Assessing the Impact of Spatial Resolution and Aggregation Method on Sentinel-2 NDVI Time Series in Grasslands of Mainland Spain.
Saved in:
| Title: | Assessing the Impact of Spatial Resolution and Aggregation Method on Sentinel-2 NDVI Time Series in Grasslands of Mainland Spain. |
|---|---|
| Authors: | Pugni-Stanek, Tomás1,2,3 (AUTHOR), Merino-de-Miguel, Silvia2,3 (AUTHOR), Recuero, Laura3,4 (AUTHOR), Magruga-Ramos, Diego1,4 (AUTHOR), Litago, Javier1,4 (AUTHOR), Palacios-Orueta, Alicia1,2 (AUTHOR) alicia.palacios@upm.es |
| Source: | Remote Sensing. Aug2026, Vol. 18 Issue 15, p2611. 26p. |
| Subjects: | Spatial resolution, Normalized difference vegetation index, Remote sensing, Grasslands, Time series analysis, Ecological impact, Phenology |
| Geographic Terms: | Spain, Southern Europe |
| Abstract: | Highlights: What are the main findings? Sentinel-2 at 10 m and 20 m preserves NDVI temporal consistency in grasslands, while 60 m causes substantial degradation and excludes over half of the plots under pure-pixel sampling. Formal Kruskal–Wallis testing and Cliff's Delta effect-size analysis confirm that spatial resolution—not the aggregation method—is the dominant factor governing time series fidelity. Parcel area and Köppen climate class are key modulators: the smallest plots (<3 ha) exhibit large effect sizes ( ε 2 = 0.273 ), while climate decisively shapes phenological trajectories (up to ε 2 = 0.447 ), with spatial degradation distorting the temporal signal unequally across climate classes. The pixel-selection strategy (pure-pixel vs. centroid) has minimal impact at 10 m and 20 m but becomes relevant at 60 m, where pure-pixel sampling slightly improves NDVI reliability. What are the implications of the main findings? A multi-scale Sentinel-2 strategy is supported: 10 m resolution for fragmented, heterogeneous grasslands (<3 ha) typical of humid regions, and 20 m as a computationally efficient alternative (78% storage reduction, 72% faster processing) for large, homogeneous plots (>10 ha) in Mediterranean and semi-arid zones. These findings outline potential implications for policy frameworks such as the Common Agricultural Policy (CAP), suggesting that 20 m products could provide a viable monitoring compromise in extensive grazing systems, while 10 m data may remain necessary for small-scale intensive pastures. High-resolution satellite imagery has substantially improved the monitoring of vegetation dynamics; however, the influence of spatial resolution and pixel aggregation on NDVI time series consistency remains insufficiently quantified, particularly across multiple native resolutions within a single sensor platform. This study evaluates how Sentinel-2 spatial resolutions (10 m, 20 m, and 60 m) and two pixel aggregation methods (pure-pixel and centroid) affect NDVI time series in 14,031 grassland plots across mainland Spain over the period 2018–2023. High-quality NDVI time series were selected using the Interpolation Efficiency Indicator (IEI), and discrepancies relative to a 10 m pure-pixel baseline were quantified through the Time Series Angle Distance (TSAD) and Root Mean Square Error (RMSE). A sensitivity check confirmed that the radiometric differences between Band 8 (10 m) and Band 8A (20/60 m) introduce negligible bias compared with genuine spatial-resolution effects. Formal non-parametric statistical testing—omnibus Kruskal–Wallis with epsilon-squared ( ε 2 ) effect sizes and pairwise Cliff's Delta comparisons—was applied to assess the magnitude and practical significance of the observed differences across plot area categories and Köppen climate groups (B, Cs, Cf). Results show that coarser resolutions (60 m) substantially reduce NDVI reliability, excluding more than half of the plots under the pure-pixel criterion and smoothing temporal variability, whereas 10 m and 20 m resolutions preserve most spectral and temporal information. The 20 m resolution introduces moderate but non-severe phenological distortion (median TSAD ≈ 0.05 rad, RMSE ≈ 0.026) with a 78% reduction in data volume and 72% reduction in processing time. The choice between pure-pixel and centroid sampling has negligible impact at 10–20 m but becomes relevant at 60 m, where pure-pixel selection reduces errors from spectral mixing at the cost of severe sample attrition. Parcel area strongly conditions the error metrics, with large effect sizes ( ε 2 = 0.273 ) in the smallest plots, while Köppen climate classification decisively shapes TSAD (up to ε 2 = 0.447 ), indicating that spatial degradation distorts phenological patterns differently across climate classes. These findings support a multi-scale monitoring strategy: 10 m for fragmented, heterogeneous grasslands (<3 ha), 20 m as a computationally efficient alternative for homogeneous areas (>10 ha), and outline potential implications for policy frameworks such as the Common Agricultural Policy (CAP). [ABSTRACT FROM AUTHOR] |
| Copyright of Remote Sensing is the property of MDPI and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
| FullText | Links: – Type: pdflink Text: Availability: 1 |
|---|---|
| Header | DbId: egs DbLabel: Engineering Source An: 196272291 AccessLevel: 6 PubType: Academic Journal PubTypeId: academicJournal PreciseRelevancyScore: 0 |
| IllustrationInfo | |
| Items | – Name: Title Label: Title Group: Ti Data: Assessing the Impact of Spatial Resolution and Aggregation Method on Sentinel-2 NDVI Time Series in Grasslands of Mainland Spain. – Name: Author Label: Authors Group: Au Data: <searchLink fieldCode="AR" term="%22Pugni-Stanek%2C+Tomás%22">Pugni-Stanek, Tomás</searchLink><relatesTo>1,2,3</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Merino-de-Miguel%2C+Silvia%22">Merino-de-Miguel, Silvia</searchLink><relatesTo>2,3</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Recuero%2C+Laura%22">Recuero, Laura</searchLink><relatesTo>3,4</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Magruga-Ramos%2C+Diego%22">Magruga-Ramos, Diego</searchLink><relatesTo>1,4</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Litago%2C+Javier%22">Litago, Javier</searchLink><relatesTo>1,4</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Palacios-Orueta%2C+Alicia%22">Palacios-Orueta, Alicia</searchLink><relatesTo>1,2</relatesTo> (AUTHOR)<i> alicia.palacios@upm.es</i> – Name: TitleSource Label: Source Group: Src Data: <searchLink fieldCode="JN" term="%22Remote+Sensing%22">Remote Sensing</searchLink>. Aug2026, Vol. 18 Issue 15, p2611. 26p. – Name: Subject Label: Subjects Group: Su Data: <searchLink fieldCode="DE" term="%22Spatial+resolution%22">Spatial resolution</searchLink><br /><searchLink fieldCode="DE" term="%22Normalized+difference+vegetation+index%22">Normalized difference vegetation index</searchLink><br /><searchLink fieldCode="DE" term="%22Remote+sensing%22">Remote sensing</searchLink><br /><searchLink fieldCode="DE" term="%22Grasslands%22">Grasslands</searchLink><br /><searchLink fieldCode="DE" term="%22Time+series+analysis%22">Time series analysis</searchLink><br /><searchLink fieldCode="DE" term="%22Ecological+impact%22">Ecological impact</searchLink><br /><searchLink fieldCode="DE" term="%22Phenology%22">Phenology</searchLink> – Name: SubjectGeographic Label: Geographic Terms Group: Su Data: <searchLink fieldCode="DE" term="%22Spain%22">Spain</searchLink><br /><searchLink fieldCode="DE" term="%22Southern+Europe%22">Southern Europe</searchLink> – Name: Abstract Label: Abstract Group: Ab Data: Highlights: What are the main findings? Sentinel-2 at 10 m and 20 m preserves NDVI temporal consistency in grasslands, while 60 m causes substantial degradation and excludes over half of the plots under pure-pixel sampling. Formal Kruskal–Wallis testing and Cliff's Delta effect-size analysis confirm that spatial resolution—not the aggregation method—is the dominant factor governing time series fidelity. Parcel area and Köppen climate class are key modulators: the smallest plots (<3 ha) exhibit large effect sizes ( ε 2 = 0.273 ), while climate decisively shapes phenological trajectories (up to ε 2 = 0.447 ), with spatial degradation distorting the temporal signal unequally across climate classes. The pixel-selection strategy (pure-pixel vs. centroid) has minimal impact at 10 m and 20 m but becomes relevant at 60 m, where pure-pixel sampling slightly improves NDVI reliability. What are the implications of the main findings? A multi-scale Sentinel-2 strategy is supported: 10 m resolution for fragmented, heterogeneous grasslands (<3 ha) typical of humid regions, and 20 m as a computationally efficient alternative (78% storage reduction, 72% faster processing) for large, homogeneous plots (>10 ha) in Mediterranean and semi-arid zones. These findings outline potential implications for policy frameworks such as the Common Agricultural Policy (CAP), suggesting that 20 m products could provide a viable monitoring compromise in extensive grazing systems, while 10 m data may remain necessary for small-scale intensive pastures. High-resolution satellite imagery has substantially improved the monitoring of vegetation dynamics; however, the influence of spatial resolution and pixel aggregation on NDVI time series consistency remains insufficiently quantified, particularly across multiple native resolutions within a single sensor platform. This study evaluates how Sentinel-2 spatial resolutions (10 m, 20 m, and 60 m) and two pixel aggregation methods (pure-pixel and centroid) affect NDVI time series in 14,031 grassland plots across mainland Spain over the period 2018–2023. High-quality NDVI time series were selected using the Interpolation Efficiency Indicator (IEI), and discrepancies relative to a 10 m pure-pixel baseline were quantified through the Time Series Angle Distance (TSAD) and Root Mean Square Error (RMSE). A sensitivity check confirmed that the radiometric differences between Band 8 (10 m) and Band 8A (20/60 m) introduce negligible bias compared with genuine spatial-resolution effects. Formal non-parametric statistical testing—omnibus Kruskal–Wallis with epsilon-squared ( ε 2 ) effect sizes and pairwise Cliff's Delta comparisons—was applied to assess the magnitude and practical significance of the observed differences across plot area categories and Köppen climate groups (B, Cs, Cf). Results show that coarser resolutions (60 m) substantially reduce NDVI reliability, excluding more than half of the plots under the pure-pixel criterion and smoothing temporal variability, whereas 10 m and 20 m resolutions preserve most spectral and temporal information. The 20 m resolution introduces moderate but non-severe phenological distortion (median TSAD ≈ 0.05 rad, RMSE ≈ 0.026) with a 78% reduction in data volume and 72% reduction in processing time. The choice between pure-pixel and centroid sampling has negligible impact at 10–20 m but becomes relevant at 60 m, where pure-pixel selection reduces errors from spectral mixing at the cost of severe sample attrition. Parcel area strongly conditions the error metrics, with large effect sizes ( ε 2 = 0.273 ) in the smallest plots, while Köppen climate classification decisively shapes TSAD (up to ε 2 = 0.447 ), indicating that spatial degradation distorts phenological patterns differently across climate classes. These findings support a multi-scale monitoring strategy: 10 m for fragmented, heterogeneous grasslands (<3 ha), 20 m as a computationally efficient alternative for homogeneous areas (>10 ha), and outline potential implications for policy frameworks such as the Common Agricultural Policy (CAP). [ABSTRACT FROM AUTHOR] – Name: AbstractSuppliedCopyright Label: Group: Ab Data: <i>Copyright of Remote Sensing is the property of MDPI and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract.</i> (Copyright applies to all Abstracts.) |
| PLink | https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=egs&AN=196272291 |
| RecordInfo | BibRecord: BibEntity: Identifiers: – Type: doi Value: 10.3390/rs18152611 Languages: – Code: eng Text: English PhysicalDescription: Pagination: PageCount: 26 StartPage: 2611 Subjects: – SubjectFull: Spatial resolution Type: general – SubjectFull: Normalized difference vegetation index Type: general – SubjectFull: Remote sensing Type: general – SubjectFull: Grasslands Type: general – SubjectFull: Time series analysis Type: general – SubjectFull: Ecological impact Type: general – SubjectFull: Phenology Type: general – SubjectFull: Spain Type: general – SubjectFull: Southern Europe Type: general Titles: – TitleFull: Assessing the Impact of Spatial Resolution and Aggregation Method on Sentinel-2 NDVI Time Series in Grasslands of Mainland Spain. Type: main BibRelationships: HasContributorRelationships: – PersonEntity: Name: NameFull: Pugni-Stanek, Tomás – PersonEntity: Name: NameFull: Merino-de-Miguel, Silvia – PersonEntity: Name: NameFull: Recuero, Laura – PersonEntity: Name: NameFull: Magruga-Ramos, Diego – PersonEntity: Name: NameFull: Litago, Javier – PersonEntity: Name: NameFull: Palacios-Orueta, Alicia IsPartOfRelationships: – BibEntity: Dates: – D: 01 M: 08 Text: Aug2026 Type: published Y: 2026 Identifiers: – Type: issn-print Value: 20724292 Numbering: – Type: volume Value: 18 – Type: issue Value: 15 Titles: – TitleFull: Remote Sensing Type: main |
| ResultId | 1 |
