A webinar series exploring new technologies that support future action for our environment.
Manaaki Whenua lead the research and development of remote sensing of the terrestrial environment in New Zealand. Working in this space since the 1970’s, it is no surprise that our research teams are at the forefront of developing and implementing cutting edge technologies and methods to monitor and better understand the environment.
We are excited to bring you a mini-series of four webinars over two days to showcase some of the recent developments in remote sensing technologies and how these are applied to answer questions for policy-makers, councils and landowners.
Register to attend the sessions using the individual links below.
Remote sensing 101
This talk will set the scene for the Manaaki Whenua short seminar series by introducing the basic remote sensing concepts. Remote sensing is no longer just a research area providing stunning images of the earth’s surface. Nowadays it is routinely used to identify, map, or otherwise quantify information about our changing world. This talk will cover the basic concepts of remote sensing, looking at
- the various technologies used to create images, their characteristic and roles
- platforms used to carry remote sensing instrumentation
- analysis techniques and approaches to extracting information
Mapping winter forage crops from time series satellite imagery: supporting decision makers and policy planners
Tuesday 3 May, 2pm
Presented by Stella Belliss.
A brief history of our winter forage mapping projects and methodology improvements, description of our current methodology steps, and some results from recent work.
Reaching into the past: deep learning and historic aerial imagery
Wednesday 4 May, 10:30am
Presented by Brent Martin.
Deep learning has revolutionised the field of computer vision, enabling imagery to be analysed quickly and efficiently to extract valuable data. At Manaaki Whenua we are increasingly using deep learning to extract features from aerial and satellite imagery for generating map layers such as land cover and tree crowns. As increasing amounts of historic aerial imagery are digitised, it becomes practical to use deep learning to extract features for snapshots in time, allowing changes to be detected and trends to be analysed.
Whilst deep learning is a highly effective approach, it is not a magic “black box”, and every new domain brings unique challenges. In one such study, we have used deep learning to extract building footprints from historic aerial images taken in the 1940s and 1980s. Whilst there is excellent imagery and building footprint information for recent years, the historic (black and white) aerial imagery is of much poorer quality, and there is no “ground truth” to train from. In this talk we show how deep learning’s ability to train on imperfect data was harnessed to overcome these limitations and achieve good results.
Geospatial landslide modelling for targeted erosion control
Wednesday 4 May, 2pm
Presented by Raphael Spiekermann and Hugh Smith.
Shallow landslides are triggered in large numbers in New Zealand’s pastoral hill country due to steep slopes, weak sedimentary rocks, and relatively frequent high magnitude rainfall events. Landslide erosion generates sediment that impacts freshwater and coastal receiving environments. Establishment of trees on slopes is therefore an important method of erosion control, since trees can greatly increase slope stability. Knowing where landslides are likely to occur in the future allows erosion control to be targeted more cost-effectively.
Raphael Spiekermann and Hugh Smith have employed machine learning algorithms using large inventories of shallow landslide scars and data on environmental factors influencing landslide occurrence to predict spatial patterns in landslide susceptibility. A range of data were used to cater for both region-wide modelling as well as farm-scale assessments that use LiDAR products to account for the influence of individual trees in silvopastoral environments. Moreover, landslide connectivity models were developed to quantify the probability of landslide-generated sediment reaching streams. We will present on how these models can be used to prioritize erosion mitigation at a range of spatial scales to reduce soil loss and delivery of fine sediment to stream networks.