Remote sensing · environmental science

Illustrated overview of remote sensing workflows: a satellite and an aircraft collect spectral data over a tailings pond and forested hills, a handheld spectrometer measures soil on site, a soil profile shows its layers, and the data flows into an AI model that produces maps and estimates.

Understanding Earth through remote sensing.

I’m Dr. Yaron Ogen, a remote sensing scientist. I connect laboratory, field, airborne and satellite observations to make environmental patterns easier to understand.

Explore my research Discuss a collaboration
From spectral measurements to Earth observation.
Soil spectroscopyHyperspectral imagingGeospatial modellingEarth observationEnvironmental scienceMachine learning

03 / Areas of expertise

Geospatial science,
put to work.

Services and methods across environmental monitoring, image analysis and spatial modelling.

01

Forest monitoring & fire assessment

Assess fire impacts and track forest condition before and after disturbance.

Burn severity west of Jerusalem · Judean Hills wildfire, August 2021
  • Forest fire severity assessment
  • Burn scar mapping
  • Post-fire recovery monitoring
  • Forest loss and gain assessment
  • Forest disturbance detection
  • Forest canopy monitoring
02

Time series & change detection

Use repeat observations to identify environmental change and long-term patterns.

Field-mean NDVI for a field in Saxony-Anhalt, from Sentinel-2 imagery, 2018–2026.
  • Satellite time-series analysis
  • Multi-temporal image analysis
  • Environmental change detection
  • Historical land surface monitoring
  • Long-term trend assessment
  • Automated change detection
03

Vegetation & ecosystem monitoring

Characterise vegetation condition, seasonal cycles and ecosystem response.

NDVI maps of vineyards. Colours run from 0 (purple) to 0.7 (red).
  • Vegetation health assessment
  • Vegetation phenology
  • Seasonal vegetation dynamics
  • Drought and vegetation stress monitoring
  • Spectral vegetation indices
  • Ecosystem condition assessment
04

Soil & environmental modelling

Combine spectroscopy, Earth observation and field measurements for soil and environmental analysis.

Modelled pH classes (A–E) across the area. Bars show the share of the area in each class.
  • Digital soil mapping
  • Soil property prediction
  • Soil spectroscopy
  • Machine learning-based soil modelling
  • Environmental modelling
  • Satellite, field and laboratory data integration
05

Land use & land cover

Map land cover and land use, and analyse how they change over time.

Forest loss north of Manaus, Amazonas, 2018–2026, from Sentinel-2 imagery.
  • Land use classification
  • Land cover mapping
  • Land cover change detection
  • Object identification
  • Feature extraction
  • Machine learning classification

04 / Research & projects

Questions pursued
across scales.

Project manager · Israel

National soil spectral library

Managed the development of a national spectral library for the soils of Israel at Tel Aviv University’s Remote Sensing Laboratory, funded by Israel’s Ministry of Agriculture.

Soil spectroscopy · Soil mapping

Project manager · Israel

National soil spectral library

Crew operating a tracked Geoprobe soil-coring rig between rows of trees in an orchard, with the drill mast raised above the ground.
Soil coring in an orchard

Managed the development of a national spectral library for the soils of Israel at Tel Aviv University’s Remote Sensing Laboratory, funded by Israel’s Ministry of Agriculture.

The library supports agriculture, environmental protection and planning, and serves students and researchers for teaching and study. In its first year, the work focused on locating soil sources across Israel, including soil survey archives and collections, bringing them to the laboratory, and cataloguing and archiving them. The samples were measured spectrally with standard methods used worldwide.

About 4,500 soil samples have been collected so far, and around 3,000 of them have been measured spectrally. The measured samples are archived digitally with their chemical and physical data and their spatial position. Selected soil properties have been mapped and statistical models built. The second year will add samples to the library and develop a web portal where end users can retrieve the archived information.

Soil spectroscopy · Soil mapping

Research · Dead Sea, Israel

Psoriasis — hyperspectral correlations with biochemical markers

I used a portable spectroradiometer (350–2,500 nm) to measure psoriatic and healthy skin on the elbow of eight participants, before treatment and four weeks later at the Dead Sea. The spectra were then compared with biochemical markers sampled from the skin.

Hyperspectral imaging · Dermatology · Biochemical markers

Research · Dead Sea, Israel

Psoriasis — hyperspectral correlations with biochemical markers

Line graph of reflectance value against wavelength from 350 to 2,500 nm. The red line shows psoriatic skin and the blue line shows healthy skin from the same participant, both before treatment. The main difference is in the visible range, 350 to 610 nm.
Psoriatic vs healthy skin before treatment; main difference in the visible range

I used a portable spectroradiometer (350–2,500 nm) to measure psoriatic and healthy skin on the elbow of eight participants, before treatment and four weeks later at the Dead Sea. The spectra were then compared with biochemical markers sampled from the skin.

In the graph, the largest difference between psoriatic and healthy skin is in the visible range (350–610 nm), where the oxyhemoglobin absorption bands at 418, 542 and 577 nm are deeper in psoriatic skin. Separately, the absorption depth near 1,777 nm and a derivative-based index (1,821 nm minus 1,727 nm) showed an inverse relation with the biochemical markers, suggesting lesion severity could be estimated without costly laboratory tests. These are preliminary results.

Hyperspectral imaging · Dermatology · Biochemical markers

Postdoctoral research · Erdenet, Mongolia

ADRIANA — Erdenet mine, Mongolia

Contributed to research developing airborne spectral approaches to monitor and predict geochemical composition at the Erdenet Cu–Mo mine.

Hyperspectral imaging · Geochemical modelling

Postdoctoral research · Erdenet, Mongolia

ADRIANA — Erdenet mine, Mongolia

Panoramic view across the open-pit Erdenet mine site in Mongolia.
View across the Erdenet mine site

Contributed to research developing airborne spectral approaches to monitor and predict geochemical composition at the Erdenet Cu–Mo mine.

ADRIANA (Airborne Spectral Detection of Reusable Industrial Materials in Waste Rock Piles and Tailings) combines modern remote sensing with classical exploration methods to quantify valuable materials in tailings and waste rock piles and to develop innovative processing methods. The studies are carried out at Erdenet in Mongolia together with German partners, coordinated by G.E.O.S. and funded by the German Federal Ministry of Education and Research.

Project announcement, G.E.O.S.

The project concluded with two workshops in October 2022, in Erdenet and Ulaanbaatar. A detailed resource model was built for a copper tailings facility of about 20 km² in Erdenet, using historical processing data, historical satellite imagery and current hyperspectral data, and it was verified by drilling. Researchers from Martin Luther University Halle-Wittenberg, my team, showed for the first time that low metal concentrations can be measured with high accuracy from hyperspectral remote sensing data. Results were published in several scientific papers.

Project completion, G.E.O.S. Hyperspectral imaging · Geochemical modelling

Soil monitoring · Mbeya, Tanzania

Soil monitoring in Tanzania

Fieldwork in a flooded rice field in Tanzania. Additional project details and findings can be added when ready to share.

Soil monitoring · Tanzania

Soil monitoring · Mbeya, Tanzania

Soil monitoring in Tanzania

Dr. Yaron Ogen standing in a flooded rice field during soil-monitoring fieldwork in Tanzania.
Soil-monitoring fieldwork in Tanzania

Fieldwork in a flooded rice field in Tanzania. Additional project details and findings can be added when ready to share.

Soil monitoring · Tanzania

Hackathon · Helsinki, Finland

Beyond Dead Wood Hackathon — 3rd place

I took part in the Beyond Dead Wood Hackathon, run through Ultrahack, with team Ecosonic. We won 3rd place and a prize package.

Hackathon · Team competition

Hackathon · Helsinki, Finland

Beyond Dead Wood Hackathon — 3rd place

A team member working at a table during the hackathon, with a laptop showing a colourful map and a notebook with handwritten notes beside it.
Working on the team analysis during the hackathon

I took part in the Beyond Dead Wood Hackathon, run through Ultrahack, with team Ecosonic. We won 3rd place and a prize package.

The hackathon was organised through Ultrahack. The challenge details are on the event page linked below.

The photos show the prize sign and the team working at the table, with a laptop displaying a colourful map beside handwritten notes.

Event page, Ultrahack

05 / Highlight publications

Four papers, and what is new.

Selected articles: the research, what was novel in the approach, and how the results stand out.

Flowchart of the workflow. Tailing spectra are preprocessed and split into training and test sets. Two regression models (PLSR and RFR) and five classifiers are fitted. The best regression R² is compared with the simulated threshold R²t: above it, the regression result is reported; below it, the classification result is reported.
Workflow from the article (Figure 1). The R²t simulation sets the decision between a regression result and a classification result.

2022 · European Journal of Remote Sensing 55, 453–470 · Open access

A novel method for predicting the geochemical composition of tailings with laboratory, field and hyperspectral airborne data

The research

Tailings from a porphyry copper basin near Erdenet, Mongolia, were measured in the laboratory, in the field and from airborne HySpex imagery. The aim was to predict about 27 geochemical, mineralogical and textural properties, and to decide for each one whether a regression or a classification model is trustworthy. Model performance drops from lab to field to airborne data, which makes the three scales hard to compare directly.

What was novel

  • A threshold instead of a fixed cut-off. A regression is accepted only if its R² exceeds a threshold (R²t) set for its measurement level: 0.58 for lab, 0.40 for field and 0.31 for airborne data.
  • The threshold comes from a simulation. Test sets of 34, 22 and 14 samples were simulated with 100%, 90% and 80% correctly classified, and each case was repeated 30 times. The extra uncertainty of field and airborne data was set at 10% and 20% above the lab error, following earlier studies.
  • Each model is judged against its own scale. Lab, field and image R² are compared with the performance potential of their own measurement level, not with each other.
  • Regression and classification in one decision. Every property gets a result: a regression where R² clears the threshold, otherwise a classification.

How the results stand out

Classification performance stayed steady across the three scales, while regression dropped at each step. Field Cu reached R² = 0.42, just above the field threshold, so it could be reported as a regression. Iron oxides (FeO and Fe₂O₃) fell below the threshold even in the lab, because their concentrations varied little, and were classified instead.

The threshold, and how I chose the method for each property

  1. Preprocess. Spectra were tested as reflectance and as first derivative, and the better preprocessing was kept.
  2. Two regressions. Partial least squares regression (PLSR) and random forest regression (RFR) were fitted on an 80:20 train/test split. Both used the same random state, so the split is identical.
  3. Pick the best regression. The model with the highest R² was kept. For the small airborne dataset, cross-validated R² was used instead.
  4. Apply the threshold. If that R² is higher than R²t for the measurement level, the property is reported as a regression. If not, it moves to classification.
  5. Pick the best classifier. k-nearest neighbour, logistic regression, support vector machine, random forest and a neural network were compared, and the one with the highest precision was kept.
Read article
Soil and compost samples laid out in sample trays. The samples range from pale sand to dark brown, reflecting different soil types and organic matter contents.
Soil and compost sample series prepared for the point and imaging spectral measurements.

2018 · Geoderma 321, 100–109

Evaluating the detection limit of organic matter using point and imaging spectroscopy

The research

Five soil types from Israel were mixed with increasing amounts of compost and measured with a point spectrometer (ASD) and an imaging spectrometer (HySpex). The aim was to find the lowest organic matter content a spectral method can detect reliably, and how clay and the starting OM of each soil change that limit.

What was novel

  • The detection limit as a soil-specific property. A soil spectral detection limit (SSDL) is estimated for each soil, rather than assuming one OM model fits all soils. The same OM content gives different spectra in different soils, so a single model is misleading.
  • A SIW/CIW ratio method. For each sample, a soil-indicative wavelength (500–600 nm) is divided by a compost-indicative wavelength (800–1000 nm). A linear fit of this ratio against added compost gives the detection limit, taken as the blank mean plus three standard deviations.
  • Two sensors compared. The SSDL values from the point and imaging sensors are correlated, so one sensor’s detection limit can be estimated from the other’s.

How the results stand out

Soil types differed in their SSDL, and this tracked their starting OM and clay content. The paper recommends classifying soils before modelling OM, and flagging samples below the detection limit. A generic imaging model gives only a rough OM estimate, because spectra are grouped by soil class rather than by actual OM content.

Read article
Four scatter plots of 1,480 soil samples placed by spectral angle (SAM, horizontal axis) and spectral gradient (vertical axis). Each plot is coloured by one property: CaCO3, organic matter, sand and clay content. Samples with similar spectra sit close together, which is the basis for forming clusters.
Samples in SAM–gradient space, coloured by soil property (Figure 4 of the article). The same space was used to group the soils into clusters.

2019 · Geoderma 340, 175–184

Cluster-based spectral models for a robust assessment of soil properties

The research

I analysed 1,480 soil samples from Israel’s climate zones, measuring calcium carbonate, organic matter (OM), clay and sand alongside their VNIR–SWIR spectra. Generic models that treat all soils alike predicted OM poorly, so the study asks whether grouping soils by spectral similarity first gives more robust models.

What was novel

  • Clustering in spectral space. Samples were grouped with the Spectral Angle Mapper (SAM), a spectral gradient and k-means, instead of by soil type. Soil type needs expert description of diagnostic horizons, which is subjective and not always available.
  • Cluster models tested against a generic model on the same samples.
  • An OM detection-limit correction applied to the predicted OM values of each cluster.

How the results stand out

Generic models reached R² = 0.71 for CaCO₃, 0.67 for clay and 0.65 for sand, but only 0.34 for OM. With clustering, the combined results reached R² = 0.82–0.89 across the properties. The detection-limit correction then lifted OM to R² = 0.82–0.93 within clusters and 0.89 for the combined data.

Read article
Map of Europe shaded by tropospheric NO2 concentration, from low (blue) to high (red), with the highest values over northern Italy and the Po Valley. Country names are marked for Spain, France, Germany and Italy.
Tropospheric NO₂ distribution over Europe (Figure 1 of the article).

2020 · Science of the Total Environment 726, 138605 · Short communication · Single author

1,056 citations on Google Scholar (checked 20 Oct 2026)

Assessing nitrogen dioxide (NO₂) levels as a contributing factor to coronavirus (COVID-19) fatality

The research

I tested whether long-term exposure to nitrogen dioxide helps explain where COVID-19 deaths were concentrated early in the pandemic. The study covered 66 administrative regions in Italy, Spain, France and Germany, with 4,443 deaths reported by 19 March 2020. The purpose of publishing it was to shift attention from pre-existing diseases as the main explanation of mortality towards environmental factors, which deserve further research.

What was novel

  • Three datasets in one analysis. Satellite NO₂ (Sentinel-5P TROPOMI, 5.5 km), vertical airflow at 850 hPa and regional death counts were combined at administrative-region level, rather than using national totals.
  • Airflow as a dispersal test. Downward airflow keeps NO₂ close to the ground, while upward airflow disperses it. Reading the two together flags regions where high NO₂ is held near the surface.

How the results stand out

Five regions (four in northern Italy and Madrid) account for 3,487 deaths, 78% of the total. They had the highest NO₂ levels, roughly 177–294 µmol/m², and downward airflow. About 83% of deaths fell in regions where NO₂ exceeded 100 µmol/m².

Limitations and the debate

The analysis is descriptive. It has no formal statistical test, no adjustment for age, population density or other confounders, and it uses absolute death counts rather than rates. The “long-term” window is only two months. A commentary by Chudnovsky raised these points, and I responded in follow-up letters in the same journal.

Read article

06 / Peer-reviewed publications

Papers, with
links to the source.

All 15 peer-reviewed publications listed in my CV. Each article links to its DOI record.

2022

A novel method for predicting the geochemical composition of tailings with laboratory field and hyperspectral airborne data using a regression and classification-based approach

Y. Ogen, M. Denk, C. Glaesser, H. Eichstaedt · European Journal of Remote Sensing, 55, 453–470

Read article
Hyperspectral
geochemistry
2022

Aggregate size distribution of arid and semiarid laboratory soils (<2 mm) as predicted by VIS–NIR–SWIR spectroscopy

E. Ben-Dor, N. Francos, Y. Ogen, A. Banin · Geoderma, 416, 115819

Read article
Soil
spectroscopy
2021

Quantification of the spectral variability of ore-bearing granodiorite under supervised and semisupervised conditions: an upscaling approach

Y. Ogen, M. Denk, C. Glaesser, H. Eichstaedt, R. Kahnt, R. Loeser, R. Suppes, R. Chimeddorj, T. Tsedenbaljir, U. Alyeksandr, et al. · Journal of Spectroscopy, 2021, e2580827

Read article
Mineral
spectroscopy
2021

Spectral assessment of organic matter with different composition using reflectance spectroscopy

N. Francos, Y. Ogen, E. Ben-Dor · Remote Sensing, 13, 1549

Read article
Organic
matter
2020

Quantitative estimation of clay minerals in airborne hyperspectral data using a calibration field

H. Eichstaedt, T. Tsedenbaljir, R. Kahnt, M. Denk, Y. Ogen, C. Glaesser, R. Loeser, R. Suppes, U. Alyeksandr, T. Oyunbuyan, et al. · Journal of Applied Remote Sensing, 14, 034524

Read article
Airborne
hyperspectral
2020

Assessing nitrogen dioxide (NO₂) levels as a contributing factor to coronavirus (COVID-19) fatality

Y. Ogen · Science of the Total Environment, 726, 138605

Read article
Environmental
science
2020

Response to the commentary by Alexandra A. Chudnovsky on “Assessing nitrogen dioxide (NO₂) levels as a contributing factor to coronavirus (COVID-19) fatality”

Y. Ogen · Science of the Total Environment, 740, 139239

Read article
Author
response
2020

Response to the commentary by Pisoni, E. and Van Dingenen, R. on “Assessing nitrogen dioxide (NO₂) levels as a contributing factor to coronavirus (COVID-19) fatality”

Y. Ogen · Science of the Total Environment, 738, 140672

Read article
Author
response
2020

An integrated methodology using open soil spectral libraries and Earth observation data for soil organic carbon estimations in support of soil-related SDGs

N. Tziolas, N. Tsakiridis, Y. Ogen, E. Kalopesa, E. Ben-Dor, J. Theocharis, G. Zalidis · Remote Sensing of Environment, 244, 111793

Read article
Soil carbon
Earth observation
2019

Removing the moisture effect on soil reflectance properties: a case study of clay-content prediction

Y. Ogen, S. Faigenbaum-Golovin, A. Granot, Y. Shkolnisky, N. Goldshleger, E. Ben-Dor · Pedosphere, 29(4), 421–431

Read article
Soil
Moisture
2019

Cluster-based spectral models for a robust assessment of soil properties

Y. Ogen, J. Zaluda, N. Francos, N. Goldshleger, E. Ben-Dor · Geoderma, 340, 175–184

Read article
Soil
Spectroscopy
2019

Preparing a soil spectral library using the Internal Soil Standard (ISS) method: influence of extreme different humidity laboratory conditions

S. Chabrillat, A. Gholizadeh, C. Neumann, D. Berger, R. Milewski, Y. Ogen, E. Ben-Dor · Geoderma, 355, 113855

Read article
Soil spectral
Library
2018

Evaluating the detection limit of organic matter using point and imaging spectroscopy

Y. Ogen, C. Neumann, S. Chabrillat, N. Goldshleger, E. Ben-Dor · Geoderma, 321, 100–109

Read article
Imaging
Spectroscopy
2017

3D spectral analysis in the VNIR–SWIR spectral region as a tool for soil classification

Y. Ogen, N. Goldshleger, E. Ben-Dor · Geoderma, 302, 100–110

Read article
Soil
Classification
2016

Integration of hyperspectral shortwave and longwave infrared remote-sensing data for mineral mapping of Makhtesh Ramon in Israel

G. Notesco, Y. Ogen, E. Ben-Dor · Remote Sensing, 8, 318

Read article
Mineral
Mapping

Publication titles and author details follow the CV. DOI links were matched against Crossref records.

02 / About me

Science is most useful
when it helps us see clearly.

I’m Dr. Yaron Ogen, a remote sensing scientist with a Ph.D. in remote sensing and soil spectroscopy and more than a decade of experience.

My path began with geography and the human environment at Tel Aviv University: a B.A. cum laude (2007), an M.A. in remote sensing (2012) and a Ph.D. in remote sensing and soil spectroscopy (2018). I then worked as a postdoctoral researcher at Martin Luther University Halle-Wittenberg, developing airborne spectral methods for the Erdenet copper–molybdenum mine in Mongolia. Until September 2026 I was Chief Research Officer and co-founder of Xsupra GmbH in Halle (Saale), an applied geospatial technology company.

My research has been carried out in Israel, Germany, Tanzania and Mongolia, and has used remote sensing to look at the environment from several angles. In mining areas, I use laboratory, field and airborne hyperspectral data to analyse and predict the geochemistry of mine tailings. In soils, I study soil properties from their spectra, from soil classification and organic matter to large spectral libraries. For forests and fire, I analyse forest condition and fire impacts from satellite data. For the atmosphere and health, I analyse satellite air-pollution data together with weather and public health records.

Across these layers I see the environment as one connected system. Rock, soil, vegetation and air each leave a spectral and spatial signature, and a deep geospatial understanding lets many of these aspects be explained together, from the ground beneath us to the air above it.

My methods combine machine learning with geostatistical approaches, including regression and classification models, and preprocessing techniques such as principal component analysis (PCA) that extract robust signals from complex spectral and satellite data. More recently, I have also begun working with agentic AI, in which intelligent systems carry out multi-step analytical workflows, to deepen our understanding of the environment.

15 peer-reviewed papers10+ years in remote sensingHE / EN Hebrew & English

Career

Research translated
into practice.

Experience across applied geospatial technology and academic research.

Jul 2024 — Sep 2026

Chief Research Officer · Co-founder

Xsupra GmbH · Halle (Saale), Germany

Led scientific and research activities, developing data-driven methods and Python workflows for agricultural applications, satellite and drone imagery, modelling and geospatial analysis.

Jun 2019 — Dec 2022

Postdoctoral Researcher

Martin Luther University Halle-Wittenberg · Germany

Worked on the BMBF-funded ADRIANA project to develop airborne spectral approaches for monitoring and predicting geochemical composition at the Erdenet Cu–Mo mine in Mongolia.

2012 — 2018

Ph.D. · Remote Sensing

Tel Aviv University · Israel

Research focused on soil spectroscopy in the visible, near- and shortwave-infrared for precise soil mapping.

07 / Media coverage

In the press,
in three languages.

Media coverage of my 2020 study on nitrogen dioxide and COVID-19 fatality. Each link opens the original article in a new tab.

English

Martin Luther University Halle-Wittenberg, press release

pressemitteilungen.pr.uni-halle.de

Read coverage

German

Spanish

08 / Contact

Have a question
worth exploring?

For consulting, research collaborations and geospatial projects, send a note.

Your email app will open with your message addressed to Yaron. No form data is sent to this website.

Remote sensing · Soil spectroscopy · Geospatial analyticsHalle (Saale), Germany