Search Results (3)

Harmful algal blooms (HABs) are increasing in frequency in the Chesapeake Bay, posing risks to marine ecosystems, water quality, and public health. Chlorophyll-a (Chl-a) is a widely used indicator of algal biomass, and satellite observations such as Sentinel-3 Ocean and Land Color Instrument (OLCI) enable large-scale monitoring of bloom dynamics. However, cloud cover and atmospheric interference frequently introduce missing pixels in daily satellite products, reducing temporal continuity and limiting monitoring reliability. Satellite-derived chlorophyll-a (Chl-a) data exhibit substantial missingness, with daily pixel gaps ranging from approximately 52.30% to 100% (mean ≈ 88.95%). This study evaluates spatial interpolation, EOF-based, supervised machine-learning, deep-learning, and convolutional autoencoder approaches for reconstructing missing Chl-a values. Sentinel-3 OLCI Chl-a data from 2023–2024 were used for model training, while data from 2025 served as a temporally independent test set to avoid spatiotemporal leakage. To simulate cloud-induced data gaps, artificial missingness scenarios ranging from 50% to 90% were applied for the Inverse Distance Weighting (IDW) and Data Interpolating Empirical Orthogonal Functions (DINEOF) baseline approaches, while machine-learning, deep-learning, and convolutional autoencoder models were evaluated using real satellite-derived missing observations. The evaluated models include IDW, DINEOF, K-Nearest Neighbors (KNN), Random Forest (RF), Extra Trees (ET), XGBoost, a Long Short-Term Memory (LSTM) network, and a Temporal Data Interpolating Convolutional Autoencoder (Temporal DINCAE). Model performance was assessed using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), prediction bias, and the coefficient of determination (R²). Results indicate that tree-based ensemble models outperform spatial interpolation and EOF-based methods, with XGBoost achieving the best overall performance (R² ≈ 0.86; RMSE ≈ 9.61 mg m⁻³). The LSTM model achieved lower prediction errors (RMSE ≈ 5.87 mg m⁻³; MAE ≈ 2.16 mg m⁻³), highlighting the benefit of incorporating temporal dependencies, although with slightly reduced variance capture. The convolutional autoencoder-based Temporal DINCAE model achieved strong reconstruction performance (R² ≈ 0.84; RMSE ≈ 11.15 mg m⁻³). Uncertainty quantification shows that Extra Trees tends to underestimate uncertainty with narrower prediction intervals, whereas XGBoost provides better-calibrated but wider intervals. Full article

Continuous daily estimates of evapotranspiration (ET) spatially distributed at plot scale are required to monitor the water loss and manage crop irrigation needs. Remote sensing approaches in the thermal infrared (TIR) domain are relevant to assess actual ET and soil moisture status but due to lengthy return intervals and cloud cover, data acquisition is not continuous over time. This study aims to assess the performances of 6 commonly used as well as two new reference quantities including rainfall as an index of soil moisture availability to reconstruct seasonal ET from sparse estimates and as a function of the revisit frequency. In a first step, instantaneous in situ eddy-covariance flux tower data collected over multiple ecosystems and climatic areas were used as a proxy for perfect retrievals on satellite overpass dates. In a second step, instantaneous estimations at the time of satellite overpass were produced using the Soil Plant Atmosphere and Remote Sensing Evapotranspiration (SPARSE) energy balance model in order to evaluate the errors concurrent to the use of an energy balance model simulating the instantaneous IRT products from the local surface temperature. Significant variability in the performances from site to site was observed particularly for long revisit frequencies over 8 days, suggesting that the revisit frequency necessary to achieve accurate estimates of ET via temporal upscaling needs to be fewer than 8 days whatever the reference quantity used. For shorter return interval, small differences among the interpolation techniques and reference quantities were found. At the seasonal scale, very simple methods using reference quantities such as the global radiation or clear sky radiation appeared relevant and robust against long revisit frequencies. For infra-seasonal studies targeting stress detection and irrigation management, taking the amount of precipitation into account seemed necessary, especially to avoid the underestimation of ET over cloudy days during a long period without data acquisitions. Full article

Currently, thermal remote sensing-based evapotranspiration (ET) models can only calculate instantaneous ET at the time of satellite overpass. Five temporal upscaling methods, namely, constant evaporative fraction (ConEF), corrected ConEF (CorEF), diurnal evaporative fraction (DiEF), constant solar radiation ratio (SolRad), and constant reference evaporative fraction (ConET_rF), were selected to upscale the instantaneous ET to daily values. Moreover, five temporal reconstruction approaches, namely, data assimilation (ET_EnKF and ET_SCE_UA), surface resistance (ET_SR), reference evapotranspiration (ET_ET_rF), and harmonic analysis of time series (ET_HANTS), were used to produce continuous daily ET with discrete clear-sky daily ET values. For clear-sky daily ET generation, SolRad and ConET_rF produced the best estimates. In contrast, ConEF usually underestimated the daily ET. The optimum method, however, was found by combining SolRad and ConET_rF, which produced the lowest root-mean-square error (RMSE) values. For continuous daily ET production, ET_ET_rF and ET_SCE_UA performed the best, whereas the ET_SR and ET_HANTS methods had large errors. The annual ET distributions over the Beijing area were calculated with these methods. The spatial ET distributions from ET_ET_rF and ET_SCE_UA had the same trend as ETWatch products, and had a smaller RMSE when compared with ET observations derived from the water balance method. Full article