Data Science — Kathiravan Meeran

Geospatial ecosystem mapping — Planetary‑scale mapping of ecosystems and canopy structure.

Time series analysis — High‑frequency time‑series: cycles, gaps, forecasts.

Bayesian inverse modelling — Bayesian inverse models for city‑scale emissions.

Machine learning — ML for classification, regression and pattern discovery.

Geospatial

Earth Engine workflows
DEM + hydrology layers
Remote sensing indices

Time‑Series

QA/QC + gap filling
Seasonal/trend modelling
Forecasting + uncertainty

Bayesian

Hierarchical models
Inverse modelling
PPCs + priors

Machine Learning

Supervised/unsupervised
Feature engineering
Model explainability

Geospatial Data

Modern environmental research utilizes remote sensing and satellite imagery. Google Earth Engine provides planetary‑scale analysis by combining a multi‑petabyte catalog of satellite imagery and geospatial datasets. We can use it to detect changes, map trends, and quantify differences on the Earth’s surface (earthengine.google.com).

I did bachelor degree in geospatial analysis to map forest types, estimate vegetation indices, and derive variables like the topographic, weather and climate parameters.

Mapping ecosystems with satellite data — Mapping ecosystems and canopy structure from multi‑spectral imagery.

Topographic wetness index from DEM — Topographic wetness index derived from DEMs supports growth and iWUE analyses.

Terrain modelling workflows — Terrain modelling and hydrological layers in geospatial workflows.

Time‑Series Analysis

Time‑series data are observations collected over time. In ecology, time‑series analysis uncovers patterns in population dynamics, species interactions, and ecosystem responses. Analysts examine autocorrelation, seasonality, and long‑term trends, using methods such as ARIMA, decomposition, and smoothing (example).

We apply models to greenhouse‑gas concentrations, fluxes, and isotope data: gap‑filling, diel/seasonal cycles, and emission forecasting.(example).

General time series patterns — Time‑series features: trend, seasonality and autocorrelation in environmental signals.

Soil respiration time series — Soil respiration dynamics under warming and drought treatments.

Structural equation model of grassland carbon dynamics — SEM linking roots, moisture, and C fluxes in grasslands.

Bayesian Modelling

Bayesian statistics treats parameters as random variables and updates prior beliefs with observed data to obtain posterior distributions. This allows direct probability statements and inclusion of prior knowledge. Frequentist methods treat parameters as fixed and base inference on the likelihood of the data .

Bayesian techniques are especially useful for small samples or complex models; they provide intuitive uncertainty quantification. I use hierarchical models and inverse models to partition CO2 and CH4 sources with isotope constraints.

Inverse model for CO2 emissions — Inverse modelling couples transport with priors to infer city‑scale emissions.

Conceptual Bayesian inference — Bayesian workflow: priors, likelihood, posterior, and predictive checks.

Code-generated image for Bayesian model — Code‑generated visual highlighting posterior distributions and uncertainty.

Machine Learning

Machine learning enables pattern discovery and prediction. Random forests combine many decision trees via random sampling and feature selection; predictions are made by averaging or voting. Gradient boosting builds models sequentially, correcting previous errors; XGBoost adds regularisation. In unsupervised learning, k‑means partitions data into k clusters by assigning points to the nearest centroid.

We classify land cover from satellite data, predict tree growth responses to climate variables, and extract patterns from multivariate ecological datasets.

Machine learning overview — Supervised and unsupervised algorithms applied to ecological datasets.

K-means clustering example — Clustering for pattern discovery and site classification.

Neural network schematic — Neural networks for non‑linear relationships and feature interactions.

Workflow

Acquire

Sensors, satellites, towers, chambers; metadata‑rich ingestion.

Process

QA/QC, harmonisation, gap‑filling, feature engineering.

Model

Bayesian inference, inverse models, ML ensembles.

Report

Interactive visuals, uncertainty communication, reproducible outputs.