Spatio-temporal continuous LAI estimation based on the remote sensing foundation model

Abstract

LAI is one of the terrestrial essential climate variables in the Global Climate Observing System and a significant indicator for model simulation of terrestrial ecosystems, estimation of crop yields, and monitoring of vegetation changes. Due to the constraints of satellite observation frequency and quality, existing LAI products focus on cloud-free estimation. Research on LAI estimation with high temporal resolution and spatial continuity is often restricted by clouds in remote sensing imagery. Based on extensive core features related to remote sensing imagery, remote sensing foundation models could provide new methodology for the downstream tasks of imagery pixel reconstruction and spatio-temporal continuous LAI estimation, which can be achieved with only a small amount of data fine-tuning the pretrained model. Compared to other models, Prithvi foundation model offers advantages such as the comprehensive and unified pre-training dataset, a classic and stable model architecture, clearly defined downstream tasks, and suggested applicability in regression tasks. In this research, we focus on the remote sensing foundation model application on spatial-temporal continuous LAI estimation under both cloud-free and cloudy conditions, utilizing the Prithvi with100 million parameters and fine-tuned with 30 m resolution Harmonized Landsat and Sentinel-2 multi-band spectral data and Hi-GLASS LAI products in the China region. After investigating the Prithvi 's capabilities to estimate the numerical values at the pixel level, we utilize the fine-tuned Prithvi foundation model to achieve cloudy pixel reflectance reconstruction and subsequently enables spatial-temporal continuous LAI estimation. We initially estimate the LAI under the cloud-free conditions. Through ablation study of Prithvi downstream model construction, the optimized model with single-phase input reconstructs the macroscopic spatial pattern, achieving a Root Mean Square Error of 0.20. For the LAI estimation under the cloud cover, a two-stage method is employed: first, reconstructing the cloudy reflectance, and then estimating the LAI values. With multi-temporal phases inputs, the cloudy pixel value reconstruction model can restore the basic reflectance features achieving a RMSE of 0.39, but show grid artifacts and lose some high-frequency details. To address them, we modify the loss function, which improves detail restoration. The reconstructed cloud-free reflectance is then used in the LAI estimation model constructed under cloud-free conditions to estimate the LAI values under cloud cover, yielding a combined RMSE of 0.69. The two-stage model improves the spatial and temporal continuity of LAI estimation, achieving spatio-temporal reconstruction of 30m resolution LAI, effectively correcting the numerical missing caused by clouds, detecting temporal anomalies and providing a method for cloudy area LAI estimation and anomaly correction in vegetation dynamic monitoring. Through the research, the downstream regression model of Prithvi is optimized to reconstruct the reflectance affected by clouds and estimate the LAI values with a resolution of 30 m and shorter temporal interval. The model optimization logic, dominated by loss function optimization, architecture adjustment and post-processing compensation enhances the model's detail capture ability and improves the LAI frequency and quality. This method provides an important reference for the construction of regression downstream models, helping researchers optimize models and obtain more accurate and reliable prediction results.