How Well do Land Classification Algorithms Detect Community Structure Patterns in Forest Ecosystems?

Last fall I had a meeting with Jean Burns and her former undergraduate student, India Johnson, about using remote sensing to study invasive species.  Among various options, we discussed using artificial intelligence methods to identify and demarcate abundance and distribution patterns of invasive species.  I had already had some experience with semi-automatic classification of land cover using the SCP Plug-in module for QGIS (Congedo, Luca, (2021). Semi-Automatic Classification Plugin: A Python tool for the download and processing of remote sensing images in QGIS. Journal of Open Source Software, 6(64), 3172, https://doi.org/10.21105/joss.03172), and I hadn't had much success with identifying patterns of distribution of the dominant tree species in the north woodlot of CWRU's University Farms.  With this post, I will provide an initial assessment of the limitations of land classification methods when applied to forested ecosystems.

For this assessment, I am using two different sources of multispectral imagery (Figs. 1 and 2).  Most of the UAV images are from a Sentera Double 4K Multispectral Sensor with  RGB (narrow band red, green, and blue bands) and NIR (red edge and near infrared bands) cameras. At 150 m, these images have a pixel resolution of 3 cm and the raster has about 240 million pixels.  I have recently discovered the availability of high resolution, multispectral images from Planet (Planet Application Program Interface: In Space for Life on Earth. San Francisco, CA. https://api.planet.com).  The image in Fig. 2 contains about 500,000 pixels with a resolution of 3m.  There are obvious patterns of land cover in both images, and there appear to be discernible patterns of distribution of different tree species in the high resolution image of the north woodlot (Fig. 1). 

Figure 1.  Ortho-mosaic image of the north woodlot of University Farms on May 26, 2020 based on 136 individual images from the RGB camera of the Sentera Double 4K Multispectral Sensor with a pixel size of 3.1 cm.

 

Figure 2.  Satellite image of the University Farm property (outlined in red).  Satellite image is a PSScene4Band image from Planet on June 8, 2020, with a pixel size of 3 m.

The SCP module in QGIS has two methods of generating land cover classifications.  Supervised classifications require an a priori training input to establish spectral signatures of land cover categories.  In contrast, unsupervised classifications builds a land cover classification by clustering pixels by similarity of spectral signatures.  Fig. 3 is an example of a supervised classification of the north woodlot.  Although the patterns are similar, the classification produces a different distribution of forest features.  The main obstacle to improving the classification is substantial overlap of spectral signatures of the various tree species used in the training input.  

Figure 3.  Comparison of part of the north woodlot in Fig. 1 (left panel) with the results of a supervised classification of the forest canopy of the north woodlot using the SCP module in QGIS (right panel).  The classification required manual specification of regions of interest for each of the categories.  

Fig. 4 provides an example of unsupervised classification of the image in Figure 1.  A challenge with is identifying land cover categories associated with clusters arising from the unsupervised classification.  Areas of grass cover or roads and other built features are easy to identify, but the structure of the forest canopy is much harder to classify by species.  Gaps in the forest canopy, however, are well demarcated.  High resolution of the image, in fact, leads to detection of gaps in the upper canopy of individual trees.  

Figure 4.  Results of supervised classification of forest features from image in Fig. 1.

The 3 m pixel resolution of the Planet satellite imagers provides a lower resolution option for exploring forest structure.  The feature categories in Fig. 5 lack correspondence with features in Fig. 4, except for gap areas.  The lower resolution satellite images result in identification of the larger areas between individual tree canopies.  The also reveal areas of similar spectral signatures that may not be species specific.

Figure 5.  Results of unsupervised classification of forest features in the north woodlot from PSScene4Band image in Fig. 2.

The reason for the lack of species specific spectral signatures in either Fig. 4 or Fig. 5 likely involves variation in the observable structure of tree canopies.  Fig 6, is a high resolution image of 76 x 57 m area within the north woodlot.  At this scale of resolution, it is possible to identify maple and oak leaves.  Although there is a slight difference in the perceived color of maple and oak branches, the coloration is not constant at different heights in the overall canopy.  Thus spectral signatures of lower leaves in the canopy will be different from upper leaves.  Also, the gaps between branches  is not species specific.  Both supervised and unsupervised classifications are limited by the overlap of spectral signatures associated with the physical structure of individual tree canopies.   

  

Figure 6. UAV image taken on July 24, 2017, in the north woodlot.  The image was acquired with a Phantom drone at 60 m altitude.  Individual pixel size is 1.9 cm and the width of the image is 76 m.  Arrows indicate maple and oak tree canopies.

Conclusion

Although individual leaves of tree species in the north woodlot are easy to distinguish by size, shape, and color, demarcation of individual tree species with remote sensing methods seems impractical.  The substantial overlap of spectral signatures of different species leads to unreliable categorization.  Because much of the overlap arises from the physical structure of the canopy, however, some categories will be more consistently identified.  For example, gap structure of the canopy varies with resolution of images, but is largely consistent (compare Figs. 4 and 5 for location of major canopy gaps).  Uppermost leaves of some tree species likely have  distinctive  spectral signatures.  Determining whether those are sufficiently distinctive of account for the other patterns in Figs 4 and 5 remains a work in progress.

Are Long-term Data Sets Worth the Effort at Biology Field Stations?

For the past 11 years, I have been maintaining equipment and curating data sets for environmental monitoring at University Farms of Case Western Reserve University.  Although the original deployment of was part of my research on Great Lakes ecosystems, the environmental monitoring program at the CWRU Farm became a general resource for research and education.  The original monitoring began with two weather stations, one in an open field and the other in the north woodlot.  Later, we added hydrology monitoring of the experimental watershed and additional environmental monitoring associated with the Farm’s Food Production Program.  The former Farm Director, Dr. Ana Locci, supported and encouraged this work as part of an effort to improve the research and education infrastructure at the Farm.  Our partnership on environmental monitoring platforms deepened with the arrival of Beech Leaf Disease (BLD) at the Farm in 2017 and led to incorporation of a remote sensing platform with the addition of an unmanned aerial vehicle and sensors to monitor changes in forest canopies associated with the spread of BLD.  I summarized preliminary findings of this work in previous posts.  In this post, I want to reflect on the value of long-term monitoring at a university field station.

The motivation for this reflection has arisen from both personal and institutional changes in priorities.  As with many things over the last two years, the Covid-19 pandemic has contributed to reassessment of research and education initiatives at the Farm.  With Dr. Locci’s retirement as Farm Director in 2020, the environmental monitoring program lost its primary institutional sponsor.  In the hiatus following her retirement, restrictions on activities associated with the University’s Covid protocols, led to shorter-term reassessment of priorities of Farm management and activities of the Biology faculty involved with the Biology Field Station.  Along with these changes and changes in my personal priorities, the environmental monitoring program has now ended.  Only the weather station in the north woodlot and a weather station at the November Greenhouse continue operating, and this will be the last year for which the remote data logger in the north woodlot will report data over a 3G network.  Recently, however, some CWRU faculty have asked for access to historical environmental data.  In particular, Dr. Mark Green and his undergraduate student, Kyle Rickert, have analyzed some of the weather station data as part of a small project to measure sap flux in beech trees of the north woodlot.  Although I had proposed providing a data archiving platform for Farm data in 2018, I haven’t kept up with documenting a wide range of problems with the monitoring platforms.  Various probe failures, data gaps, and difficulties accessing historical data made it difficult to give others access to a consistent set of time series from various probes.  Stimulated by Green and Rickert’s preliminary analyses of partial data sets, I have now finished a complete review of the data and am distributing the consolidated data from the weather station to interested researchers.  As of January 20, 2023, the data from the Farm weather stations is now publicly available (see References below).

In their partial analysis of soil moisture data from the north woodlot weather station, Green and Rickert found some recent increases in the soil moisture that reinforced their observations of reduced sap flux in BLD infected beech trees.  Previously, Dr. Locci and I had identified a pattern of increased solar radiation in the north woodlot where damage to beech was acute.  To me, the following Figures 1 and 2, prove the importance of maintaining long-term environmental monitoring at the Farm.  

In my first post on BLD incidence at the Farm (2017 post), I noted that the ratio of solar radiation measured at the weather stations of the north woodlot and open field had nearly doubled from a about 6% prior to 2017 to about 12% in 2017 during the summer.  As illustrated in Fig. 1, this trend continued after 2017 with the ratio approaching 19% in the summer of 2021.  Fig. 2 also shows a substantial change in soil moisture from 2017 to 2021.  Over the period of 2009 to 2016, minimum fall soil moisture (m3/ m3) ranged from 0.02 to 0.1, except for 2014.  From 2017 to 2021, minimum soil moisture increased to 0.1 in 2017 and 2018 and then to 0.22 in 2021.  

Conclusion

The eastern part of the north woodlot is a remnant primary forest with a predominant beech-maple assemblage. Beech Leaf Disease is only the latest stressor on beech in northeastern Ohio.  Beech bark disease has also impacted the beech area of the north woodlot.  The cumulative stress on beech trees is evident is the loss of large old beeches.  Without the long-term data sets from the weather stations, however, it would have been difficult to detect the timing of the consequences of these stressors.  The increases in the fall soil moisture from 2017 correspond most closely with the outbreak of Beech Leaf Disease.  It would thus seem prudent to continue this environmental monitoring program.

References

Koonce, J.F. 2023. Weather data for the period 2016 to 2022 from the HH1 Greenhouse location at University Farms, Case Western Reserve University ver 1. Environmental Data Initiative. https://doi.org/10.6073/pasta/33776fc6b8588fecc802db8405cfad3f (Accessed 2023-01-18).

Koonce, J.F. 2023. Weather data for the period 2009 to 2022 from the North Woodlot location at University Farms, Case Western Reserve University ver 4. Environmental Data Initiative. https://doi.org/10.6073/pasta/abe2eb0ba198d3faf682e0245b5eedaa (Accessed 2023-01-20).

Koonce, J.F. 2023. Weather data for the period 2009 to 2022 from the Open Field location at University Farms, Case Western Reserve University ver 3. Environmental Data Initiative. https://doi.org/10.6073/pasta/67e5f60b87c38157f5240a5546ece0af (Accessed 2023-01-20).