Three-Dimensional Landscape Visualizations: New Technique towards Wildfire and Forest Bark Beetle Management

After a century of fire exclusion, western US forests are vulnerable to wildfire and bark beetles. Although integrated fire and pest management programs (e.g., prescribed burning and thinning) are being implemented efficiently, damage to forests continues. Management challenges come in the forms of diverse land ownership, dynamic forest landscapes, the uncertainty effect of management strategies, and social interaction of the increasing wildland-urban interface. Three-dimensional (3-D) landscape visualization is comprised of multi-spatial, multi-temporal, and multi-expression elements. Supplemented with GIS database, remote sensing images, and simulation models, this technique can provide a comprehensive communication medium for decision makers, scientists, stakeholders, and the public with diverse backgrounds on the wildfire and forest bark beetle management. The technique we describe here can be used to organize complicated temporal and spatial information, evaluate alternative management operations, and OPEN ACCESS

the characteristics of 3-D landscape visualization for contributing to wildfire and bark beetle management decision-making processes.

Characteristics of 3-D Landscape Visualization
3.1.1 Multi-spatial 3-D visual simulators allow the flexibility to choose several perspectives in representing forest landscapes ( Figure 1). This takes advantage of the tendency of observers to view an area from different directions, locations, and distances [20]. In addition, viewer movement is a normal way of experiencing the forest landscape. Usually, visual landscape simulators use camera position to provide various perspective viewsheds from any point of view [21]. We can therefore assign any specified pathway and duration for the movement of the camera to provide animation by flythrough or walkthrough of the viewshed of interest (http://people.clemson.edu/~cchou/VStream.avi).

Multi-temporal
Quantitative and information-based 3-D landscape visualizations can visualize stand succession, landscape transformation, and regional planning [19]. They are capable of visualizing forest changes caused by management activities and disturbances. And they are capable of demonstrating future development using time-series databases and predictive models. Visualizing the past, present, and future conditions of forest landscapes provides the ability to display potential outcomes that are difficult to illustrate in the field [22] (Figure 3). Spatial scale is also a key issue in developing forest visualizations. The amount and types of data needed for stand versus landscape scale visualizations differ because the purposes differ [8]. Stand scale visualization focuses on accurately displaying the vertical structure and dynamics of the stand, including stand density, species composition, and tree height. In contrast, landscape scale visualization emphasizes relative landscape components, such as the arrangement and interaction of patches, corridors, and matrixes. Hence, depending on the purpose, 3-D visualizations can be created for specified foliage effects (e.g., different species, ages, and statuses), understory and overstory ecotypes, and ground effects for appropriate landscape elements under assigned visualization scales ( Figure 2).

Multi-expression
In the past, tables, figures, and maps have been the predominant methods used to communicate management alternatives, but these types of data include a high level of abstraction. Recently, 3-D visualizations have allowed the creation of perspective views to achieve more natural and direct depictions, enhance communications, and make complex information more easily understood by both experts and the general public [16]. Usually, a static diagram can only display a maximum of three factors [16]. In reality, natural phenomena result from many interactive factors and effects. For instance, bark beetle spot growth is typically discussed by referencing distances between pines, susceptible species, temperatures, and seasonal effects [11]. However, overall time-series effects (i.e., extended drought, infestation history) and relative spatial effects (i.e., arrangement of landscape patches, landforms, slopes, and soil characteristics) are seldom considered in beetle-infected forests [2,10,23]. 3-D landscape visualization allows the overlaying and integration of the combined effects of diverse interactive geo-information into simplified and integrated 3-D media [24] (Figure 4).

3-D Landscape Visualization
The 3-D landscape visualizations in Figures 1-4 were generated using a 3-D simulator software called Visual Nature Studio (VNS, 3D Nature Inc. http://3dnature.com/), with the required data to delineate a landscape consisting of the terrain, vegetation, water, built structures, animals, and light [16,21,25].
VNS is a premium photo-realistic and landscape-visualization software package. It was chosen from a vast number of visualization tools for some of its specific qualities: (1) integration with georeferenced GIS datasets, (2) flexibility of land cover type development, (3) use of raster and vector formats to drive rendered vegetation components, and (4) including both motion and time-series animation ability [26]. Although, VNS provides the flexible and various models to bring a scene to life with vivid photo-realistic visualization, it requires skilled operation, high-end hardware, and long rendering time for high quality animation.
In order to visualize the base layer of landscapes, terrain was obtained based on digital elevation model (DEM; the most common source of digital terrain models, [27]) from high-resolution remote sensing images [16,25,28]. In addition, vegetation visualization is the critical technique to determine if the 3-D landscape visualization is convincing or not [24]. To realistically visualize the vegetation,

Incorporating Remote Sensing Images and GIS
GIS is a computer program that allows one to efficiently manage, process, analyze, and represent data. The data can be referenced to a location on the earth, including any thematic attribute that may be connected to the location [21]. GIS can be used to develop data sources for projecting 3-D landscape visualizations. With GIS one can utilize information from field or remotely sensed data to help classify different land cover types, such as agricultural land, rangeland, forestland, and wetlands. The GIS can also maintain detailed forest stand characteristics, including cover type, tree species, crown diameter, and stand height and density. Compatibility of 3-D visualization simulators and GIS data layers allows the rendering of landscapes based on information collected from the actual landscape.
Remote sensing images, including aerial photography and satellite imagery, constitute the basis for the creation of a variety of spatial data. The DEM, which forms the basis of most landscape visualization, is generally derived from remotely sensed data such as aerial photography, LiDAR (Light Detection and Ranging), and Interferometric Synthetic Aperture Radar. The elements of landscape include the physical materials or objects on the surface of the land, and are also known as land covers [23]. By interpreting remote sensing images, we can generate land cover maps. These images can include a widespread area with multi-spatial resolutions (the finest resolution can be less than 1 m). Information not apparent with visible light can be also obtained from multispectral satellite imageries that record the detailed physical characteristics of ground-features [29,30]. Based on these images, we can identify, recognize, and delineate land cover maps on multilevel land cover classification systems to support different landscape scale managers with appropriate resolution information on a nationwide, interstate, or countywide basis [29]. Moreover, foliage color usually changes with damage from wildfire or bark beetle outbreaks, and these widespread infestation phenomena can be easily detected using aerial photographs [30]. For these reasons, aerial photo interpretation usually supports the detection and assessment of stand health, forest vigor, the different stages or degrees of damage, and predicting the spreading region of a bark beetle or wildfire damage [30].
The analysis of landscape pattern change is an important method for understanding significant ecological dynamics, such as natural and human disturbances, forest succession, and recovery from previous disturbances [31]. Satellite imagery and aerial photography have been classified according to vegetation or land cover types, and they provide an excellent source of data for performing structural studies of landscapes [26]. When comparing these remote sensing images over time, they become especially useful for describing types of landscape changes and indicating the resulting impacts on surrounding habitats [26] (an example shown in Figure 3).
Therefore, remote sensing images and GIS database can help us effectively monitor forest changes according to type, duration, and intensity [32]. They can facilitate the representation of these highly dynamic temporal and spatial phenomena at varying scales ranging from an individual tree to an extensive forest landscape [32,33]. Illustrating landscape change is one of the most beneficial applications of 3-D landscape visualization [34]. Using a time-series GIS database and remote sensing images, we can delineate the appearance of terrain, land cover, and vegetation to compare the spatial and temporal changes in past and present forest landscapes [17] (Figure 3). In attempts to visualize future landscapes, the visual projection must be driven by dynamic models that can simulate the recovery, succession, or growth situations under different management scenarios.

Incorporating Simulation Models
In the following discussion, we examine how simulation models can be linked with 3-D landscape visualization.

Forest Vegetation Simulator (FVS)
FVS is a distance-independent growth and yield model at the individual tree scale [35]. It can simulate growth and yield for major forest tree species, forest types, and stand conditions for all national forests in the US [35]. For instance, Wang et al. [19] used a Forest Inventory Analysis (FIA) dataset to simulate the dynamics of tree size (diameter and height) for different forest types in FVS models and VNS.

Fire Area Simulator (FARSITE)
FARSITE is a fire growth simulation model that can incorporate existing models of surface fire, crown fire, point-source fire acceleration, spotting, and fuel moisture to simulate fire behavior and represent fire growth and effect over an entire landscape [36]. Williams et al. [37] used VNS and shapefiles of different wildfire stages generated from FARSITE to visualize fire spread and intensity across the New Jersey Pine Barrens. In this study, in order to describe the different effects of fire severity and crown fire on forest stands, they created burned tree image models using Photoshop to visualize the different flame effects on either individual trees or clusters of trees [37]. As a result, both still frame and animated views of wildfire visualizations were established by combining the burned tree models with different flame models [37].

CLEMBEETLE
The CLEMBEETLE model simulates the stand damage caused by southern pine beetle (SPB), including estimating the number of attacked and killed trees per spot, percentage of stand killed per acre, and expected yield per acre with or without SPB attacks [23]. Chou et al. [38] visualized the active spot growth with different affected-stage from SPB infestation. By using CLEMBEETLE, a GIS-based spot growth model, VNS, and ArcGIS (Environment System Research Institute Inc. http://www.esri.com/), 3-D visualizations of SPB spot growth with different stand densities, species compositions, and stand ages were generated [38]. This new GIS-based spot growth model created realistic views with stereo viewsheds and vivid foliage images, and helps us understand the dynamics of SPB spot growth under different silvicultural scenarios [38].

Landscape Disturbance and Succession Simulation Model (LANDIS)
LANDIS is a spatially explicit and stochastic landscape model for simulating large-scale and longterm forest landscape processes with species level vegetation dynamics. It can generate the time curves for heterogeneous spatial patterns and species abundance to represent the interaction of disturbances and succession with changing forest patterns over long periods of time and a wide range of landscape scales [39]. It can also provide species information, including age class, abundance percentage, diameter, and density on different land types within a specified environmental situation [15,40], to support the required data of landscape visualization. LANDIS has been used to simulate the disturbance influences of wildfire and SPB on successional dynamic landscapes [40,41], and it also generates major required data for the 3-D landscape visualization.

3-D visualization that incorporates a GIS database, remote sensing images, and simulation models
can provide a more comprehensive, practical, and applicable approach for monitoring spatial pattern changes due to disturbances caused by wildfires and bark beetles. It can be used to evaluate alternative management strategies and to effectively communicate the impacts of those strategies to diverse stakeholder groups. In addition, 3-D visualization can depict the structure and composition of landscapes, and also portray spatial and temporal changes resulting from different natural disturbances or management strategies [5,15,18,42,43] [43,46,47]. The 3-D visualization has been used in the latter two applications to provide visual representation (i.e., photo-realistic visualizations) to extend our power of perception to consequences of non-visual processes (i.e., stand growth and yield model, forest restoration, ecosystem succession, etc.) [45][46][47]. Consequently, it can synthesize different dimensions (i.e., time-series, spatial scale, purpose, etc.) and provide a well-organized technique. It can be used to combine complex information into a comprehensive media to enhance the integration of information, processes, and strategies.
The value of such 3-D landscape visualization depends on accuracy and realism, which will depend on the quality of the supporting data and the validity of the simulation models [22,48]. In order to produce 3-D visualization that can be viewed with confidence by various public groups, we must be assured of the accuracy of the underlying forest data and the application of this data to simulation models. Especially, the visual representation should be defensible through making the projection process and assumptions transparent to the audiences, and by clearly describing the expected level of accuracy and uncertainty [49].
In the future, the research group would aim to improve the quantitative analysis of (1) whether the 3-D visualization (comparing to the text, tabular, or 2-D map) could help the participants articulate more clearly their preferences for landscape conditions [50], (2) whether the 3-D visualization could increase the perception of multi-purpose, multi-temporal, and multi-spatial alternative forest management strategies [45,51], and (3) the accuracy of assessment (i.e., the ability of the simulation model to capture the essence or details of the scene) by comparing static views of the projected landscape visualization with known photorealistic viewpoints [52]. Furthermore, although the 3-D visualization is recognized as a helpful and meaningful medium to forest management plans and other activities, it is still a new technique for forest research with limited used. More widespread studies are needed to extend its applicability, as well as the development of standard guidance and validation for its use in practice [15,50].