Geospatial analyst working at the intersection of remote sensing, spatial modelling and machine learning.

My background is in Environmental Science (BSc) and Geomatics (MSc). I use spatial data to map change, model risk, and turn satellite imagery into decisions people can act on.

Burnaby, BC · Portfolio · agnesschisale8@gmail.com

Aflatoxin is a carcinogenic fungal toxin that contaminates maize under specific climate conditions. It is invisible and the only current detection method is physical sampling. This is slow, expensive and impossible to scale. It remains a major public health and food security problem across East and West Africa.

This project predicts where and when contamination is likely, using machine learning trained on georeferenced field measurements and satellite climate data. In effect, a weather forecast for crop toxicity.

Deliverables

• Geodatabase integrating agronomic field data with satellite climate data
• Trained machine learning model for aflatoxin risk classification
• Risk maps under current and projected climate scenarios
• Interactive Shiny dashboard for exploring risk by region and year
• Annotated, reproducible code repository
• Technical report

R · Random Forest · Google Earth Engine · ERA5-Land · terra · sf · Shiny

Vegetation change & forest dynamics in western Canada Detected forest disturbance across two western Canadian study areas — BFAST break detection applied to a 20-year MODIS NDVI time series in Alberta, and 33 years of Landsat-derived forest gain and loss in British Columbia. Distinguished stand-replacing disturbance from vegetation recovery. R · terra · sf · bfast · ggplot2

Salmon stream network analysis, Vancouver Island Modelled Pacific salmon spawning habitat by deriving a stream network from a 25 m DEM, modelling dams as network barriers, and tracing accessibility from the ocean. Characterised segments by gradient and Strahler order using linear referencing to separate accessible from blocked habitat. ArcGIS Pro · PostgreSQL/PostGIS · QGIS · Network Analyst

Cartographic modelling of old growth forests, Vancouver Island Classified crown forest stands by seral stage using BEC zone age thresholds, then compared old growth percentages against provincial targets by Landscape Unit. Revealed spatial patterns of old growth deficit reflecting historical logging pressure. SQL · PostgreSQL · ArcGIS Pro · Python

Further geospatial projects completed during my MSc in Geomatics.

• least_cost_path_modelling — Grizzly bear movement corridors across the Yellowhead Bear Management Area, using a weighted cost surface built from slope, land cover and road proximity. GDAL · QGIS · ArcGIS Pro
• riparian_area_management — Riparian reserve and management zones in the Nahmint Valley, derived from a high-resolution DEM and BC stream classification guidelines. ArcGIS Pro · Python
• geocoding_and_spatial_data — Geocoded Vancouver street trees using a custom address locator built from a regional road network, and calculated park tree density citywide. ArcGIS Pro · PostgreSQL

Languages R · Python · SQL

Geospatial analysis Spatial modelling · Hydrological modelling · Network analysis · Cost surface modelling · Spatial interpolation · Geocoding · Cartography

Remote sensing Google Earth Engine · Change detection · Time series analysis (BFAST) · NDVI & vegetation indices · LiDAR processing · MODIS · Landsat · ASTER

Machine learning Random Forest · Classification & regression

Software & databases ArcGIS Pro · QGIS · GDAL · PostgreSQL / PostGIS · RStudio

Tools & workflow Git · GitHub · Shiny · tidyverse · tidymodels · terra · sf