While satellite imagery provides a powerful tool for ecologists studying the impact of human activity or climate change, to understand the pace of change, much older photographs taken prior to the availability of satellite data, are needed.
Recently, scientists have begun to use old photographs, some going back to the mid 19thcentury, to classify landscape change.
Challenges of Comparing Satellite Imagery to Historical Photographs
There are challenges when comparing satellite images to oblique photographs in particular.
The main challenge is not only the perspective is different but pixel size creates difficulties in making measurements when it comes to land-based photographs. When someone takes a oblique-level picture, pixels near the camera cover a much smaller area than pixels that are more distant.
Reconciling these differences in spatial scale is critical if a broader classification is to be made that can then be compared to satellite imagery.
Free weekly newsletter
Fill out your e-mail address to receive our newsletter!
By entering your email address you agree to receive our newsletter and agree with our privacy policy.
You may unsubscribe at any time.
Despite this challenge, many past images of landscapes can be geo-referenced and classified through comparison with modern images. These photographs can then be compared to classification of satellite-based images, such as from Landsat, to correct for and adjust spatial resolution.
Once that is completed, the classification of vegetation is possible which can then be compared using old and modern photographs (a method known as repeat photography) as well as satellite images.
Historical photos from the Canadian Rockies show changes in forests
Recent work by scientists has resulted in determining that in the Canadian Rockies, over the last 100 years, forested landscapes grew from about 40% of the total area to 52%, while alpine meadows (16% to 9% today) and swamps (5% to 4% today) declined.
This is because warming temperatures have enabled the tree line to creep upwards, resulting in more forested landscape but also a more homogeneous landscape with declining other land cover types.[1]
In the research paper related to this study, the scientists did find that oblique photographs often showed key and important differences from satellite imagery. Mainly, narrow landscape features were often identified in the photographs and there were higher estimates for the proportions of rock surfaced identified in comparison to satellite imagery.
This might not be an issue, as smaller features and variation from satellite imagery might be expected, given that photographs can capture more detail in near ranges.[2]
The paper highlights the use of manual, supervised classification that enabled land cover to be determined in black and white photographs that generally have less variation in spectral data to enable more typical automated or unsupervised machine learning classification methods.
Methods for Classifying Historical Aerial Photographs
Methods have also been developed to classify historical aerial photographs so that they can be utilised to compare and classify land cover across different regions.
The use of deep convolutional neural networks has been applied to align landscapes in older aerial photographs that are then classified so that land cover is identified. This has enabled a useful, historical-based method for scientists to compare long-term vegetation change across wide landscapes.[3]
Similarly, aerial photography has been used and combined with recent imagery to give a long-term land cover change assessment in the United States.
In Colorado’s Northern Front Range, it was determined that forest cover has increased since the 1930s, similar to the Canadian Rockies; however, there was widespread variability and recent trends suggest declines in forest cover.
Furthermore, 14.3% of land cove has experienced fire damage since the 1970s. Fire activity is a growing problem in the West and rising temperatures and diseased trees may make much of the land cover susceptible to major fires in the coming decades.[4]
What recent research has shown is that old photographs, both older aerial photographs as well as oblique, land-level photographs can be very useful in demonstrating land cover change over long time horizons.
For scientists, this gives us a powerful tool to better document the pace of change experienced due to climatic and other anthropogenic factors. The challenge is to now expand coverage so that more regions of the globe can be covered for longer-term land cover classification, allowing us to better assess at a global scale how much the planet has changed.
References
[1] For more on recent work comparing old photographs with satellite imagery, see: https://rsecjournal.blog/2018/12/03/different-viewpoints-same-landscape-do-land-based-photos-and-landsat-imagery-paint-the-same-picture/.
[2] For more on the methods and variation between satellite imagery and oblique photographs, see: Fortin, Julie A., Jason T. Fisher, Jeanine M. Rhemtulla, and Eric S. Higgs. “Estimates of Landscape Composition from Terrestrial Oblique Photographs Suggest Homogenization of Rocky Mountain Landscapes over the Last Century.” Edited by Ned Horning and Jian Zhang. Remote Sensing in Ecology and Conservation, December 2, 2018. https://doi.org/10.1002/rse2.100.
[3] For more on classification of older photographs and imagery using convulsion neural networks and filtering methods, see: Ratajczak, Remi, Carlos Fernando Crispim-Junior, Elodie Faure, Beatrice Fervers, and Laure Tougne. “Automatic Land Cover Reconstruction From Historical Aerial Images: An Evaluation of Features Extraction and Classification Algorithms.” IEEE Transactions on Image Processing28, no. 7 (July 2019): 3357–71. https://doi.org/10.1109/TIP.2019.2896492.
[4] For more on land cover change in Colorado and using historical aerial imagery, see: Rodman, Kyle C., Thomas T. Veblen, Sara Saraceni, and Teresa B. Chapman. “Wildfire Activity and Land Use Drove 20th-Century Changes in Forest Cover in the Colorado Front Range.” Ecosphere10, no. 2 (February 2019): e02594. https://doi.org/10.1002/ecs2.2594.
