GEOG 560 GIScience I: Introduction to Geographic Information Science
Yang Han
Option 1 – Annotated Bibliography
My name is Yang Han, a first-year PhD candidate in the Sustainable Forest Management program under the Department of Forest Engineering, Resources and Management. My academic exploration centers on geospatial technologies and their practical implementation in forest ecosystems. Specifically, I aim to investigate how Geographic Information Systems (GIS) can enhance contemporary forestry practices while identifying operational constraints that may affect their real-world adoption.
My future researches will focus on three core aspects: current GIS applications in forest resource management, their potential for optimizing engineering solutions in silvicultural operations, and systemic challenges that might hinder technological integration. This comprehensive approach seeks to bridge theoretical frameworks with practical field applications, addressing both the technical capabilities and institutional limitations of spatial analysis tools in forestry contexts.
Article 1
Gonzalez-Benecke, C. A., Gezan, S. A., Martin, T. A., Cropper, W. P., Jr., Samuelson, L. J., & Leduc, D. J. (2014). Individual tree diameter, height, and volume functions for longleaf pine. Forest Science, 60(1), 43–56. https://doi.org/10.5849/forsci.12-074
Gonzalez-Benecke et al. (2014) mainly develops individual tree models to predict the height, diameter of longleaf pine bark, stem volume and merchantable volume ratios. This individual tree model really valuable to my interested research, as I plan to utilize GIS to explore the business value of individual trees in the forest. In addition, the article also compares local model using tree-level variables with general models using stand-level variables. There are 3 key findings of this article that may relevant to my future research. First, the study shows the general models outperformed local models. Some variables such as age, stem diameter outside bark at breast height, basal area, site index and so on, reduced RMSE of the model to 7.61%. Secondly, the study also shows models with dbh and height were more accurate. But if height data were unavailable, tree density and site index also can improve the accuracy of the model. Last and most important, Gonzalez-Benecke et al. (2014) utilize GIS in the study and prove that GIS can map stand variables, such as site index, basal area and so on, by applying these variables at landscape scales. Also, GIS can identify spatial patterns in stand productivity to guide thinning, harvesting, and restorations in the forest management. To conclude, Gonzalez-Benecke et al. (2014) explores the importance of stand-level variables in forest modeling, aligning with GIS’s capacity to manage spatial data. By integrating these models into GIS platforms, researchers can achieve more precise, scalable, and actionable outcomes and insights. However, it is also noticeable that data acquisition and technical integration may impose negative impacts on unlocking GIS’s full potential in advancing forest management and engineering.
Article 2
Gonzalez-Benecke, C. A., Gezan, S. A., Leduc, D. J., Martin, T. A., Cropper, W. P., & Samuelson, L. J. (2012). Modeling survival, yield, volume partitioning and their response to thinning for longleaf pine plantations. Forests, 3(4), 1104–1132. https://doi.org/10.3390/f3041104
Gonzalez-Benecke et al. (2012) built a stand-level growth and yield model for longleaf pine. Researchers utilize this model to tackle survival, dominant height, basal area and merchantable volume issues. Gonzalez-Benecke et al. (2012) integrating thinning effects and merchantable volume breakdowns into the comprehensive stand-level growth and yield model and using 40 years US Forest Service dataset to predict site index, mortality, basal area growth post-thinning and volume partitioning. There are several key findings from the article. Firstly, the accuracy of the stand-level growth and yield model is very high with 94.6% to 99.8% R square. At the same time, the merchantable volume breakdown functions outperformed to models for other southern pines. Second, Gonzalez-Benecke et al. (2012) conclude that survival rates decreased with higher dominant heights, which indicates site quality and stand age. While basal area growth relies on stand density and dominant height. From the forest management aspect, this stand-level growth and yield model can be used for long-term silvicultural planning, thinning schedules and product yield optimization. In addition, the study also can be applied to pine plantation management, especially for forest restoration. Last, assume future research can align this model with GIS, it may create dynamic tools for real-time management adjustments and build GIS-based simulations to test how climate change impacts growth trajectories. To sum up, this study encapsule a growth and yield framework that may combine with GIS to make precision forest management. However, future research challenges may lie in scaling plot-based models to explore hybrid models and ensure seamless integrations with GIS.
Article 3
Sun, C., Huang, C., Zhang, H., Chen, B., An, F., Wang, L., & Yun, T. (2022). Individual Tree Crown Segmentation and Crown Width Extraction From a Heightmap Derived From Aerial Laser Scanning Data Using a Deep Learning Framework. Frontiers in Plant Science, 13, 914974. https://doi.org/10.3389/fpls.2022.914974
Sun et al. (2022) mainly discuss the important role of AI-enhanced GIS in forest management, which highly related to my study interests. Authors points out that AI-enhanced GIS can offer a more scalable, accurate method for tree crown segmentation. Their study combines LiDAR with deep learning to deal with the challenges in spatial data analysis. From my perspective, this AI-enhanced GIS may provide a blueprint for utilizing advanced GIS technology to improve forest management and operational efficiency. There are several key points about this study. First of all, it develops a deep learning framework combining YOLO-v4 with generative adversarial networks (GANs) to segment individual tree crowns (ITCs) from heightmaps derived from airborne LiDAR data. At the same time, GAN variants, such as Cycle GAN, WGAN-GP, SinGAN, are used to augment training data to improve the model performance. As a consequence, this model achieved 83.6% recall 81.4 % accuracy among nursery forest, forest landscape and mixed plantation. Moreover, the model also indicates crown width estimation has strong correlation with field measurements with 79.93% R squares, which outperforming traditional methods. In addition, GAN-based data augmentation reduced the reliance on manual labeling and also show adaptability to diverse forest structures, such as complex canopies with overlapping crowns. Last, the applications of this model are promising. For instance, the LiDAR-derived heightmaps integrated with deep learning for crown width parameter extraction and tree density data collection may help researchers to do better forest inventory and monitoring. This framework also supports automated, high-accuracy mapping of individual trees, critical for precision forestry, carbon stock estimation, and biodiversity monitoring. This model provides an economic method for large-scale forest management by utilizing AI technology combined with GIS. For future researches, remoting sensing, AI technology and GIS may work together to deal with forest engineering issues, such as optimizing harvest planning, pest outbreak detecting through crown health analysis and so on. however, interdisciplinary cooperation may be needed to overcome barriers like occlusion in dense canopies or understory vegetation interference.
Article 4
Jurado, J. M., Lopez, A., Padua, L., & Sousa, J. J. (2022). Remote sensing image fusion on 3D scenarios: A review of applications for agriculture and forestry. International Journal of Applied Earth Observation and Geoinformation, 112, 102856. https://doi.org/10.1016/j.jag.2022.102856
Jurado et al. (2022) discuss a comprehensive overview of how 3D remote sensing fusion enhances forest management and engineering. It validates the potential of GIS to integrate multi-source data for advanced spatial analysis while highlighting technical and operational barriers. There are several important takeaways from the article. First of all, 3D models can enhance spatial and spectral analysis in forestry and agriculture, such as LiDAR, photogrammetric point clouds and so on. Also, 3D models ensure detailed structural assessments, such as tree height and canopy volume, which can help vegetation dynamics, disease detection and biomass estimation. In this study, Unmanned Aerial Vehicles (UAVs) are vitally important tools for collecting high-resolution, multi-temporal datasets. Sensors like multispectral cameras, thermal imagers, and hyperspectral devices are critical for capturing data on plant health, water stress, and species classification. These UAVs and sensors may also play important roles in my future studies. This article provides very meaningful insights and guidance for my study. Another important lesson is that machine learning and deep learning is the future trends, so as the Real-time GIS integration with 3D models. These interdisciplinary technology and tools may provide future direction for dynamic forest monitoring, automated feature extraction, forest inventory, diseases detection and fire risk management. For future research, the insights on UAV-based 3D modeling, machine learning integration, and thermal/hyperspectral applications offer actionable pathways to innovate GIS workflows in forestry. However, challenges may lie in data interoperability and computational scalability.
Article 5
Beloiu, M., Heinzmann, L., Rehush, N., Gessler, A., & Griess, V. C. (2023). Individual Tree-Crown Detection and Species Identification in Heterogeneous Forests Using Aerial RGB Imagery and Deep Learning. Remote Sensing, 15(5), 1463. https://doi.org/10.3390/rs15051463
Beloiu et al. (2023) explore the feasibility of combining Convolutional Neural Networks (CNNs)—specifically the Faster R-CNN algorithm—with open-source aerial RGB imagery to detect individual tree crowns and identify four key tree species, which are Norway spruce, silver fir, Scots pine, European beech, in heterogeneous temperate forests. They used single-species and multi-species models to train more than 10,000 trees datasets, validated across diverse forest structures and illumination conditions. Norway spruce shows the highest accuracy in the single-species model, with Scots pine demonstrates the lowest accuracy in the same model. As to multi-species models, it increased detection for underrepresented species and decreased misidentification errors. The study outcome shows that forest tree density and canopy overlap significantly influenced model accuracy, while illumination had minimal impact. This study highlights how GIS-integrated deep learning can automate tree species identification and geolocation. It also provides a framework for large-scale forest management, monitoring and carbon stock estimation. Beloiu et al. (2023) underscores the importance of GIS and remote sensing in modern forestry. Future studies may also leverage GIS and remote sensing to combined with advanced AI technology to explore more scalable and lower costs solutions for species mapping and ecosystem monitoring. However, future challenges may come form high quality data acquisition, model adaptability, and so on. Overall, this study may provide actionable insights for leveraging AI-driven GIS tools to enhance ecological research and operational forestry practices.
Article 6
Dong, C., Zhao, G., Meng, Y., Li, B., & Peng, B. (2020). The effect of topographic correction on forest tree species classification accuracy. Forests, 11(3), 287. https://doi.org/10.3390/f11030287
In Dong et al. (2020) study, cosine, C, SCS+C, and empirical rotation models are evaluated for correcting Landsat 8 imagery in mountainous terrain. Among these four models, the SCS+C model and empirical rotation model works best in reducing band standard deviations, adjusting reflectance distributions, and improving visual consistency. Another key finding of Dong et al. (2020)’ study is that topographic correction significantly enhanced tree species classification accuracy, by using the SCS+C model, the overall accuracy increased 4% with ull-coverage training data and 13% with shadowless training data. In summary, this article highlights the transformative role of GIS in forest management by improving the accuracy of remote sensing workflows through topographic correction. It highlights the feasibility of integrating cloud computing (GEE) and machine learning for scalable, cost-effective forest monitoring. For my future study, this study provides an important guidance that topographic correction can work as a critical step to improve the mapping accuracy. At the same time, the use of Google Earth Engine (GEE) demonstrates the scalability of cloud-based GIS platforms for large-area forest analyses, reducing computational barriers. However, future challenges may lie in data resolution, technical complexity, and model adaptability. For instance, moderate-resolution DEMs and satellite imagery may limit the model precision, higher resolution DEMs might further improve study results.
Article 7
Michałowska, M., & Rapiński, J. (2021). A Review of Tree Species Classification Based on Airborne LiDAR Data and Applied Classifiers. Remote Sensing, 13(3), 353. https://doi.org/10.3390/rs13030353
Michałowska, & Rapiński (2021) explore the utilization of Airborne Laser Scanning (ALS) LiDAR data in the tree species classification. Michałowska, & Rapiński (2021) review 44 studies to identify the most effective LiDAR-derived features and classifiers for species discrimination. They found that Full-waveform (WF) metrics has the highest classification accuracy, while Geometric features has the lower accuracy. The Radiometric features work in the moderate level. For classifiers, Random Forest (RF) and Support Vector Machine (SVM) algorithms both achieve the highest accuracy by leveraging complex feature interactions. From Michałowska, & Rapiński (2021)’s study, it shows that LiDAR can works well to automate species mapping, enabling precise biomass estimation, growth modeling, and biodiversity monitoring. In addition, by integrating ALS LiDAR with hyperspectral or multispectral imagery, they may provide an actionable strategy to GIS-based forest inventories. For future study, researchers may consider to use multisensor fusion (LiDAR + optical imagery) and AI-driven classifiers, such as deep learning to deal with species overlap and structural heterogeneity in complex forests. But the major challenge may come from the need for cost-effective UAV-LiDAR solutions.
Article 8
Landsberg, J. J., & Waring, R. H. (1997). A generalised model of forest productivity using simplified concepts of radiation-use efficiency, carbon balance and partitioning. Forest Ecology and Management, 95(3), 209–228. https://doi.org/10.1016/S0378-1127(97)00026-1
Landsberg & Waring (1997) introduce 3-PG (Physiological Principles in Predicting Growth) model to make forest productivity forecasting. This 3-PG integrates physiological principles (e.g., radiation-use efficiency, carbon allocation) with empirical relationships to estimate forest growth over more than 100 years. The major inputs of this simplified processed based model including radiation, precipitation, vapor pressure deficit, soil parameters, and stand characteristics. The output of the model are stem biomass, leaf area index (LAI), and carbon allocation patterns. This study related to my future study interests, as this 3-PG model can be parameterized using remotely sensed LAI and weather data. Hence, it may suitable for scaling predictions across landscapes via GIS. In addition, this old model is also a practical tool for forest managers to make decision supports. For example, the model’s outputs can inform GIS-driven forest management plans, such as optimizing harvest schedules or identifying areas vulnerable to drought. At the same time, GIS can spatially infer 3-PG model to assess climate change impacts, such as precipitation pattern changes, on forest productivity. However, it is also noticeable that high-resolution weather and soil data vitally important for the accuracy of the model. Hence, the data quality may limit the overall performance of the 3-PG (Physiological Principles in Predicting Growth) model. In addition, if apply this model in the large geographic areas via GIS, huge demand for computational resources is also a challenge.
Article 9
Novkovic, Ivan, et al. “GIS-Based Forest Fire Susceptibility Zonation with IoT Sensor Network Support, Case Study—Nature Park Golija, Serbia.” Sensors (Basel, Switzerland), vol. 21, no. 19, 2021, pp. 6520-, https://doi.org/10.3390/s21196520.
Novkovic et al. (2021) mainly combine GIS, multi-criteria decision analysis (MCDA), and Internet of Things (IoT) technologies to assess forest fire susceptibility in the protected Nature Park Golija, Serbia. Researchers utilize the Index-based Forest Fire Susceptibility Index (RC) and the Fuzzy Analytic Hierarchy Process (AHP) methods to conduct their study. They identify the high-risk zones, accounting for 25% to 26% of the Nature Park Golija. These areas are much more likely to have forest fires. Furthermore, researchers also find that These zones correlate with coniferous forests, steep south-facing slopes, and proximity to human activity. In addition, IoT sensor network also applied in this study that are used to collect real-time data, such as temperature, humidity, wind speed, rainfall and environment data, like CO, CO₂, O₂ levels. One of the key finding of the Novkovic et al. (2021) study is that GIS-based modeling combined with fuzzy AHP and TOPSIS provided a robust framework for spatial fire risk assessment, enabling precise forest fire monitoring. For my future research, this study helps me realized the important role of GIS in the multi-criteria spatial analysis in the forest management field. Moreover, the combination of remote sensing and machine learning also highlight the important role of GIS in dealing with large datasets on the forest management area. Moreover, the application of the IoT sensor network also provide me some insights of developing smart forest monitoring systems, which may narrow the gap between GIS analysis and on-ground engineering implementations.
Article 10
Teich, Michaela, and Peter Bebi. “Evaluating the Benefit of Avalanche Protection Forest with GIS-Based Risk Analyses—A Case Study in Switzerland.” Forest Ecology and Management, vol. 257, no. 9, 2009, pp. 1910–19, https://doi.org/10.1016/j.foreco.2009.01.046.
The study evaluates the feasibility of applying GIS-based risk analysis in forests against snow avalanches. The authors combined forest structure mapping, avalanche simulation modeling (AVAL-2D), and risk assessment to quantify how changes in forest cover may impose impacts on avalanche risk. The key findings of their study including upper-slope forests have much higher risks than lower-slope forest. Moreover, deforested issue can increase annual collective risks by around 2.5 million CHF. Last, variations in avalanche friction parameters significantly affected simulated run-out distances and risk outcomes. Hence, this study shows that spatially explicit risk analysis can effectively quantifies the protective role of forests management. Forest cover is vitally important for snow avalanche mitigation. For the future research, this study indicates that GIS-based risk analysis may also appliable to other hazard mitigations, such as rockfall, landslides and so on. However, it is also noticeable that there are uncertainties and long-term forest changes, so future study may need to pay attention to collect dynamic forest data.
Article 11
Nakao, Katsuhiro, et al. “Assessing the Regional-Scale Distribution of Height Growth of Cryptomeria Japonica Stands Using Airborne LiDAR, Forest GIS Database and Machine Learning.” Forest Ecology and Management, vol. 506, 2022, pp. 119953-, https://doi.org/10.1016/j.foreco.2021.119953.
Nakao et al. (2022) build a model to evaluate regional-scale height growth patterns of planted Japanese cedar, by using technologies including airborne LiDAR, forest GIS databases, and machine learning. Airborne LiDAR-derived canopy height data were combined with climate factors, such as warmth index, precipitation, topographic factors, such as slope, aspect, solar radiation, and stand age variables to build a random forest (RF) model. The model indicates that in cooler climate, 20.9% stand age and 10.6% warmth index were the most influential factors for height growth. While in the warmer climate, with 23.8% slope aspect and 16.6% stand age, southwestern slope is likely to reduce growth. The outcomes of this study indicates that the integration of LiDAR, GIS, and machine learning technologies can help researchers to make accurate, large-scale assessment of forest productivity. In addition, Climatic and topographic factors may impose different impacts on different regions. Hence, it is necessary to build local models for forest management. This outcome aligns with my future research focus on GIS application in forest management. This article shows me a replicable method to create site-specific growth maps, aiding timber yield predictions and climate adaptation planning. I may apply this method to my future researches.
Article 12
Ren, Yin, et al. “Effects of Rapid Urban Sprawl on Urban Forest Carbon Stocks: Integrating Remotely Sensed, GIS and Forest Inventory Data.” Journal of Environmental Management, vol. 113, 2012, pp. 447–55, https://doi.org/10.1016/j.jenvman.2012.09.011.
Ren et al. (2012) explore the impacts of rapid urban sprawl on urban forest carbon stocks in Xiamen, China. They utilizing remote sensing, GIS, and forest inventory data to conduct research throughout 40 years. Carbon stocks and densities were analyzed across three urbanization zones, which are urban core, suburb, exurb, to assess the effects of human disturbance and natural regrowth. The find that Carbon stocks and densities declined significantly in the urban core area, while Carbon stocks and densities increased in the exurbs. As to suburbs, mixed effects were observed and carbon dynamics in the suburbs are mainly influenced by human activity proximity. From this article, I realized that spatial analysis can be used to quantify carbon dynamics. In addition, the multi-source data can help researchers to do holistic forest monitoring. GIS also ply important role in shaping sustainable urban-forest interface policies. I may adopt similar approaches to work on advanced GIS-driven solutions for carbon-smart forest engineering and forest management. In addition, this study also reflects that GIS can deal with large datasets and quantify human-natural activities. This finding may also apply to my future research.
Article 13
Zerouali, Bilel, et al. “A Cloud-Integrated GIS for Forest Cover Loss and Land Use Change Monitoring Using Statistical Methods and Geospatial Technology over Northern Algeria.” Journal of Environmental Management, vol. 341, 2023, pp. 118029–118029, https://doi.org/10.1016/j.jenvman.2023.118029.
Zerouali et al (2023) study the integration of remote sensing (RS) and geographic information systems (GIS) in assessing forest and grassland ecosystem health. The study highlights the potential of RS and GIS to tackle with large-scale monitoring and standardization of ecosystem health metrics. The key findings of their study including that climate change and insect infestation are the major stressors for forests, while for grassland, the major stressors are grazing and climate change. Their review also finds that multispectral sensors such as Landsat, MODIS are widely used in 45% to 53% studies. At the same time, GIS is also frequently used in mapping fragmentation, disturbance regimes, road density and land cover changes. In the future researches, I may leverage scalable, repeatable, and cost-effective RS/GIS workflows, to address critical challenges in monitoring ecosystem health, optimizing resource allocation, and enhancing climate resilience. However, there are also challenges. For example, currently, we may need a unified framework to align field-based measures with remote sensing (RS) and geographic information systems (GIS) proxies. Future work may focus on standardizing indicators, improving sensor integration, and aligning outputs with policy frameworks to maximize impact.
Article 14
Soubry, Irini, et al. “A Systematic Review on the Integration of Remote Sensing and GIS to Forest and Grassland Ecosystem Health Attributes, Indicators, and Measures.” Remote Sensing (Basel, Switzerland), vol. 13, no. 16, 2021, pp. 3262-, https://doi.org/10.3390/rs13163262.
Soubry et al. (2021) evaluate forest cover loss and land use changes in Northern Algeria among 2 decades. They utilize GIS, remote sensing, and statistical methods to analyze trends across 17 river basins. They find that forest area shrink 64.96% from 2000 to 2020. Even worse, Constantinois-Seybousse-Mellegue basin experienced the 1,018 km² absolute loss, while the Seybouse basin had the highest percentage loss, which is up to 40%. At the same time, urban areas and croplands expanded, while shrublands, savannas, and grasslands declined. They conclude that human activities and climate change are the major reasons lead to serious forest loss in northern Algeria between 2000 and 2020. In addition, researchers find that GEE and GIS are able to efficiently monitor large-scale forest change, and tracking those changes. Morevoer, the study demonstrates GIS’s capability to process multi-decadal satellite data, such as Landsat, MODIS, for detecting deforestation hotspots and LULC trends. At the same time, the Random Forest algorithm also highlights GIS can be used to improve land classification accuracy, even in complex ecosystems. GEE’s cloud computing power also explore how GIS can deal with big data efficiently, and conduct real-time or historical analysis without local storage constraints. Based on this study, my future studies may further explore advanced GIS technology applications, such as predictive modeling of forest recovery or integrating IoT sensors with spatial analytics for real-time forest health monitoring.
Article 15
Noroozi, Farzaneh, et al. “Forest Fire Mapping: A Comparison between GIS-Based Random Forest and Bayesian Models.” Natural Hazards (Dordrecht), vol. 120, no. 7, 2024, pp. 6569–92, https://doi.org/10.1007/s11069-024-06457-9.
Noroozi et al. (2024) assess the effectiveness of two machine learning models, which are Random Forest (RF) and Bayesian approaches, for mapping forest fire susceptibility in the Firouzabad region of Iran. In the research, they utilizing GIS technology to integrate 12 environmental and anthropogenic variables, such as elevation, rainfall, distance to roads, temperature, vegetation indices to predict fire-prone areas. As a consequence, the Random Forest model outperformed the Bayesian approach, with higher 0.876 AUC. In addition, the Random Forest model identify 37.86% of the study area as high risk, while Bayesian approaches classified 48.46% of the study area as high risk. The study also identifies the most critical factors lead to forest fire, which are Elevation, annual rainfall, distance to roads, and temperature, while slope aspect, topographic wetness index (TWI), and slope percentage are the least important factors. Another important finding is that human activities and climate change are the two factors amplified forest fire risks. Based on this study, we can know that GIS-based model can conduct quick and data-driven fire risk assessments. This model can help government to take proactive forest management and resource allocations. From this study, it demonstrates that GIS combined with machine learning to build predictive models like Random Forest (RF) model can optimize forest fire risk mapping and work as an important role of sustainable forest management. In the future studies, I may adopt those similar methodologies to address complex challenges in forest engineering and forest management.
