Assessing vegetation seasonality in Tropical Dry Forests: Multi-site comparison of MODIS NDVI and CHIRPS Precipitation Time Series.

Tropical Dry Forests (TDFs) exhibit strong couplings between precipitation and vegetation phenology. TDFs are broadly defined as ecological formations characterized by -but not limited to- the presence of deciduous tree species, occurring in tropical regions where precipitation is highly seasonal, with a distinct dry season lasting several consecutive months (Miles et al., 2006), and mean annual rainfall ranging from 250 to 2000 mm (Murphy, 1986; Holdridge, 1969). This broad definition covers multiple bioclimatic types depending on authors, ranging from woodland savannah to moist semi-deciduous forests. A final agreement on their extent and definition may remain unattainable (Murphy, 1986; Blasco, 2000), and areas locally recognized as TDFs sometimes defy the most commonly used thresholds (Pando-Ocon, 2021).

TDFs provide numerous critical ecosystem services including carbon sequestration, support for local livelihoods, maintenance of biodiversity through habitat provision, high levels of floristic endemism, and a buffering effect against desertification (Siyum, 2020; Mendes, 2025). Despite these vital services, TDFs have been described as one of the most threatened biomes worldwide, experiencing extensive loss, fragmentation, and degradation driven by agricultural conversion, fire, and other anthropic pressures, with less than one-third of original forest area remaining (Stan et al., 2024). However, these valuable ecosystems have long suffered from a lack of public interest and from limited attention in conservation policies and research (Santos et al., 2011).

In this context, our study aims to assess and compare the dynamics of vegetation phenology, precipitation and their relationships across multiple TDF hotspots: Santa Rosa national park (Costa Rica), the Caatinga ecoregion (Brazil), Bandipur and Mudumalai national parks (Southwestern Ghats, India), and the Chizarira and Sijarira national parks within the Miombo and Mopane woodlands (Zimbabwe). These sites collectively span more or less pronounced gradients in rainfall seasonality, topography, edaphic conditions, tree density and forest composition. Using two decades (2003-2023) of MODIS NDVI time-series and CHIRPS precipitation data, we investigate inter and intra-site variability based on the visual interpretation of weekly mean NDVI and precipitation time-series over multiple points along a rainfall gradient, and compare metrics characterizing both phenological dynamics (amplitude of the vegetation index time-series and temporal phenometrics) and rainfall regimes (precipitation values and temporality of the rainy season). Additionally, we cross both times-series to assess their relationship (lag time between rainy season onset and vegetation response).

Other drivers of phenology are mentioned as TDF definitions are not limited by the presence of deciduous vegetation, and phenological dynamics in these systems may be impacted by other factors such as access to groundwater or atmospheric moisture, floristic composition and stand age (Hasselquist et al., 2010; Cuba et al., 2017; Siyum, 2020; Parthasarathy et al., 2008).

Overall, understanding and spatializing the links between phenology, environmental drivers and the associated plant functional traits in TDFs has important implications for the assessment of carbon flux and storage, the projection of ecosystem resilience and redistribution under climate change as well as accurate description of land-cover, ecotones and habitat connectivity (Li et al., 2024; Pereira Dos Santos et al., 2025; Ribeiro et al., 2025).