A5002459081- Soil Moisture and Remote Sensing
- Precipitation Measurement and Analysis
- 3D Modeling in Geospatial Applications
- Geographic Information Systems Studies
- Computational Physics and Python Applications
- Meteorological Phenomena and Simulations
- Climate variability and models
- Synthetic Aperture Radar (SAR) Applications and Techniques
- Remote Sensing and LiDAR Applications
- Atmospheric and Environmental Gas Dynamics
- Remote-Sensing Image Classification
- Remote Sensing in Agriculture
- Plant Water Relations and Carbon Dynamics
- Advanced Computational Techniques and Applications
- Geophysical Methods and Applications
- Landslides and related hazards
- Geological Modeling and Analysis
- Geophysics and Gravity Measurements
- Smart Agriculture and AI
- Soil Geostatistics and Mapping
- Distributed and Parallel Computing Systems
- Geochemistry and Geologic Mapping
- Flood Risk Assessment and Management
- Soil and Unsaturated Flow
- Genetic and phenotypic traits in livestock
Clark University
2023-2025
Radiant Earth
2018-2023
Columbia University
2016-2019
Massachusetts Institute of Technology
2014-2019
Radiant Research (United States)
2018-2019
Earth Island Institute
2019
Rutgers, The State University of New Jersey
2019
University of California, Los Angeles
2019
Environmental Earth Sciences
2018-2019
Wageningen University & Research
2019
Abstract. Satellite-retrieved solar-induced chlorophyll fluorescence (SIF) has shown great potential to monitor the photosynthetic activity of terrestrial ecosystems. However, several issues, including low spatial and temporal resolution gridded datasets high uncertainty individual retrievals, limit applications SIF. In addition, inconsistency in measurement footprints also hinders direct comparison between gross primary production (GPP) from eddy covariance (EC) flux towers...
Much of the progress in development highly adaptable and reusable artificial intelligence (AI) models is expected to have a profound impact on Earth science remote sensing. Foundation are pre-trained large unlabeled datasets through self-supervision, then fine-tuned for various downstream tasks with small labeled datasets. There an increasing interest within scientific community investigate how effectively build generalist AI that exploit multi-sensor data observation applications. This...
Abstract. A new global estimate of surface turbulent fluxes, latent heat flux (LE) and sensible (H), gross primary production (GPP) is developed using a machine learning approach informed by novel remotely sensed solar-induced fluorescence (SIF) other radiative meteorological variables. This the first study to jointly retrieve LE, H, GPP SIF observations. The uses an artificial neural network (ANN) with target dataset generated from three independent data sources, weighted based on triple...
Solar-induced fluorescence (SIF) observations from space have resulted in major advancements estimating gross primary productivity (GPP). However, current SIF remain spatially coarse, infrequent, and noisy. Here we develop a machine learning approach using surface reflectances Moderate Resolution Imaging Spectroradiometer (MODIS) channels to reproduce normalized by clear sky irradiance the Global Ozone Monitoring Experiment-2 (GOME-2). The resulting product is proxy for ecosystem...
Abstract. Validation of precipitation estimates from various products is a challenging problem, since the true unknown. However, with increased availability wide range instruments (satellite, ground-based radar, and gauge), it now possible to apply triple collocation (TC) technique characterize uncertainties in each products. Classical TC takes advantage three collocated data same variable mean squared error each, without requiring knowledge truth. In this study, triplets among NEXRAD-IV,...
Remote sensing, or Earth Observation (EO), is increasingly used to understand system dynamics and create continuous categorical maps of biophysical properties land cover, especially based on recent advances in machine learning (ML). ML models typically require large, spatially explicit training datasets make accurate predictions. Training data (TD) are generated by digitizing polygons high spatial-resolution imagery, collecting situ data, using pre-existing datasets. TD often assumed...
Abstract. Characterizing soil moisture at spatiotemporal scales relevant to land surface processes (i.e., of the order 1 km) is necessary in quantify its role regional feedbacks between and atmospheric boundary layer. Moreover, several applications such as agricultural management can benefit from information fine spatial scales. Soil estimates current satellite missions have a reasonably good temporal revisit over globe (2–3-day repeat time); however, their finest resolution 9 km. NASA's...
Climate change, increasing population and changes in land use are all rapidly driving the need to be able better understand surface water dynamics. The targets set by United Nations under Sustainable Development Goal 6 relation freshwater ecosystems also make accurate monitoring increasingly vital. However, last decades have seen a steady decline situ hydrological availability of growing volume environmental data from free open satellite systems is being recognized as an essential tool for...
Tackling data challenges and incorporating physics into machine learning models will help unlock the potential of artificial intelligence to answer Earth science questions.
Mapping agricultural fields using high-resolution satellite imagery and deep learning (DL) models has advanced significantly, even in regions with small, irregularly shaped fields. However, effective DL often require large, expensive labeled datasets, which are typically limited to specific years or regions. This restricts the ability create annual maps needed for monitoring, as changes farming practices environmental conditions cause domain shifts between locations. To address this, we...
There is a significant growth in development and utilization of foundation models for geospatial applications. These are trained on large scale unlabeled data commonly evaluated downstream tasks using labeled datasets. While this approach provides platform to assess the performance model specific tasks, there has been limited effort quantify characteristics after pre-training. Explainable AI (XAI) approaches aim increase accuracy transparency make their results interpretable. In...
Regularly updated and accurate land cover maps are essential for monitoring 14 of the 17 Sustainable Development Goals. Multispectral satellite imagery provide high-quality valuable information at global scale that can be used to develop classification models. However, such a application requires geographically diverse training dataset. Here, we present LandCoverNet, dataset based on Sentinel-2 observations 10m spatial resolution. Land class labels defined annual time-series Sentinel-2,...
An approach for estimating vertically continuous soil moisture profiles under varying vegetation covers by combining remote sensing with (hydrological) modeling is proposed. The uses decomposed scattering components, after the removal of components from fully polarimetric P-band SAR observations. By comparing these hydrological simulations, surface until a depth 30 cm (assumed average penetration depth) are estimated. Here, model HYDRUS-1D, as representative any model, employed to simulate...
Recent progress in self-supervision has shown that pre-training large neural networks on vast amounts of unsupervised data can lead to substantial increases generalization downstream tasks. Such models, recently coined foundation have been transformational the field natural language processing. Variants also proposed for image data, but their applicability remote sensing tasks is limited. To stimulate development models Earth monitoring, we propose a benchmark comprised six classification...