An image difference map (Date2 - Date1) of NDVI values shows large negative values in a forest region. Which interpretation is most likely?
AThe forest grew significantly between dates
BVegetation loss occurred (deforestation, fire, disease) causing NDVI to decrease
CAtmospheric conditions improved between the two dates
DThe sensor gain was reduced for the second acquisition
NDVI decreases when vegetation health or cover declines (reduced NIR reflectance relative to red). Large negative NDVI differences in a forest indicate substantial vegetation loss -- from logging, fire, insect damage, or storm damage. Atmospheric and sensor artifacts should have been removed by preprocessing; seasonal phenology is controlled by selecting anniversary dates.
Question 2 True / False
Comparing two images acquired in different seasons is sufficient for detecting land cover change because seasonal vegetation differences average out.
TTrue
FFalse
Answer: False
Seasonal differences (phenology) can overwhelm actual land cover change signals. A deciduous forest photographed in summer (leaf-on, high NDVI) and winter (leaf-off, low NDVI) shows dramatic spectral differences that have nothing to do with land cover change. Reliable change detection typically requires anniversary date images (same time of year) or time-series approaches that model and remove the seasonal cycle.
Question 3 Short Answer
What advantage does time-series change detection (e.g., LandTrendr, BFAST) offer over simple bi-temporal comparison?
Think about your answer, then reveal below.
Model answer: Time-series approaches use all available observations (potentially hundreds of images over decades) to model the temporal trajectory of each pixel, distinguishing gradual trends (forest degradation, urban sprawl) from abrupt events (fire, harvest) and from seasonal noise. Bi-temporal comparison can only detect net change between two dates, missing the timing, duration, and nature of change events. Time-series methods are also more robust to individual noisy observations because they fit models through many data points rather than relying on just two.
Dense time-series analysis transforms change detection from 'did it change?' to 'when, how fast, and what type of change occurred?'