Calculating the Shannon Diversity Index with Data from Remote Sources

Introduction

Gaining knowledge of biodiversity is a complex and essential task for the conservation of ecosystems. With the advance of remote sensing technologies, particularly the use of ultra-high resolution satellite imagery, we have access to unprecedented detail on features of the Earth's surface. They allow us to identify species, assess habitats and monitor changes in the environment with ever greater precision. Along with high resolution satellite images, three-dimensional (3D) landscape information and three-dimensional segmentation are also proving to be powerful tools for analyzing biodiversity. These technologies give us more comprehensive and detailed views of ecosystems, to more effectively manage environments and devise better informed conservation strategies.  

Ultra-high resolution satellite images are taken by satellites equipped with sensors capable of detecting fine details on the Earth's surface. These images offer a spatial resolution down to centimeters, enabling the precise identification of different natural and artificial elements present on the ground. Combined with three-dimensional information from techniques such as photogrammetry and LiDAR (Light Detection and Ranging), we can build digital terrain models to portray topographies and vertical structures of vegetation and other landforms. 

In this context, three-dimensional segmentation is the process of dividing a 3D image into meaningful segments, which can represent different types of vegetation, bodies of water, buildings, and other features. This technique provides a more detailed and contextualized analysis of ecosystem components, making it easier to identify species and to assess the health of habitats. The combination of high-resolution images, three-dimensional data and 3D segmentation has the potential to revolutionize how we study and preserve biodiversity, through insights that are inaccessible with traditional methods. 

The problem

Characterizing biodiversity accurately and efficiently over large geographical areas is a significant challenge for environmental conservation.

State of the Art

The use of ultra-high resolution satellite imagery to characterize biodiversity has gained significant ground in recent years. Sensors such as those used by WorldView-3 and WorldView-4 satellites, with spatial resolutions of down to 31 centimeters, give us a detailed view of the Earth's surface. Combining these images with LiDAR data on the vertical structure of vegetation helps create accurate three-dimensional models of terrains and plant cover.

One of the most promising applications of this technology is to identify plant species. Spectral analysis of high-resolution satellite images allows us to distinguish different species based on their unique spectral signatures. Studies have shown that the integration of spectral and three-dimensional data significantly improves the accuracy of species identification, for a more precise assessment of plant biodiversity. 

In addition to species identification, three-dimensional information and three-dimensional segmentation are being used to map habitats and assess the health of ecosystems. Three-dimensional segmentation in particular allows for a detailed analysis of habitat structure by identifying different types of plant cover, bodies of water and other ecosystem components. This approach has been an effective way to identify critical areas for conservation and to monitor changes over time. 

The combined application of these technologies has also broadened into the study of marine ecosystems. Satellites with high-resolution sensors are used to map coral reefs, assess coral health and monitor changes in these sensitive ecosystems. The combination of high-resolution optical images with underwater sonar and LiDAR data offers a detailed view of reef structures and the distribution of marine species. 

Despite major strides, there are still challenges to overcome. Interpreting high-resolution satellite data demands sophisticated image processing and data analysis techniques. Machine-learning algorithms, such as convolutional neural networks (CNNs) and deep learning, have been used to boost precision for image segmentation and classification. The need for large volumes of training data and the high cost of remote sensing technologies, however, are still obstacles to the widespread application of these techniques. 

In a study carried out in the Lages Municipal Park, using images captured by sensors onboard remotely piloted aircraft and three-dimensional point clouds, the Shannon index was determined indirectly. By extracting vegetative indices and segmentation metrics, it was possible to determine the Shannon index with an accuracy of over 90%. 

Prospects for future studies

The outlook for future studies to determine biodiversity with ultra-high resolution satellite images is promising. Ongoing development of remote sensing technologies, along with advances in artificial intelligence and machine learning, point to further improvements in these techniques' precision and efficiency. By integrating different data sources, such as hyperspectral images, LiDAR data and climate information, we can expect an even more comprehensive and detailed view of ecosystems. 

Another highly promising field is the application of these technologies in real time for environmental monitoring. With the development of smaller and more affordable satellites, such as CubeSats, it will be possible to obtain high-resolution satellite data more frequently at a lower cost. This will enable more dynamic and responsive monitoring of environmental changes, facilitating early detection of threats to biodiversity and the rapid implementation of conservation measures.

Finally, interdisciplinary collaboration will be key to further progress in this area. Combining knowledge from biology, ecology, remote sensing, data science and engineering will generate innovative and effective solutions for biodiversity conservation. Educating and training new professionals in remote sensing techniques and data analysis will be essential to meet future challenges and guarantee the preservation of ecosystems for future generations.

Case study

In order to assess the biodiversity of a Mixed Ombrophilous Forest area located in the Lages Municipal Park, we decided to determine the Shannon index for the upper canopy. Thirteen forest inventory plots (20m x 20m each) were set up on the site.

Aerial images were acquired by a remotely piloted aircraft (RPA) equipped with an RGB sensor. The images were taken at a height of 120 meters, with longitudinal and lateral coverage of 80% and 60% respectively. The figure below illustrates the layout of the inventory plots at the study site.

We obtained the digital orthoimage with a spatial resolution of 4 cm and the digital surface model (DSM) with a spatial resolution of 10 cm. The co-occurrence matrix was generated for the orthoimagery, and a dataset was generated containing all the aforementioned data.

Following a multiscale segmentation, we calculated segment parameters, which were then processed using the EM (expectation maximization) simple cluster algorithm. This algorithm was chosen because there is no need to define the number of classes (of species), which was exactly the information we wanted. Using the number of species (in the Upper Canopy), it was possible to calculate the Shannon index.

The Shannon indices were very close to those obtained in the traditional inventory, with consistent values. Differences in the number of species for each methodology can be explained by the fact that the automatic methodology considers the features of images that use only visible spectrum bands. 

* Marcos Benedito Schimalski is researcher and associate professor in the Department of Forest Engineering at Santa Catarina State University