Grasslands in the Iberian Peninsula are some of the most valuable ecosystems in Europe and are vulnerable because of their location in arid-semiarid regions. Remote sensing techniques have the potential for monitoring grasslands using vegetation indices (VIs), which can reveal bare soil and non-photosynthetic vegetation reflectance in these regions. The temporal variability of the VI time-series is commonly measured as the standard deviation of the records, insufficient to study the system dynamics. Recurrence plots (RP) and recurrence quantification analysis (RQA) allow us to visualize and quantify system dynamics based on topology. These advanced analyses calculate stochasticity through determinism (DET) and predictability degree based on the average length of diagonal structures (LT). This study aims to evaluate RPs, Cross Recurrence Plots (CRP), and RQA to visualize and quantify VIs and climatic series and their anomalies responses dynamics.

The approach was tested at two sites with different agro-climatic characteristics in Central Spain: Guadalix de la Sierra (ZGU) and Soto del Real (ZSO). The Normalized Difference Vegetation Index (NDVI) and the Modified Soil Adjusted Vegetation Index (MSAVI), which includes a soil factor, were calculated from MODIS images. Temperature (TEMP) and precipitation (PCP) series were acquired from two meteorological stations close to each site. All the anomalies series were obtained seasonally adjusted to the original ones.

RPs showed the different VIs responses in both areas. The MSAVI patterns showed a clear structure, while NDVI presented a noisy pattern showing a higher random behaviour. To quantify these visual differences with RQA, four parameters were calculated, including DET and LT. The MSAVI series and the corresponding anomaly series presented higher DET and LT than the NDVI in both sites. CRPs of VIs with TMIN shows a higher DET in ZGU, suggesting that time synchronization is higher than in ZSO. When CRPs were applied over anomalies series, we could observe that precipitation is more synchronized with VIs anomalies, most likely because of the influence of soil moisture. LT of CRPs in all the cases was low and close to 2, indicating low predictability. Overall, our results suggest that VI time series show a more evident pattern and differentiate both sites better when a variable soil adjustment factor is included. Therefore, in arid-semiarid grasslands, a soil factor should be considered in VI calculations.

伊比利亚半岛的草原是欧洲最有价值的生态系统之一，由于它们位于干旱半干旱地区，因此很脆弱。遥感技术具有使用植被指数 (VI) 监测草原的潜力，可以揭示这些地区的裸土和非光合植被反射率。VI 时间序列的时间可变性通常作为记录的标准偏差来衡量，不足以研究系统动力学。递归图 (RP) 和递归量化分析 (RQA) 使我们能够基于拓扑可视化和量化系统动力学。这些高级分析通过确定性 (DET) 和基于对角结构平均长度 (LT) 的可预测度计算随机性。本研究旨在评估 RP，

该方法在西班牙中部具有不同农业气候特征的两个地点进行了测试：Guadalix de la Sierra (ZGU) 和 Soto del Real (ZSO)。归一化差异植被指数 (NDVI) 和修正土壤调整植被指数 (MSAVI) 包括土壤因子，是根据 MODIS 图像计算得出的。从每个站点附近的两个气象站获取温度 (TEMP) 和降水 (PCP) 系列。所有的异常序列都是根据原始序列进行季节性调整的。

RP 在这两个领域显示出不同的 VI 响应。MSAVI 模式显示出清晰的结构，而 NDVI 则呈现出噪声模式，显示出更高的随机行为。为了用 RQA 量化这些视觉差异，计算了四个参数，包括 DET 和 LT。MSAVI 系列和相应的异常系列在两个站点中呈现出比 NDVI 更高的 DET 和 LT。具有 TMIN 的 VI 的 CRP 在 ZGU 中显示更高的 DET，表明时间同步高于 ZSO。当 CRP 应用于异常系列时，我们可以观察到降水与 VIs 异常更同步，很可能是因为土壤水分的影响。所有病例中 CRP 的 LT 都很低，接近 2，表明可预测性低。总体，我们的结果表明，当包含可变的土壤调整因子时，VI 时间序列显示出更明显的模式，并且可以更好地区分两个地点。因此，在干旱半干旱草原，VI 计算中应考虑土壤因素。