Field Spectroscopy for Assessing Midday Leaf Water Potential in Amazonian Forest Environments: Preliminary Results

The increasing severity of droughts and their direct impact on the health of Amazon forest ecosystems underscore the urgent need to understand this phenomenon and to develop tools for large-scale monitoring. Leaf water potential (ψ_leaf) is a critical indicator of plant water status. However, traditional methods for measuring ψ_leaf are often logistically challenging and costly. Field spectroscopy offers a more efficient means of assessing plant water status, allowing scaling of information through predictive models that can be combined with imaging spectroscopy techniques from orbital and suborbital sensors. This study collected hyperspectral leaf data from three Amazonian forest environments during the El Niño period in October 2023: White Sand Forest, Flood Forest, and Upland Forest, all located at the Atto site. We collected two species from the forest canopy in each environment, resulting in six species and 43 samples. The reflectance measurements were taken immediately after the ψ measurement using a Scholander pump, around midday, with an ASD spectroradiometer covering the range from 350 nm to 2500 nm. The prediction model was developed using the entire data set by applying an optimized Partial Least Squares (PLS) regression model in Python. This was done after pre-processing the spectral data, which included jump correction functions, a Savitzky-Golay filter, and first derivative analysis. The resulting model showed good performance, with an R² of 0.73 and a mean squared error (MSE) of 0.21, although it still showed moderate generalization ability. The spectral bands that provide the most information about water potential are found in the near-infrared (NIR) range between 780 and 1100 nm, and the shortwave infrared (SWIR) range around 1700 and 2250 nm. These preliminary results support the idea that spectroscopic techniques can effectively indicate plant responses to water stress, which is critical in climate change. Such studies may facilitate more efficient monitoring of water status in Amazonian forest ecosystems. Future research should improve the use of spectroscopy in ecological studies of plant responses to environmental change by expanding sampling to more tree species and considering additional variables that reflect water stress, such as fuel moisture content (FMC), leaf water content (LWC), equivalent water thickness (EWT), and relative water content (RWT).