Data from: Grazing and climate interact to regulate greening trends in Mediterranean grasslands

Rincon-Madroñero, Marina 1 2 ; Sánchez-Zapata, José Antonio1 2; Morant, Jon1 2 3; Barbosa, Jomar1 2

Published Dec 10, 2025 on Dryad. https://doi.org/10.5061/dryad.0gb5mkmfq

Data files

Dec 10, 2025 version files 52.55 KB

grazingData.csv

46.82 KB
README.md

5.72 KB

Abstract

A widespread increase in vegetation photosynthetic activity and biomass, known as greening, has been detected since the 1980s. While global climate change explains some of this trend, regional-level land-use management also plays a significant role. In grasslands, the fine-scale movement of livestock is a key driver of vegetation dynamics, likely affecting greening. Here, we investigate how spatial changes in grazing pressure interact with regional climate to determine long-term vegetation trends. Our study focuses on a Mediterranean mountainous region in south-eastern SpainF, with a long history of traditional grazing. We used GPS collar data from 35 livestock herds to create a high-resolution map of herbivory pressure. We then analysed vegetation dynamics from 1985 to 2024 using NDVI time series and evaluated vegetation responses to climate and herbivory with multivariate autoregressive models and logarithmic regressions. Greening was the dominant trend occurring in 90 % of the study area. However, increasing grazing pressure diminished this greening tendency. We found a non-linear response: the greening trend reduced sharply even under low levels of grazing and then stabilized at medium to high intensities. Despite hotspots of grazing pressure, no significant browning (vegetation decline) was detected. Vegetation became more sensitive to climate under higher herbivory pressure. In heavily grazed areas, vegetation greenness responded more strongly to increases in precipitation and showed grater negative responses to rising temperatures. This suggests that grazing regulates how grassland vegetation responds to climatic shifts. Policy implications. Our results show that traditional grazing systems can maintain grassland stability without causing degradation, but they can also increase the ecosystem’s vulnerability to climate change. We provide clear evidence that GPS-based livestock monitoring is a powerful and scalable tool for sustainable rangeland management. It provides land managers with fine-scale data on grazing pressure, enabling them to identify areas at risk, enhance ecosystem resilience to climate change, and support both conservation goals and productive landscapes at local and regional scales.

Dataset DOI: 10.5061/dryad.0gb5mkmfq

Description of the data and file structure

This dataset contains the data on the 35 herbivory pressure index of 500 x 500 m grid cells across the study area, Los Campos de Hernán Perea and the greenness and climatic variable between 1985 and 2024. To create the herbivory index map, we calculated monthly utilisation distributions (UDs) for each herd using the kernel density estimator in the R package amt (Signer, 2018). Monthly NDVI was obtained from Landsat 5, 7, and 8 satellites (www.usgs.gov), and precipitation and temperature data were sourced from the ERA5-Land dataset. All remote sensing data were processed using Google Earth Engine. We used a multivariate autoregressive state-space (MARSS) model implemented in the MARSS package in R to infer the underlying vegetation trends from incomplete datasets without prior variance estimates . We also quantified how vegetation in each grid cell responds to the climatic covariates (precipitation and temperature). To assess vegetation trends modulated by herbivory, we identified the long-term trend (greening, browning, or no trend) in the 39-year NDVI time series for each grid cell using the Mann-Kendall test for trend detection, and the Theil–Sen to quantify the trend magnitude.

Files and variables

File: grazingData.csv

Description: Parameters of the MARSS models (Hinrichsen and Holmes, 2009) resulting from time series data between 1985 and 2024, based on 500 × 500 m cells distributed in a regular grid across the study area Los Campos de Hernán Perea. Id corresponds to the name of the resulting model for each grid cell; Herb indicates the herbivory index of each cell; Iter is the number of iterations until model convergence; Conv indicates whether the model converged; R.NDVI5, R.NDVI7, and R.NDVI8 represent the observation noise matrices for the NDVI sensors from Landsat 5, 7, and 8, respectively; Q refers to the hidden process noise; and rain and temperature indicate the effect of these covariates on the hidden NDVI process. Finally, Theil-Sen refers to the slope of the NDVI time series as estimated by the Theil-Sen and Mann-Kendall statistics. All parameters showed statistical support.

Variables

Id: Name of the resulting model for each grid cell
Herb: Herbivory index of each cell
Iter.: Number of iterations until MARSS model convergence used to built the NDVI time series of the cell
AIC: MARSS model AIC
Convergence: Indicative of the model convergence
R.NDVI5.Est: Observation noise estimate for the NDVI sensors from Landsat 5 form the MARSS model used to built the time series of each cell
R.NDVI5.lowCI: Observation noise low credibility interval (CI) for the NDVI sensors from Landsat 5 form the MARSS model used to built the time series of each cell
R.NDVI5.highCI: Observation noise high credibility interval (CI) for the NDVI sensors from Landsat 5 form the MARSS model used to built the time series of each cell
R.NDVI7.Est: Observation noise estimate for the NDVI sensors from Landsat 7 form the MARSS model used to built the time series of each cell
R.NDVI7.lowCI: Observation noise low credibility interval (CI) for the NDVI sensors from Landsat 7 form the MARSS model used to built the time series of each cell
R.NDVI7.highCI: Observation noise high credibility interval (CI) for the NDVI sensors from Landsat 7 form the MARSS model used to built the time series of each cell
R.NDVI8.Est: Observation noise estimate for the NDVI sensors from Landsat 8 form the MARSS model used to built the time series of each cell
R.NDVI8.lowCI: Observation noise low credibility interval (CI) for the NDVI sensors from Landsat 8 form the MARSS model used to built the time series of each cell
R.NDVI8.highCI: Observation noise high credibility interval (CI) for the NDVI sensors from Landsat 8 form the MARSS model used to built the time series of each cell
Q.Est: Hidden process noise estimate form the MARSS model used to built the time series of each cell
Q.lowCI: Hidden process noise low credibility interval (CI) form the MARSS model used to built the time series of each cell
Q.highCI: Hidden process noise high credibility interval (CI) form the MARSS model used to built the time series of each cell
Prec.Est: Precipitation estimate form the MARSS model used to built the time series of each cell
Prec.lowCI: Precipitation low credibility interval (CI) form the MARSS model used to built the time series of each cell
Prec.highCI: Precipitation high credibility interval (CI) form the MARSS model used to built the time series of each cell
Temp.Est: Temperature estimate form the MARSS model used to built the time series of each cell
Temp.lowCI: Temperature low credibility interval (CI) form the MARSS model used to built the time series of each cell
Temp.highCI: Temperature high credibility interval (CI) form the MARSS model used to built the time series of each cell
TheilSen: Slope of the NDVI time series as estimated by the Theil-Sen and Mann-Kendall statistics using MARSS model

Code/software

Data can be open using Excel or R

Access information

Other publicly accessible locations of the data:

None

Data was derived from the following sources:

Monthly NDVI were obtained from Landsat 5, 7 and 8 satellites (www.usgs.gov) and precipitation and temperature data were sourced from the ERA5-Land dataset. All remote sensing data were downloaded using Google Earth Engine.