Mitigating NDVI saturation in imagery of dense and healthy vegetation

Abstract

The Normalized Difference Vegetation Index (NDVI) is a widely used tool for assessing vegetation in remote sensing. However, its non-linear response to increasing vegetation vigor, especially in dense and healthy canopies, often leads to inaccurate estimations of vegetation status — a phenomenon known as NDVI saturation. This study investigates the underlying causes of NDVI saturation and proposes a two-stage saturation mechanism. In contrast to the first-stage optical saturation caused by biophysical constraints, we demonstrate that second-stage mathematical saturation can be mitigated through functional optimization. To address this issue, we introduce a new vegetation index (VI), Saturation Mitigated NDVI (NDVIsm), which modifies the NDVI structure by integrating an anti-saturation module. This modification enhances the index’s sensitivity to changes in vegetation vigor dynamics by amplifying variation among high NDVI values, thereby mitigating saturation effects. The performance of NDVIsm was validated across multiple remote sensing platforms and land cover types. NDVIsm eliminated saturation (defined as areas where more than 80% of pixels fall within less than 20% of the index range) and exhibited a more dispersed distribution with a pronounced right tail in the histogram. The effectiveness of NDVIsm in mitigating saturation was further confirmed by improved Pearson correlations (r) with canopy structure (0.3010 increase in Leaf Area Index (LAI) and 0.2452 increase in Leaf Structure Parameters (N)) and vegetation vigor (0.5575 increase in Chlorophyll Content (Cab)), as well as enhanced performance in machine learning (ML)-based yield prediction models when combined with non-optical and spatial features. Overall, NDVIsm demonstrates improved sensitivity in detecting subtle variations in dense and healthy vegetation and offers a promising solution to the saturation problem in environmental remote sensing.