Ph.D. Candidate · Johns Hopkins University

Remote sensing, hydrology, and AI for the global water cycle

I work across satellite remote sensing, numerical hydrological modeling, and machine learning — including foundation models — to estimate precipitation and soil moisture from space. My models embed infiltration, routing, and soil-water balance physics directly into learned, interpretable systems, and are built to work globally, including where rain gauges do not reach.

Satellite observation → land surface response → physics-informed inference

Remote SensingHydrologyAI & Machine LearningFoundation ModelsNumerical ModelingPrecipitationSoil Moisture

Publications Research projects Contact

Recognition

Awards & honors

Competitive research funding and recognition for peer-review contributions to the field.

2026

Research Grant Awardee

Virginia Space Grant Consortium (VSGC) / NASA

Competitive research funding supporting satellite-based precipitation retrieval and soil moisture science.

2024

Outstanding Reviewer

Vadose Zone Journal

Selected for exceptional peer-review contributions, one of eleven journals served as reviewer.

Selected service

Session ChairAGU Fall Meeting 2026 — Geospatial Foundation Models and Remote Sensing for Hydrologic Prediction
Session ChairAGU Fall Meeting 2025 — Advances in Remote Sensing, AI, and Modeling for Hydrology
Review CommitteeIEEE IGARSS 2026 — Scientific Review Committee Member

Research focus

Four questions driving the work

From algorithm development to global-scale deployment — physical rigor first, learning second.

🛰

Remote sensing

Soil moisture, precipitation, evapotranspiration, and terrestrial hydrology from SMAP, ASCAT, and AMSR2.

🧠

Physics-informed AI

Deep learning architectures with hydrological formulations embedded as interpretable, constrained components.

🌧

Hydroclimate extremes

Drought–flood transitions and precipitation reconstruction in data-scarce regions worldwide.

◫

Validation frameworks

Foundation-model-driven validation, downscaling, and data assimilation for Earth observation products.

Flagship project

A global daily precipitation model that calibrates itself

Multiple physics-based expert systems — Green-Ampt infiltration, Muskingum-Cunge routing, soil-water balance — fused through a regime-aware gating network that selects the right expert for each hydroclimatic regime.

Complements the GPM mission and upcoming NASA Earth observation missions
Extends reliable precipitation estimates to ungauged and data-sparse regions
Interpretable by construction — no black-box parameterization

Method

Inverse hydrological inference

Rather than retrieving rainfall from the top down, SM2RAIN-NWF and its successors read rainfall out of the soil's own water balance — inverting the hydrological process to recover the forcing that produced it.

Rainfall estimated from satellite soil moisture as the only input
Streamflow records inverted to reconstruct historical precipitation
Active and passive sensor fusion to close observational gaps

8Peer-reviewed journal articles

10Conference presentations

11Journals served as reviewer

6Active research projects

Latest

Recent publications

All publications →

Interested in collaborating on Earth observation and hydrology?

I welcome collaboration on physics-informed retrieval algorithms, satellite soil moisture validation, and precipitation reconstruction — and I am always glad to review for journals in the field.

Email me Google Scholar

About · Doctoral research

AI-driven and physics-informed rainfall retrieval for data-scarce regions

Supporting GPM through NASA multi-mission observations — my dissertation at Johns Hopkins, advised by Venkataraman Lakshmi.

Satellite soil moisture holds an indirect record of the rainfall that produced it — but reading rainfall back out of that record is an ill-posed inverse problem. Soil moisture also responds to drainage, evapotranspiration, runoff, retrieval noise, and delayed land-surface response, and intense storms can leave a weak or even negative same-day signal.

My work makes that inversion tractable by constraining it with physics rather than replacing physics with a black box. Hydrological formulations — Green-Ampt infiltration, Muskingum-Cunge routing, analytical net water flux — stay intact inside learned architectures, while neural operators, graph attention networks, and geospatial foundation models supply the parameterization those formulations need in places where no gauge exists.

Increasingly the question is not just how much rain but how much to trust the answer — so recent work produces calibrated distributions rather than single numbers, and uses conformal risk control to retain only the retrievals that meet a stated error tolerance.

Interests

Research interests

Physics-informed learning and geospatial foundation models sit at the centre; everything else follows from applying them to real Earth observations.

🧠

Physics-informed AI & neural operators

Hydrological formulations embedded as constrained, interpretable components — with Fourier-domain operator learning, learned advection, and source–sink corrections for land-surface memory.

◫

Geospatial foundation models

Frozen encoders such as Prithvi-EO-2.0 over HLS imagery to represent persistent land-surface conditions, and foundation-model-driven validation frameworks for Earth observation.

📈

Uncertainty-aware retrieval

Probabilistic and ensemble retrieval through conditional flow matching, plus conformal risk control and selective prediction to bound error at a user-chosen tolerance.

🛰

Satellite & GNSS-R remote sensing

SMAP, ASCAT, AMSR2, CYGNSS delay–Doppler maps, and emerging NISAR and OPERA products for soil moisture and land surface state.

⇄

Inverse modeling & data assimilation

Recovering rainfall from soil moisture, streamflow, and learned subsurface boundary conditions, with regionalization that transfers to ungauged basins.

🌧

Hydroclimate extremes

Drought–flood transitions, heavy-event recall, and precipitation reconstruction across data-scarce regions worldwide.

Projects

Research projects

Current work carried out as a Graduate Research Assistant at Johns Hopkins and the University of Virginia, plus earlier algorithm development and fieldwork.

Graduate Research Assistant

Johns Hopkins University · Whiting School of Engineering

2026 — present

01In progress

Physics-informed hybrid deep learning for rainfall retrieval

A physics-informed learning framework that replaces fixed process-model parameters with learned, interpretable ones — keeping the hydrological process model intact while letting data determine how it is parameterized.

Physics-informed DLRainfall retrieval

02In progress

Risk-calibrated selective prediction for GNSS-R soil moisture

A selective prediction framework that reads CYGNSS delay–Doppler maps, observation geometry, and land-surface characteristics to identify which soil moisture retrievals are reliable, using conformal risk control to retain observations under a user-defined error tolerance.

CYGNSS / GNSS-RConformal risk controlUncertainty

Graduate Research Assistant

University of Virginia · Civil & Environmental Engineering

2024 — 2026

01Completed

Calibration-free regionalization for SM2RAIN-NWF

A self-calibration framework that eliminates long calibration periods by regionalizing model parameters with unsupervised learning — K-means, Gaussian mixture models, agglomerative clustering, and Growing Neural Gas. Hold-out and leave-one-out protocols demonstrate improved transferability to ungauged regions.

Unsupervised learningGeneralization

02Completed

Green-Ampt-enhanced retrieval and uncertainty quantification

Physically based infiltration modeling integrated into SM2RAIN for robustness under heterogeneous soil and antecedent-moisture conditions, with systematic sensitivity analysis and Monte Carlo uncertainty quantification across climate, soil, and land-cover regimes.

Green-AmptMonte Carlo UQ

03Completed

Inverse river-network modeling for precipitation reconstruction

An inverse hydrologic framework that recovers effective rainfall from outlet discharge by coupling Muskingum–Cunge routing with Green–Ampt infiltration, with emphasis on river-network dynamics, runoff-generation processes, and data-scarce basins.

Inverse modelingStreamflowMuskingum–Cunge

04In progress

Mixture-of-Physics Experts for global precipitation retrieval

A global daily precipitation model that fuses multiple physics-based experts through a regime-aware, reliability-weighted gating network, with gauge-only magnitude supervision, multi-timescale sequence encoding, and domain adaptation for physically interpretable prediction in data-sparse and ungauged regions.

Mixture of expertsDomain adaptationGlobal

Foundations

Earlier work

Algorithm development and field measurement from the M.Sc. period that the current research builds on.

SM2RAIN-NWF: integrating SM2RAIN with the net water flux model

2020 — 2022

Building on the net water flux analytical model of Sadeghi et al. (2019), SM2RAIN-NWF estimates rainfall using satellite soil moisture as its only input — the algorithm the later work extends.

In-situ soil moisture vs. satellite and model products, Lake Urmia basin

2019 — 2020

A field measurement campaign compared against AMSR2, SMAP L3, SMAP L4, and GLDAS, stratified by climate, soil texture, and land cover.

“What we observe is not nature itself, but nature exposed to our method of questioning.”
— Werner Heisenberg

Updated April 2026

Publications

Peer-reviewed journal articles and conference contributions. Filter by year or search titles, venues, and co-authors.

Peer-reviewed journals

⌕

Conferences

AGU Annual Meeting and IEEE IGARSS presentations

Looking for a preprint or full text?

Most articles are available open access through the DOI links above. For anything paywalled, email me and I will send a copy.

Request a paper ResearchGate

Network

Collaborators & co-authors

Across NASA Goddard, US Army ERDC, CNR Italy, University of Virginia, GIST, Wageningen, and Reading.

Advisor

Venkataraman Lakshmi

B. Howell Griswold, Jr. Professor, Whiting School of Engineering, Johns Hopkins University · President, Hydrology Section, AGU

Profile →

Co-author

John Bolten

Chief, Hydrological Sciences Laboratory, NASA Goddard Space Flight Center

Profile →

Co-author

John Eylander

Coastal and Hydraulics Laboratory, US Army Engineer Research and Development Center, Vicksburg, MS, USA

Co-author

Luca Brocca

Director of Research, Research Institute for Geo-Hydrological Protection, CNR Italy

Co-author

Hyunglok Kim

School of Environment and Energy Engineering, Gwangju Institute of Science and Technology (GIST)

Co-author

Ameneh Tavakol

Water Resources Control Engineer, California Water Resources Control Board, Sacramento, CA

Co-author

Sina Nabaee

Doctoral candidate, Wageningen University & Research

Co-author

Hamidreza Mosaffa

Postdoctoral Researcher, University of Reading, UK

Curriculum Vitae

Experience, education, awards & service

Research experience

2026 — present

Graduate Research Assistant

Johns Hopkins University

Physics-informed hybrid deep learning for rainfall retrieval
Risk-calibrated selective prediction for GNSS-R soil moisture retrieval

2024 — 2026

Graduate Research Assistant

University of Virginia

Calibration-free regionalization for SM2RAIN-NWF
Green-Ampt-enhanced retrieval and uncertainty quantification
Inverse river-network modeling for precipitation reconstruction
Mixture-of-Physics Experts for global precipitation retrieval

Education

2026 — present

Johns Hopkins University

Ph.D., Geography & Environmental Engineering

Department of Environmental Health and Engineering · Baltimore, MD, USA
Dissertation: AI-Driven and Physics-Informed Rainfall Retrieval for Data-Scarce Regions: Supporting GPM Through NASA Multi-Mission Observations
Advisor: Venkataraman Lakshmi

2024 — 2026

University of Virginia

Ph.D., Civil Engineering

Charlottesville, VA, USA

2018 — 2021

Science and Research Branch, IAU

M.Sc., Civil Engineering — Water Resources Engineering & Management

Tehran, Iran · GPA 3.88 / 4.0
Thesis: Estimation of rainfall based on water balance equations and net water flux in soil using satellite-based soil moisture data

B.Sc.

Islamic Azad University

B.Sc., Civil Engineering

Awards & honors

Research Grant Awardee, Virginia Space Grant Consortium (VSGC) / NASA — 2026
Vadose Zone Journal Outstanding Reviewer — 2024

Service, session chairing & peer review

Session Chair, AGU Fall Meeting 2026 — Geospatial Foundation Models and Remote Sensing for Hydrologic Prediction
Session Chair, AGU Fall Meeting 2025 — Advances in Remote Sensing, AI, and Modeling for Hydrology
Scientific Review Committee, IEEE IGARSS 2026
IEEE Transactions on Geoscience and Remote Sensing
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Science of the Total Environment · Journal of Hydrology
Scientific Data · PLOS ONE · Vadose Zone Journal
Environmental Modelling & Software
Water Resources Management · International Journal of Climatology
American Journal of Remote Sensing

Teaching & mentoring

Teaching assistant · co-instructor

Fluid Mechanics (CE 3210)

University of Virginia

Co-taught roughly half the semester through lectures and in-class instruction, under the instructor of record
Designed midterm and final exam questions aligned to the course learning objectives
Graded exams and returned detailed, timely feedback

Graduate mentoring

Research mentoring

Johns Hopkins University

Mentored a first-year student in the research group on selecting a research topic and developing a proposal
Advised on framing the research question and scoping a tractable first study

Get in touch

Contact

Happy to hear about collaborations, data sharing, and reviewing requests.

✉

University emailmsaeedi1@jh.edu

✉

Personal emailmohammadsaeedi.wrm@gmail.com

GS

Google ScholarCitations & metrics

iD

ORCID0000-0001-8100-6102

Sc

ScopusAuthor profile

RG

ResearchGatePreprints & full texts

in

LinkedInmohammad-saeedi-wrm

✕

X / Twitter@MSaeediiii

Prospective collaborators and students

If you have a keen interest in the intersection of climate change, satellite remote sensing, and hydrological modeling, I am glad to talk about joint work, data, and code.

Start a conversation