A Web‐Based Integrated Modeling and Simulation Method for Forest Growth Research

To facilitate forest research, simulations of the whole forest growth process can be employed to analyze forest dynamics and predict forest yields. Different forest growth models can be integrated for comprehensive process simulation and thus can assist forest growth research. With the development of network technologies, a web environment can provide cross‐platform capability and wide availability for distributed researchers. In order to serve the simulation of complex forest growth processes and help online forest growth research, this article proposes a web‐based integrated modeling and simulation method for forest growth research. The proposed method includes three steps, namely, model preparation, model integration, and forest growth simulation. The corresponding implementation strategies are designed to prepare forest growth models, integrate different models, preprocess model data, and implement forest growth simulations for integrated modeling and simulations via the web. Two applications in the comprehensive prediction of forest growth and comparison of different forest management decisions are introduced to verify the feasibility and capability of the proposed method. The results show that the proposed web‐based integrated modeling and simulation method can be used conveniently for comprehensive simulations of forest growth research.


Introduction
Forest growth simulations can provide insights into forest dynamics, predict forest yields, guide forest management decisions, and analyze the impact of the environment on forests (Huang et al., 2019;Jandl et al., 2013;Weiskittel et al., 2011). To provide additional economic benefits and improve the ecological environment, the whole forest growth process is usually separated into different subprocesses, which have been the focus for forest experts when analyzing changes of forest growth, including tree diameter growth (Adame et al., 2008;Almeida et al., 2010), height growth (Kearsley et al., 2017;Marziliano et al., 2013), crown width growth (Fu et al., 2013;Sharma et al., 2016), volume estimation (Del Río & Sterba, 2009;Neumann & Jandl, 2005), mortality (Bohlman & Pacala, 2012;Zhao et al., 2004), and deforestation (Bavaghar, 2015;Behera et al., 2018). Therefore, based on various modeling methods, including empirical modeling and process modeling, many forest growth models have been built for these subprocesses (Korzukhin et al., 1996;Pretzsch, 2009;Weiskittel et al., 2011;Burkhart & Tomé, 2012;Adams et al., 2013). With these many different models, forest growth subprocess can be simulated; however, the overall forest growth might not be simulated effectively in this manner (Rasinmäki et al., 2009). Thus, generating a comprehensive simulation of complex forest growth is challenging. One possible solution for developing such comprehensive simulations is the integration of different subprocess-represented forest growth models.
Model integration is an efficient method to implement comprehensive simulations and solve complex problems (Argent, 2004;Chen et al., 2011;McColl & Aggett, 2007;Yu et al., 2016). In the forestry domain, there is a tendency to integrate different models for forest growth research. Several systems and frameworks have been developed that can integrate models for forest growth modeling and simulation. SIMO (Rasinmäki et al., 2009), OntoPlant (Tang et al., 2011;Tang et al., 2018), sIMfLOR (Faias et al., 2012), CAPSIS (Dufour-Kowalski et al., 2012) and DSS-SADfLOR (Garcia-Gonzalo et al., 2014) are notable examples. The OntoPlant software package can combine different individual-tree models to simulate the growth of the tree height, diameter and the number of branch whorls. sIMfLOR and DSS-SADfLOR can integrate forest growth models for forest management decision making. SIMO and CAPSIS are two integrated modeling and simulation frameworks that can integrate different types of models for forest growth. These models are desktop systems or frameworks that need to be installed by each user and simulate forest growth on a local host.
With the development of web technologies, a web environment can provide opportunities for modeling and simulation through the network (Byrne et al., 2010;Walter et al., 2018). Compared to desktop modeling and simulation systems, a web-based system has several advantages, such as cross-platform capability, ease of use, wide availability, model reuse, and information exchange (Byrne et al., 2010). Specifically, web-based modeling and simulation can increase participation from different stakeholders because of its efficient accessibility through web browsers without the need to install software and configure a runtime environment. These advantages have led to the development of a number of web-based applications in forest research. For instance, SIMANFOR is a web-based application that supports different forest growth models to promote sustainable forest management (Bravo et al., 2012), and Rammer et al. (2014) developed a web-based toolbox to support forest management. Additionally, EucaTool is used in a web environment to estimate the growth of Eucalyptus globulus plantations (Rojo-Alboreca et al., 2015). ForestMTIS can help to create web-based forest simulators, and FlorNExT is a web-based application of ForestMTIS for forest growth and thinning simulation (Gómez-García et al., 2018). However, web-based forest growth modeling and simulation still require further researches. A key point is that these web-based applications usually implement forest growth simulations using a single model without the function of model integration. The feasibility of web-based integrated modeling has been confirmed in many other fields (Belete et al., 2017;Granell et al., 2013;Laniak et al., 2013). For researches on complex forest growth, a web environment could create conditions that would enable forestry scientists and ecological experts to access various forest growth models easily and integrate them to implement a comprehensive simulation. Moreover, considering the trends of open modeling and simulation which call for the sharing and reusing of models (David et al., 2013;Moore & Tindall, 2005;Peckham & Hutton, 2009;Stasch et al., 2016), if there is a web-based standardized method for model sharing, users could share their own models on the web and reuse them for integrated forest growth simulation by others. Therefore, combining model integration with web-based forest growth is meaningful.
This article proposes a web-based integrated modeling and simulation method for forest growth. This method in a web environment supports efficient integrated modeling and simulations of complex forest growth processes. A number of forest growth models can be prepared and integrated, and a comprehensive simulation of forest growth can be implemented.
The remainder of this article is structured as follows. Section 2 describes the main steps for achieving integrated forest growth modeling and simulation. Sections 3, 4, and 5 explain the steps of the proposed method, including forest growth model preparation, model integration, and simulation implementation, respectively. The feasibility and capability of the proposed method are evaluated in section 6 with two applications. This article concludes in section 7 with further discussion.

The Basic Idea of Web-Based Integrated Modeling and Simulation
The proposed web-based integrated modeling and simulation method for forest growth consists of three main steps: model preparation, model integration, and forest growth simulation ( Figure 1).
First, forest growth models are prepared as standardized model services for web-based integrated modeling. Forest growth models play an essential role in the simulation of different forest growth process; for example, height models can simulate the height change of tree and crown width models can forecast the width of the tree crown in the future. An integrated forest growth model should integrate these types of forest growth models to represent the whole process of tree growth. However, forest growth models from various authors are generally developed using different programming languages and different model running environments. These unstandardized models are thus difficult to use in integrated modeling, which requires considerable repetitive work in programming. Therefore, forest growth models need to be wrapped into standardized models so that they have higher availability for integrated modeling. To implement a forest growth simulation on the web and promote model sharing, models also need to be published as model services so that they can be accessed in the web environment.
Second, model integration is the next main step for integrated forest growth simulation. The analysis of complex forest growth processes is necessary before integrating models since required models, data and different subprocesses need to be determined. As shown in Figure 2, to describe these subprocesses and related data, in this article, the term "model item" represents a category of models that can be used to simulate similar growth processes, and the term "data item" describes the same type of model data. In this case, every subprocess contains a corresponding model item (e.g., diameter model item and height model item) and several data items that correspond to the actual model data during simulation implementation. Subsequently, integrated forest growth modeling requires users to understand the logical relationships among those model items and data items. According to the relationships, users can design the model integration logic. Then, specific models of each model item in this logic can be chosen, and the corresponding standardized model services that are encapsulated from forest growth models can be configured on the web for the integrated forest growth simulation in the next step.
The implementation of a forest growth simulation on the web is the final step. Once those model services are configured, different forest growth models are linked as parts of an integrated model. According to the data item in the model integration logical graph, specific model data are uploaded. After that, the simulation can  be implemented on the web, and the output data of one model can be used by the next. However, because of the heterogeneity of the model data, especially the data formats, the model data are sometimes incompatible when the integrated model is running, which poses obstacles to the simulation. Therefore, data preprocessing to overcome the heterogeneity of the model data is also required for model integration. After an integrated forest growth model is completed and the model data are prepared, this model can be implemented and the simulation results can be visualized to evaluate the simulation.
It is worth mentioning that, as shown in Figure 1, an integrated forest growth simulation can be implemented on the web with the help of four support tools, including model service containers, logical graph editing tools, data service containers, and integrated modeling tools. 1. The model service container furnishes standardized model preparation, model service management and model service invocation in an open web environment (Zhang et al., 2019). Heterogeneous forest growth models can be encapsulated into standardized models based on the model encapsulation strategy (Yue et al., 2016) and deployed in the model service container as web services. This article uses a model service container developed by the Open Geographic Modeling and Simulation (OpenGMS) team (Wen et al., 2013;Wen et al., 2017;Yue et al., 2016;Zhang et al., 2019). 2. The logical graph editing tool helps to integrate forest growth models at the logical level. With this tool, forest growth model items and data items can be illustrated by graphics, and the relationships among them can be represented by arrows. The logical model integration graphs can be designed via web browsers to facilitate the integrated modeling and simulation in the next steps. This logical graph editing tool was developed based on Mxgraph (https://github.com/jgraph/mxgraph) and designs different forest growth model items and related data items. 3. The data service container can publish and manage web services for model data preprocessing.
Specifically, a data service container offers functions that include data mapping, data refactoring, and data visualization (Wang et al., 2018). Because different forest growth models generally require dissimilar data formats for input and output, these data service containers can help prepare appropriate model data by providing format-conversion-related web services for integrated modeling and simulation. The data service container used in this article was also developed by the OpenGMS team (Wang et al., 2018;Yue et al., 2015;Yue et al., 2018). 4. An integrated modeling tool can configure integrated resources and control the implementation of simulations on the web. With the help of the integrated modeling tool, forest growth model services and data preprocessing services can be accessed and configured for integrated modeling. After an integrated model is created, the progress of the forest growth simulation and the results of each part in the integrated model can be monitored by this tool. This tool was developed based on jtopo (https://github.com/wondery/jtopo) and creates connections with the model and the data service containers.

Analysis of Forest Growth Model Resources
A complex forest growth process consists of several subprocesses (i.e., tree growth, mortality and felling), which can be represented by corresponding forest growth model items. This article analyzed different forest growth model resources and classified them as several model items for integrated modeling and simulation (Table 1). Specifically, in this article, individual-tree growth models are selected to simulate forest growth processes, because they contain considerable detailed information compared with the whole-stand models (Munro, 1974). Other related models that can help forest growth simulations, including estimating individual-tree information, making decisions for felling trees, and estimating individual tree volume and biomass, are also analyzed and classified.
The individual-tree growth models provide a detailed simulation of the forest growth processes by selecting individual trees as a research object (Munro, 1974). For example, these individual-tree growth models can estimate the diameter at breast height (DBH; or basal area), tree height, under branch height (or crown ratio), tree crown width, crown height, and mortality of individual trees in a forest.
The other related models are also helpful for implementing forest growth simulations.
(1) According to the selected forest management measures, such as thinning and shelterwood cutting, the decision-making model for felling trees can calculate the felling probability of individual trees . (2) The individual-tree information estimation model can calculate individual-tree locations and DBHs based on forest stand attributes (e.g., mean DBH and dominant DBH) and the spatial distribution of the trees; thus, it can perform estimates from forest stand data to individual-tree data (Li et al., 2013;Tang et al., 2018).
(3) The individual-tree volume models and biomass models can estimate the volume and biomass of each tree in a forest based on other attributes of trees, such as the DBH and height (McRoberts & Westfall, 2013;Meng, 2006;Repola, 2009;Zhang, 2018).
To integrate and couple these models using the common data, the model data requirement needs to be analyzed. For example, the integration of the DBH model and Height model requires common data that contain the DBH attribute. These model data can be classified into different data categories, including tree attribute data, stand attribute data, location data, DEM data, and climate data (Table 2).
(1) These individual-tree growth models need various model data. The individual tree attributes (e.g., ages, species, diameters, heights, and crown heights) are generally used as input data and output data. In addition, each individual-tree growth model requires its own specific input. For example, the DBH model and mortality model may need tree locations to assess the competition intensity and DEM data to consider the aspect and slope in the models.
(2) The decision-making models for felling trees, which generally calculate the felling probabilities of trees based on the tree sizes and forest structures, requires the tree attributes and locations as input data to estimate the felling probabilities. The tree attributes are also used as output data for storing the felling results. (3) The forest stand attributes and DEM data are the input data for the individual-tree information estimation models. The location and attribute data of the individual trees can be computed and used as the output. (4) The tree attribute data are needed for the volume models and biomass models. Then, the tree attribute data, including the volume and biomass values, are the output. The result of analysis shows that these analyzed models commonly share tree attributes as their input/output data, and output data from one model can be the input data for the next. Therefore, different models can be integrated through tree attribute data.

Forest Growth Model Encapsulation
Based on the analysis results, the encapsulation of the forest growth models is the next step. In general, these forest growth models and related models are developed in different programming languages and running environments, which hinder the implementation of model integration.
In brief, integrating one model developed in C# with another developed in python is very difficult. The integration of heterogeneous models necessitates encapsulating these models into standardized models to overcome obstacles in programming languages and running environments. In addition, the forest growth models must be published as web services to be accessed in the web environment to facilitate model sharing and reuse for web-based integrated modeling.

Earth and Space Science
According to the service-oriented model encapsulation strategy and deployment strategy provided by the OpenGMS team (Wen et al., 2017;Yue et al., 2016), forest growth models can be encapsulated and deployed into a model service container as a standardized model service. This article uses the following method.
First, the forest growth models need to be described for model encapsulation so that potential model users and model service container can understand models correctly, such as determining the model purpose and types of required data. The encapsulation module used in this article helps to encapsulate the original forest growth models into standardized models based on a description in a model description language (MDL) document. The MDL document is an XML-based document that contains the AttributeSet node, Behavior node and Runtime node which describe the model metadata, behavior, and runtime demands, respectively (Yue et al., 2016). For example, an individual-tree diameter model (Qin et al., 2014) can be described by the MDL document ( Figure 3). Specifically, the AttributeSet node describes keywords and abstracts, which include the author, modeling time, forest type, original research area, and dominant tree species information. The Behavior node explains several behaviors of models, such as inputting specific forest growth model data (e.g., DEM, location, and attribute data), setting the model parameters (e.g., stand density, crown density, site index, temperature, and precipitation), and outputting the forest growth results (Figure 3). The Runtime node describes the dependency environments for model execution, including the demands in the operation system, memory size and software environment.
Second, the forest growth model needs to be wrapped as a standardized model service package. Three standardized interfaces (data interfaces, behavior interfaces and log interfaces) can be used to create the standardized model service package (Zhang et al., 2019). Specifically, the data interface handles data input and output while the behavior interface takes the responsibility of model states and events. The log interface records information during model execution, such as warnings and errors. Based on these interfaces, the encapsulated models can be controlled by the model service container and used on the web with the Finally, with the deployment strategy (Wen et al., 2013;Wen et al., 2017), the standardized models are deployed in the service container and published as web services. Then, users can access and directly use the forest growth model from model service containers via web browsers for forest growth simulation.
Based on the web-based forest growth model encapsulation, users can encapsulate their own forest growth models for web-based integrated modeling. Additionally, when new models are built, they can also be made accessible on the web using this web-based method.

Design Logic for Model Integration
The integration of forest growth models first requires the proper selection of model items and clear logical connections among them. After forest growth model encapsulation, different model services are accessible on the web and can be used for integrated forest growth modeling. However, to integrate models and implement a comprehensive simulation, the complex forest growth process must be divided into different subprocesses in advance to reduce complexity. In accordance with the process division results, the forest growth model items and related data items that correspond to these subprocesses can be selected. Subsequently, these model items and data items are also required to be integrated with an appropriate connection logic, in this case, a clear understanding of the relationships among the subprocesses is needed. For example, to estimate the forest yields, the volume estimation and biomass estimation are achieved based on the DBH and height growth simulation. Moreover, decision-making for felling trees can also influence the forest growth; thus, the subprocess of felling trees should cooperate with other subprocesses of forest growth for deforestation projects, as shown in Figure 4.
Logical graphs are useful for integrated forest growth modeling at the logic level. Selected model/data items and their relationships can be represented on logical graphs. Thus, this article developed a logical graph editing tool, by which users can select graphics to signify the logical connections among the model items and data items in a web environment. Figure 5 shows the use of the logical graph editing tool. This tool drew model items and data items (according to Tables 1 and 2) with different shapes and colors as nodes in the graph. Hence, users could easily understand the logic of different models and data items. By selecting the proper models and data based on the meaning of each node in logical graphs, the graphs of model integration can guide the model services configuration and model integration in the next step.

Integration of Forest Growth Models
In accordance with the integrated logical graph, users can select specific forest growth models of each model item for integrated forest growth modeling, and corresponding model services need to be configured and integrated on the web.

Earth and Space Science
With a designed integrated modeling tool, the forest growth model services deployed in the distributed model service containers can be accessed and configured by users. The integrated modeling tool is developed with a user-friendly graphic user interface (GUI) to assist with model selection and integration. This GUI consists of three main parts that can facilitate model service selection, selected service listing and model service configuration. In the first part, the model service containers are connected; thus, all prepared model services can be accessed and selected for model integration by the integrated modeling tool. The second part lists the selected model services. According to the model items in the logical graphs, users can choose the proper model services and add them to the list. The third part supports dragging model services into the integration scene and implementing integration by configuring model services. In particular, the integration scene is the main workspace for integrated modeling where a rectangle represents a model operation, a circle represents input/output data, and an arrow indicates the data flow in the integrated model. By setting the data flow among model operations and input/output data, model services are configured and the model integration is realized. Figure 6 shows the process of selecting and configuring model services for integrated simulation. Users fill in the host and port to connect with a model service container so that different prepared forest growth model services can be accessed. In Figure 6, there are several types of forest growth model services, including DBH models, tree height models and CW models. According to the integrated logical graph, proper model services are added to the service list, and each required forest growth model service can be dragged to the integration scene for integrated modeling and simulation. These model services are configured and different models are linked as an integrated forest growth model.

Data Preprocessing for a Forest Growth Simulation
Model data are crucial in model integration and forest growth simulations. However, for the input and output, forest growth models that shared by different modeler generally require dissimilar data formats (e.g., ESRI shapefiles, CSV, NetCDF, and Excel files), which obstructs the integrated forest growth simulation.
To overcome this difficulty of model data, data preprocessing is necessary to prepare the proper data for an integrated forest growth model. Model data format conversion is the most important while data preprocessing. The direct data format conversion with programming methods is one feasible method. But it requires considerable programming efforts to address various data formats, and these programming works are hard to subsequently reuse. In addition, with the increasing number of shared heterogeneous models, the model data will become more complex. A large number of complicated and repetitive programming works will hinder the integration of models. Therefore, a universal and reusable data preprocessing method is needed for integrated modeling and simulation.
The Universal Data eXchange (UDX) model can be used in the universal and reusable data preprocessing method for integrated forest growth modeling and simulation. The UDX model designed for the flexible description of heterogeneous model data was proposed by the OpenGMS team (Yue et al., 2015), and can help prepare proper data format for an integrated model by decreasing the cost of understanding the model data (Wang et al., 2018). For example, there are two forest models that require different data formats respectively (e.g., NetCDF and shapefile) need to be integrated. Using the UDX model, the forest growth data formats and contents are described in a uniform and structural way for easily understanding. Compared with the directly programming method, according to the description of UDX, the data conversion method can be realized easily and published as web service for sharing and reuse.
Based on the UDX model, two data preprocessing methods are developed for CSV data and ESRI shapefile data. Figure 7 shows the data format conversion between these two types of raw data and their description in the UDX model. A UDX model consists of two main components, UDX schema and UDX data. For example, data values of the tree attributes (e.g., ages, DBHs, basal areas, tree heights, crown widths) are extracted from raw data and stored in the UDX data, and the structural data description is defined in the UDX schema to

10.1029/2019EA000748
Earth and Space Science reduce the difficulty of understanding heterogeneous CSV data and ESRI shapefile data. Based on the UDX model, two different raw data types can also be converted to each other with these two components.
These data preprocessing methods based on the UDX model can be applied to two steps of the proposed webbased integrated modeling and simulation method. Therefore, two approaches are available to prepare the proper data for integrated forest growth modeling.
One approach works in the forest growth model encapsulation step. With this approach, the UDX model is used as a kind of data format to standardize the input and output data of the forest growth model. The conversion of the data format between raw data and UDX data is realized during the model encapsulation. Therefore, standardized models can use common data (UDX data) as the input and output data, and these models can be integrated directly.
The other approach is used in the forest growth model integration step. In this approach, a data preprocessing method based on the model data description in UDX model needs to be bundled into a package. After uploading the package into a data service container, a data preprocessing service used for data format conversion can be published and shared on the web (Wang et al., 2018). Then, within the integrated modeling tool, users can select proper data preprocessing services to integrate with model services. Therefore, the output data format of one model can be converted to the other data format of the following model during the period of simulation implementation.

Forest Growth Simulation and Result Visualization
When forest growth model services are configured and proper model data can be accessed, input data and model parameters need to be provided to the integrated model for a web-based forest growth simulation. Using the integrated modeling tool, the model data (e.g., tree attribute data and DEM) and related model parameters (e.g., site index, stand density, temperature, precipitation and soil) need to be set for the integrated model. Subsequently, the integrated model can be executed.
While running an integrated forest growth model, all model services and data preprocessing services (if needed) are invoked in order according to the data flow that is configured in the integrated model. With the use of the HTTP protocol, the integrated modeling tool can communicate with the model service containers and data service containers in real time. Therefore, the running status of the integrated model can be monitored to support a successful simulation. Finally, the running of the integrated model is finished, and the simulation results are obtained for further forest growth research.
Furthermore, forest growth simulation results can be visualized to support the analysis of the results. The provided visualization methods include histograms, line charts, scatter diagrams and 3D visualization. Thence, the size class distributions of trees, forest growth trends, relationships of tree attributes and 3D

10.1029/2019EA000748
Earth and Space Science forest stand scenes can be shown on the web (Figure 8). The visualization method can also support heterogeneous model data. In detail, users can upload the results of forest growth simulations, and the data preprocessing services can be used to convert the uploaded data to UDX data for data visualization.

Experiments
To apply the web-based integrated modeling and simulation method for forest growth, a web system was developed based on these four support tools. Forest growth and related models, which have been deployed in a model service container, can be accessed and used in this system (the number of encapsulated models is still increasing). Several data preprocessing methods (e.g., CSV data preprocessing and ESRI shapefile data preprocessing) are also deployed in a data service container for integrated modeling and simulation. Additionally, users can share newly developed models in a model service container, and these models can also be used for integrated modeling and simulation in this system.
This system has four main parts, including basic information, models, model data and applications ( Figure 9). The first part introduces the basic information about the models (e.g., model types and their basic principles). The second part lists all prepared model resources, and the third part shows the collection of the data that can be used in a forest growth model (e.g., DEM, forest stand and forest plot data). In the last part, users can use a single model or integrate multiple models for forest growth simulation. The simulation results can also be downloaded and visualized in this last part.
The proposed method for forest growth was verified with two applications. These applications implemented a simulation of comprehensive forest growth processes and a comparison of different integrated forest growth models for forest management.

An Application for Model Integration and Growth Simulation
This method was applied in a state-owned forest farm in Hunan province, south-central China ( Figure 10). The area is in a zone with a subtropical monsoon humid climate. The average annual temperature is 17.8°C and the average annual precipitation is 1,410.8 mm. The Chinese fir (Cunninghamia lanceolata) that is a prized timber tree and grows rapidly can adapt to this climate very well. Therefore, there are a lot of plantations of Chinese fir in this forest farm.
Some documents about the forest farm and inventory data supplied by forest managers. From these documents, the Chinese fir plantations in this farm were mainly planted for timber production. The

10.1029/2019EA000748
Earth and Space Science plantations include lots of pure even-aged forest stands whose ages are from 4 to 50 years. The inventory data contains sample plot data and forest stand data. These sample data were established in 2010 for promoting the forest management and research. And forest stand data records attributes of all forest stands in this forest farm (e.g., stand densities, stand ages, mean DBHs, and stand areas). The summary statistics of the inventory data are shown in Table 3.
In order to fell trees at appropriate time, the requirement that is to predict the approximate growth states and yields of forest stands via a comprehensive simulation were established by the forest managers.
Without installing a special system and configuring related runtime environments for the comprehensive forest growth simulation, the proposed method is adopted in this application. We first selected several forest growth model items and related model items, including DBH-Height (defined by users), CW, HUB, individual-tree information estimation, volume, and biomass, and data items to analyze the forest growth process and design the integration logic. Figure 11 shows the logical integration of these forest stand growth simulations.
Based on the logic graph, specific forest growth models, which were developed for the Chinese fir in southcentral China (Li et al., 2013;Lv, 2002;Ma et al., 2018), are selected. After model encapsulation, the corresponding model services are configured, as shown in Figure 12. Because the selected individual-tree information estimation model uses CSV data as output data, to integrate this model with other models that use ESRI shapefile data, data preprocessing services that can convert CSV data to shapefile data must be configured and integrated with model services, as shown in Figure 12.
After the integrated forest growth modeling and data uploading, the comprehensive simulation is implemented, and the results are obtained. Because the original forest stand attributes (e.g., stand spatial structure and density) are different, the estimated growth states are dissimilar. Simulation results of the integrated model for three 10-year-old forest stands are shown in Table 4 as an example. It shows the mean DBH, stand density, and stand area at the age of 10. The mean DBH, tree height, CW, HUB, tree volume, and tree biomass of these forest stands at 30 years old are also shown. Abbreviation: DBH: diameter at breast height. Figure 11. Logical integration of these forest stand growth simulations.

10.1029/2019EA000748
Earth and Space Science MA ET AL. 2154

An Application for the Comparison of Different Forest Management Measures
Different forest management measures have dissimilar influences on the forest growth, due to the influence on competition. Therefore, a good forest management decision-making is very important. Still working with the state-owner forest farm in Hunan province, apart from the comprehensive prediction of forest yield, the comparison of different forest management measures is also implemented using the proposed method. In this application, two simple forest management measures are compared for verifying the comparison capability. In the first measure, the thinning in the 10th year with a thinning intensity of 15% needs to be conducted. The other measure involves selective felling in the 15th year, and all trees whose DBH is less than 10 cm will be selected for felling.
Two new decision-making models for felling trees that correspond to those two forest management measures must be used. With the proposed method, these two models are encapsulated in model service containers for integrated modeling on the web. After the encapsulation of these two models, under the guidance of the web-based integrated modeling and simulation steps, the integration logic of two forest management measures are analyzed, and the model services are configured as two integrated models for forest management measure comparison, as shown in Figure 13. Then the simulations are implemented, and the results are visualized for comparison ( Figure 14). Figure 14 shows the influences of these forest management measures on the forest growth from the 10th year to the 20th year. Specifically, in the simulation results of integrated model 1, there are 143 felled trees, and the maximum DBH, minimum DBH, and mean DBH increase to 37.26, 18.94, and 25.62, respectively. In the simulation results of integrated model 2, the number of felled trees is 259, and the maximum DBH,  minimum DBH, and mean DBH increase to 36.03, 21.15, and 25.24, respectively. And the distributions of the DBH class are also dissimilar. In summary, following the second decision, more trees are felled, but the mean DBH is smaller. It is easy to explain the reason that the number of trees is changed due to the different felling activities. For the difference of the DBHs, due to the earlier felling activity in the first measure, trees can grow with less competition so that they have a better yield at the end. And because there is a selective felling activity in the 15th year according to the second measure, the measure has an obvious effect on the distribution of DBH classes. The selective felling removes all small DBH trees, which make the DBH class distribution different from the normal distribution in the result of integrated model 1.

Conclusion
This article proposes a web-based integrated modeling and simulation method for forest growth. The method can be used in a web environment through three steps, including model preparation, model integration and forest growth simulation. Compared with existing forest growth systems and frameworks, the proposed method combines the advantages of integrated modeling methods and web systems, including the cross-platform capability, operating system independence, and the ability for complex process simulation.
It not only supports a web-based simulation for forest growth, but realizes the integrated modeling for a complex forest growth process. Moreover, with this method, each forest growth modeler can share their own models for model reuse. The feasibility of the proposed method has been verified by two applications. The two applications show that heterogeneous forest growth models can be integrated conveniently on the web and that forest growth simulations can be implemented successfully.
This research can be extended in the future for more complex forest growth processes.
1. The preparation of forest growth models needs to be more convenient for comprehensive forest growth simulation. Although a large number of forest growth models have been created by forestry and ecology experts, this article only analyzed and encapsulated limited forest growth models. Thus, additional forest growth models are still required to be encapsulated into model service containers by users. Additionally, this proposed method also needs to be applied and verified by more complex forest researches. Therefore, for convenient forest growth model preparation, some web-based tools that can help in the web-based encapsulation need to be developed in the future. 2. For a comprehensive forest growth simulation, the web-based integrated modeling and simulation method supports collaborative work among forestry scientists and ecological experts (Chen et al., 2012). Based on network communication technologies, this collaboration helps users to establish more appropriate integrated forest growth models and make better forest management decisions. Specifically, this proposed method allows the design of integration logic, the configuration of model services, and the setting of model parameters collaboratively on the web. This collaborative method of integrated modeling and simulation for forest growth still needs to be researched.