datadi.us
Independent consulting
Earth observation and machine learning for water resources and environmental decision-making.
My work concerns the defensibility of inference from satellite, airborne, and learned-model observations: whether a retrieved signal carries sufficient information, with quantified uncertainty, to support a regulatory, operational, or investment decision in water quality, hydrology, soils, and land use.
Services
Areas of practice
Principal areas of engagement. Each is grounded in peer-reviewed work, listed under Selected work.
Validation of Earth observation and machine learning claims
Independent technical review of remote sensing retrievals and machine learning models prior to regulatory, operational, or financial commitment: identifiability of the target variable, calibration and validation design, transferability across sites and sensors, and characterization of predictive uncertainty.
Agencies, engineering firms, investorsWater quality and TMDL practice
Model selection and review for TMDL development, specification of the margin of safety, load and wasteload allocation under uncertainty, and post-implementation monitoring of BMP performance, including where remote sensing can credibly inform regulatory programs.
States, EPA regions, utilities, consultantsStreamflow and hydrologic forecasting
Streamflow prediction in ungauged and data-scarce basins using LSTM, Transformer, and ensemble approaches, including loss-function design, process-guided hybrid models, and detection of nonstationarity arising from wildfire and land use change.
Water managers, flood and drought riskSoil carbon measurement and accounting
Soil organic carbon baselines, attainable stock estimation, and evaluation of SOC models used in MRV and crediting protocols, including the representativeness of legacy soil data and the detection limits of proximal and remote sensing for stock change.
Carbon programs, MRV, agricultureHyperspectral and drone sensing
Sensor and mission selection, radiometric, atmospheric, and glint correction, and retrieval of soil, crop, and water quality parameters from VNIR-SWIR imaging spectroscopy and UAV platforms, with explicit assessment of spectral and spatial resolvability.
Agriculture, irrigation districts, sensor vendorsMonitoring and measurement design
Design of sampling frequency, spatial network, and parameter selection to achieve a stated statistical power; fluvial load estimation; and evaluation of whether available observations can support the intended inference.
Watershed programs, research teamsArid-region water, irrigation and salinity
Evapotranspiration and crop water use mapping, root-zone salinity assessment through data assimilation, irrigation with saline and reclaimed water, and water availability analysis in transboundary basins, including the Middle Rio Grande and Hueco Bolson.
Growers, irrigation districts, water utilitiesDecision-support tools and data platforms
Decision-support systems that couple process models with monitoring data: real-time water quality portals, web-based interfaces to legacy models such as HSPF and CE-QUAL-W2, and natural language processing of regulatory documents.
Utilities, agencies, monitoring programsExpert opinion and technical due diligence
Written expert opinion on water quality, hydrologic modeling, and remote sensing evidence, and technical due diligence on Earth observation, agricultural technology, and climate ventures prior to investment or procurement.
Law firms, investors, procurement teamsAbout
Saurav Kumar
I am a civil and environmental engineer (PhD, Virginia Tech) and a licensed Professional Engineer. My work lies at the intersection of Earth observation, machine learning, and water systems modeling, and is organized around the observation-to-decision chain: a measurement acquires value only when it supports a defensible action.
Published work includes deep learning and explainable AI for crop and land cover classification, hyperspectral and UAV-based prediction of soil salinity and soil organic carbon, machine learning for streamflow forecasting and groundwater potential mapping, and analysis of TMDL practice in U.S. water quality regulation.
Current work and publications: waterdmd.info.
Consulting is provided independently through Datadius LLC. It is not offered on behalf of, and carries no endorsement by, any university or professional society named on this page.
Selected work
The published basis for each area
A representative selection. The complete record is available on Google Scholar and ORCID.
Trusting Earth observation and AI in regulatory practice
- Kumar, S., Imen, S., Sridharan, V.K., et al. (2024) Perceived barriers and advances in integrating Earth observations with water resources modeling. Remote Sensing Applications: Society and Environment, 33. doi
- Kumar, S., Imen, S., Ahmadisharaf, E., et al. (2025) Rethinking TMDLs: perspective based on community survey. Journal of Environmental Engineering. doi
- Sridharan, V.K., Kumar, S., Kumar, S.M. (2022) Can remote sensing fill the United States' monitoring gap for watershed management? Water, 14(13). doi
- Sridharan, V.K., Quinn, N.W.T., Kumar, S., et al. (2021) Selecting reliable models for Total Maximum Daily Load development: holistic protocol. Journal of Hydrologic Engineering, 26. doi
- Nunoo, R., Anderson, P., Kumar, S., Zhu, J.-J. (2020) Margin of safety in TMDLs: natural language processing aided review of the state of practice. Journal of Hydrologic Engineering. doi
Streamflow and groundwater forecasting with machine learning
- Dahal, K., Gupta, A., Bokati, L., Kumar, S. (2026) Ensemble streamflow forecasting with diverse loss functions. Applied Soft Computing. doi
- Dahal, K., Sharma, S., Shakya, A., et al., Kumar, S. (2023) Identification of groundwater potential zones in a data-scarce mountainous region using explainable machine learning. Journal of Hydrology, 627. doi
Soil organic carbon baselines and model reliability
- Somenahally, A.C., Bokati, L., Kumar, S. (2025) Estimating soil organic carbon deficits at the continental scale using legacy-data-driven dynamic baseline and attainable projections. Geoderma. doi
- Bokati, L., Somenahally, A., Kumar, S., et al. (2025) Temporal adjustment approach for high-resolution continental scale modeling of soil organic carbon. Scientific Reports, 15. doi
- Bokati, L., Somenahally, A., Kumar, S. (2026) Soil carbon modeling at crossroads: building reliable methods for policy and practice. European Journal of Soil Science, 77(2).
Land cover, crop mapping and explainable AI
- Ebrahimi, S., Kumar, S. (2025) What helps to detect what? Explainable AI and multi-sensor fusion for semantic segmentation of simultaneous crop and land cover land use delineation. IEEE JSTARS. doi
- Ebrahimi, S., Kumar, S. (2025) Semantic segmentation for simultaneous crop and land cover land use classification using multi-temporal Landsat imagery. Remote Sensing Applications: Society and Environment. doi
- Ebrahimi, S., Khorram, M., Neri Barranco, R., et al., Kumar, S. (2025) Thirty years of simultaneous crop and land cover land use maps for the Middle Rio Grande, 1994 to 2024. Scientific Data. doi
Hyperspectral and drone sensing of soils and crops
- Dayal, D., Palmate, S.S., Luera, E.D., Ganjegunte, G.K., Kumar, S. (2025) A spatially aware Bayesian deep learning framework for UAV-based soil salinity prediction. Smart Agricultural Technology. doi
- Khorram, M., Kumar, S., Shrestha, R.K., et al. (2026) Harnessing hyperspectral imaging and machine learning to enhance salinity stress detection in canola. Computers and Electronics in Agriculture. doi
- Ebrahimi, S., Khorram, M., Palmate, S., et al., Kumar, S. (2024) Assessing field scale spatiotemporal heterogeneity in salinity dynamics using aerial data assimilation. Agricultural Water Management. doi
Monitoring design and pollutant load estimation
- Kumar, S., Godrej, A.N., Grizzard, T.J. (2013) Watershed size effects on applicability of regression-based methods for fluvial loads estimation. Water Resources Research. doi
- Kumar, S., Godrej, A.N., Grizzard, T.J. (2015) A web-based environmental decision support system for legacy models. Journal of Hydroinformatics. doi
- Poulose, T., Kumar, S., Ganjegunte, G.K. (2021) Robust crop water simulation using system dynamic approach for participatory modeling. Environmental Modelling and Software. doi
Arid-region water, irrigation and salinity
- Palmate, S.S., Kumar, S., Poulose, T., et al. (2022) Comparing the effect of different irrigation water scenarios on arid region pecan orchard using a system dynamics approach. Agricultural Water Management, 265. doi
- Alger, J., Mayer, A., Kumar, S., Granados-Olivas, A. (2020) Urban evaporative consumptive use for water-scarce cities in the United States and Mexico. AWWA Water Science, 2. doi
- Talchabhadel, R., McMillan, H., Palmate, S.S., et al., Kumar, S. (2021) Current status and future directions in modeling a transboundary aquifer: a case study of Hueco Bolson. Water, 13. doi
How I work
Scope and documentation
Scope, data, methods, and deliverables are defined in writing before work begins. Deliverables document data sources, assumptions, and the conditions under which the conclusions would change.
Contact
Inquiries
Please describe the problem and the decision it informs, together with:
- your organization and role
- the timeline
- the data currently available