarly, post-2017 recovery phases correspond to increases in non-forest vegetation, indicating temporary regrowth responses after extreme precipitation events. Conversely, strong La Niña events, such as those in 2007, 2013, and 2021–2022, appear to have contributed to reductions in cropland extent and an increase in bare soil, reflecting drier conditions and potential land abandonment (Cai et al., 2020; Salazar Zarzosa, 2018). These findings align with previous studies highlighting ENSO’s role in shaping land cover dynamics in northern Peru (Takahashi and Martínez, 2019).

The comparison between MODIS, ESA-CCI, and the National Ecosystem Map of Peru reveals systematic differences in the LULC classes. MODIS’s urban and non-forest vegetation classifications align more closely with the National Ecosystem Map, while ESA-CCI’s cropland classification better matches agricultural zones. Yet, only CCI exhibits a slight increase in urban areas in both the upper (0.53%) and mid-lower (0.14%) Piura Basin (Figure 8) which can reflect the demographic growth in urban areas, with an increase of the people in Piura city of 2.3% between 2013 and 2017 (Zuchetti and Freundt, 2018). The increase in urban areas might have been missed in MODIS LULC due to its coarser spatial resolution as Piura city, the largest city in the Piura Basin has 70% of its urbanized land occupied by informal or spontaneous constructions (Rivera Saavedra, 2016), which might be fragmented. However, the most notable discrepancy is in the extent of bare soil, significantly larger in MODIS and ESA-CCI compared to the National Ecosystem Map. This difference likely results from spatial resolution variations, as the 30-m resolution of the National Ecosystem Map enables finer-scale classification, whereas the coarser 300-m and 500-m resolutions of ESA-CCI and MODIS, respectively, may lead to an overgeneralization of barren areas (Hansen et al., 2013). Additionally, classification methodologies differ, with MODIS using smoothed time-series reflectance and ESA-CCI applying a probability-based classification with a 2-year temporal filter, both of which can overestimate persistent barren landscapes (Giri et al., 2013). Thematic classification discrepancies also contribute to these differences, as the coastal desert category in the National Ecosystem Map includes sparse vegetation patches, which MODIS and ESA-CCI classify as bare soil, leading to apparent barren area overestimation. These findings emphasize the importance of integrating high-resolution regional datasets for land cover validation, reinforcing previous studies highlighting the limitations of coarse-resolution global land cover products in heterogeneous landscapes (Song et al., 2018). Nevertheless, they provide consistent annual changes for cropland (rapid decrease) in the upper Piura Basin (Figure 5) and for forest (slow increase), non forest vegetation and cropland (rapid decrease) in the middle-lower Piura Basin (Figure 6).

Future research should consider combining these products with higher-resolution sensors such as Sentinel-1 and 2 to improve cropland expansion and deforestation detection. Additionally, expanding in-situ validation efforts would most likely enhance classification accuracy, particularly in highly heterogeneous landscapes like the Piura River Basin. Integrating multiple datasets would enable a more comprehensive understanding of land cover change, capturing short-term fluctuations and long-term transformations. Further research should also explore the incorporation of climate indices, such as the Coastal El Niño Index (ICEN), into land cover change models to quantify ENSO’s direct influence on vegetation dynamics.

This study analyzed land cover changes in the Piura River Basin from 2001 to 2022 using MODIS, ESA-CCI, and the National Ecosystem Map of Peru, harmonizing their classifications into six macroclasses to enable a standardized comparison. The results highlight significant differences between the global satellite-based products, which are primarily attributed to their classification methodologies, temporal resolution, and detection sensitivity.

MODIS exhibits greater sensitivity to interannual changes, capturing more frequent land cover transitions, which may also include conditional changes within a specific land cover class. ESA-CCI presents a more stable representation of longer-term land cover changes due to its probability-based classification and 2-year temporal filter. These differences led to notable discrepancies in certain categories, particularly in the extent of Bare Soil, which is more extensive in MODIS and ESA-CCI compared to the National Ecosystem Map. The observed inconsistencies underscore the importance of integrating high-resolution regional datasets to refine global land cover assessments.

The division of the study period into 2001–2011 and 2011–2022 reveals an increase in land cover dynamics during the second period, which coincides with a higher frequency of El Niño events after 2014. These climatic fluctuations influenced the expansion of Bare Soil and the reduction of Cropland during strong warm phases, particularly in the lower basin, highlighting the sensitivity of land use dynamics to ENSO-driven hydroclimatic variability. The distinct precipitation regimes and elevatio nal gradients across the upper and lower basins further contribute to spatial heterogeneity in land cover transitions.

The findings of this study emphasize the need for multi-source data integration in land cover change analysis. While MODIS is valuable for detecting short-term variability, ESA-CCI provides a long-term, stable baseline for monitoring gradual transformations. Future research should combine these products with higher-resolution sensors such as Sentinel-3, enabling more precise assessments of cropland expansion, deforestation, and ecosystem degradation. Additionally, integrating climate indices like the Coastal El Niño Index (ICEN) could improve understanding of ENSO-driven land cover changes. Strengthening in-situ validation efforts would also enhance classification accuracy, ensuring more reliable assessments for land management and conservation planning in climate-sensitive regions like the Piura River Basin.