Mohanad Albughdadi

Mohanad Albughdadi

The bottleneck of Earth Observation processing chains is not the arrival of new imagery but whether the surface is actually visible when the image arrives. We study this as an observability forecasting problem on EarthNet2021. Given recent multispectral imagery and exogenous weather drivers, the goal is to predict whether the next acquisition will be usable and, if not, when a usable view is likely to return. To do this, we adapt LeWorldModel, a joint-embedding predictive architecture world model, to cloud-aware Earth Observation sequences. The final pipeline converts raw minicubes into episodic HDF5 sequences with five image channels (blue, green, red, near-infrared, cloud mask) and eight meteorological and calendar covariates. The resulting model has 18.0M trainable parameters and is trained from scratch on 23,904 training episodes. The trained leWorldModel is evaluated under a locked protocol: linear probes are fit on train only, calibration choices are set on an internal validation split, and the fitted heads are then frozen for valsplit, IID, OOD, and extreme evaluation. On the full frozen-bundle observability benchmark, LeWorldModel consistently outperforms persistence. For next-step usability, balanced accuracy ranges from 0.769 to 0.887, compared with 0.493 to 0.556 for persistence. For exact first-usable-horizon prediction, accuracy ranges from 0.602 to 0.806, compared with 0.120 to 0.369 for persistence. Against a frozen LightGBM baseline fit on the same training windows, LeWorldModel is better on continuous clear/cloud regression and on exact recovery timing on valsplit, IID, and extreme, while LightGBM is stronger on the simpler binary any-usable-within-six task and is more robust on OOD. In separate sampled diagnostic analyses, LeWM also produces strong ranking-based anomaly signals under synthetic temporal inconsistencies.

Mohanad Albughdadi

Recent advances in Earth Observation have focused on large-scale foundation models. However, these models are computationally expensive, limiting their accessibility and reuse for downstream tasks. In this work, we investigate compact architectures as a practical pathway toward smaller general-purpose EO models. We propose a Metadata-aware Mixture-of-Experts Masked Autoencoder (MoE-MAE) with only 2.5M parameters. The model combines sparse expert routing with geo-temporal conditioning, incorporating imagery alongside latitude/longitude and seasonal/daily cyclic encodings. We pretrain the MoE-MAE on the BigEarthNet-Landsat dataset and evaluate embeddings from its frozen encoder using linear probes. Despite its small size, the model competes with much larger architectures, demonstrating that metadata-aware pretraining improves transfer and label efficiency. To further assess generalization, we evaluate on the EuroSAT-Landsat dataset, which lacks explicit metadata, and still observe competitive performance compared to models with hundreds of millions of parameters. These results suggest that compact, metadata-aware MoE-MAEs are an efficient and scalable step toward future EO foundation models.

Florian Mouret, Mohanad Albughdadi, Sylvie Duthoit et al.

Missing data is a recurrent problem in remote sensing, mainly due to cloud coverage for multispectral images and acquisition problems. This can be a critical issue for crop monitoring, especially for applications relying on machine learning techniques, which generally assume that the feature matrix does not have missing values. This paper proposes a Gaussian Mixture Model (GMM) for the reconstruction of parcel-level features extracted from multispectral images. A robust version of the GMM is also investigated, since datasets can be contaminated by inaccurate samples or features (e.g., wrong crop type reported, inaccurate boundaries, undetected clouds, etc). Additional features extracted from Synthetic Aperture Radar (SAR) images using Sentinel-1 data are also used to provide complementary information and improve the imputations. The robust GMM investigated in this work assigns reduced weights to the outliers during the estimation of the GMM parameters, which improves the final reconstruction. These weights are computed at each step of an Expectation-Maximization (EM) algorithm by using outlier scores provided by the isolation forest algorithm. Experimental validation is conducted on rapeseed and wheat parcels located in the Beauce region (France). Overall, we show that the GMM imputation method outperforms other reconstruction strategies. A mean absolute error (MAE) of 0.013 (resp. 0.019) is obtained for the imputation of the median Normalized Difference Index (NDVI) of the rapeseed (resp. wheat) parcels. Other indicators (e.g., Normalized Difference Water Index) and statistics (for instance the interquartile range, which captures heterogeneity among the parcel indicator) are reconstructed at the same time with good accuracy. In a dataset contaminated by irrelevant samples, using the robust GMM is recommended since the standard GMM imputation can lead to inaccurate imputed values.