Research on Construction and Application of Water Processes Based on Knowledge Graph: Analysis of Dynamic Paths and Impact Factors

Abstract

The water process refers to the movement and changes in water on Earth, encompassing changes among its three states and its spatial movement. This process is vital for human society as it directly influences water resources, environmental sustainability, and climate regulation. Previous studies have used various related factors to analyze the water process but have not explained the rationale behind selecting these factors from the perspective of pathways. Based on this, the paper explores the construction and application of a top-down water process knowledge graph to clarify the changing process of water movement and the sources of influencing factors. Firstly, we define the concept of the water process and classify its entities based on the concept of water boundaries. Secondly, we identify key knowledge components of the water process, including water bodies, processes, and influencing factors. Finally, we construct and analyze a knowledge graph of the water process and its influencing factors. Results show that (1) the paths of water process help us understand the movement and change process of the water bodies; (2) the number of paths increases with the length of the connection between entities, reflecting the complexity of water process relationships; and (3) tracing these pathways can help identify their influencing factors, providing a data foundation for applying deep learning algorithms in water process research.

1. Introduction

Water is not only the foundation of survival and civilization but also the important support and basic guarantee of economic and social development. The water process is the movement and change process of the water body on the earth, which mainly includes the three-state change in the water body and its position movement. With the development of digital society and the rapid progress of artificial intelligence technology, more and more studies began to pay attention to the water process. This is because the study of the water process can not only more effectively manage water resources but also better protect the water environment and predict and prevent natural disasters such as floods and droughts.

At present, the water process research can be mainly divided into two types, including process-based models and data-driven models. The process-based models are considered the powerful tools to study the water process, with the capability of simulating both spatial and temporal distribution patterns based on coupled physical process and the interactions between different state variables [1]. However, they require large computational resources to compute the complex interactions between state variables in the different processes [2], and the previous assumptions no longer meet the current demand. By contrast, data-driven models only depend upon historical hydro-meteorological data without directly considering underlying physical processes and, thus, entailing much less input and parameter data [3] and have been widely used in the water process. These methods include multiple linear regression (MLR) [4], back propagation neural network (BPNN) [5], multilayer perceptron (MLP) [6], support vector regression (SVR) [7,8], and extreme learning machine (ELM) [9]. However, they belong to the “shallow” learning category, in which the instinct information represents insufficiently.

To overcome the problem, deep learning techniques, which can gain better forecasting performance owing to its “deeper” representations, are attracted much attention in hydrological prediction [3]. For example, Yue et al. [10] adopted the deep belief network (DBN)with partial least-squares regression (PLSR) and used key factors obtained by the partial mutual information (PMI) as its input for mid-to long-term runoff prediction. Gao et al. [11] applied three models including LSTM, Gated Recurrent Unit (GRU), and artificial neural network to simulate runoff in the Yutan station control catchment, and showed that LSTM and GRU have better performance. Lv et al. [12] proposed a long Short-Term memory cyclic model with mutual information for forecasting flood flow of Xixian Basin, China. Hu et al. [13] proposed a method based on a convolutional neural network-bidirectional long short-term memory-difference analysis (CNN-BiLSTM-DA) model for water level prediction analysis and flood warning. Even though, due to the features of deep learning, e.g., high accuracy, high efficiency, low cost, and low barriers, the above-mentioned models offer not only highly accurate results but orders-of-magnitude lower cost in terms of both model preparation/validation and run-time computation efforts; however, it is difficult to provide an easy-to-understand explanation for its results [14]. In addition, there is no physical water routing process being formulated in above-mentioned models, which leads to insufficient information describing the water process and affects the accuracy of the model [15,16]. Thus, to make the model explainable, this paper studies water process based on physical mechanism. On this basis, all the data describing the water process are given to obtain sufficient information.

Knowledge graph (KG) is a graph-based data structure used to describe concepts and their relationships in the physical world. It can mine entity relationships from vast, fragmented information and present them as structured semantic networks [17]. At present, it has been widely used in various fields, such as in the field of fault diagnosis, Tang et al. [18] proposed a new intelligent auxiliary diagnosis method based on knowledge graph which helps grass-roots maintenance engineers to quickly and accurately locate the fault unit of aircraft. In the field of biomedicine, Lin et al. [19] proposed a KG-based neural network framework to solve the problem of drug–drug interaction prediction. Zeng et al. [20] summarized knowledge graph-based works that implement drug repurposing and adverse drug reaction prediction for drug discovery. In addition, there are a few studies in the field of water conservancy. Feng et al. [21] proposed a knowledge graph construction method for water conservancy object data, aiming at the problem that it is difficult for water conservancy industry to use the relationship between objects for information retrieval based on keyword search technology. Duan et al. [22] put forward the description method of water conservancy comprehensive knowledge system, constructed the construction framework and key technology system of water conservancy knowledge graph, and realized the cross-domain query and retrieval of water conservancy knowledge query the relationship between discipline graph and water network graph, and mine the implicit relationship between different water conservancy entities. Rondon Diaz and Vilches-Blazquez [23] characterized water quality datasets through multi-dimensional knowledge graphs for bridging data silos. However, to our knowledge, there are no studies that apply knowledge graph to water motion change. Hence, to obtain sufficient information, it is necessary to carry out the research based on the knowledge graph of water process.

The important point for solving the above problems is to determine the water body and its relations that describe the water process. For this reason, based on the top-down scheme [24], this paper firstly gives the relevant definition of the effluent process and divides it according to the different boundaries of the carrying water body. Secondly, the knowledge of water process is extracted, including water body, process and influence factors. Then, the knowledge graph of water process and the knowledge graph of influence factors of water process are constructed. Finally, the path and influence factors of water process are analyzed. This can further provide new insights into the understanding of water processes.

2. Methods

2.1. Problem Formalization

For the convenience of the study, this paper makes some hypotheses and gives some definitions. The specific contents are as follows:

Suppose that the water body is enclosed in a space, and it can undergo three-state change by itself, but it will not disappear. In addition, the water process mentioned in this paper is assumed to be continuous.

Definition 1.

The water process is the movement and change process of the water body on the earth. Among them, motion refers to the position change of water body, such as water body from point a to point b; change refers to the three-state transformation of water body, such as the transformation of water body from solid to liquid.

Definition 2.

The water body is a water structure composed of unit water in a certain range, which is composed of gas–liquid-solid single state or multiple states coexisting unit water. In addition, due to the different boundaries of the carrying water bodies, various water bodies such as rivers and lakes have been formed. This will prepare for the subsequent division of the water body.

Definition 3.

The water process KG can be represented as WP = (P, E) where P is the set of water process entities, and E is the set of their relations.

Definition 4.

The water process path refers to the spatial trajectory left by the change of the position of the water body. A path $M_{1,L}$ defines a sequence of entity types and relationship types. Taking $M_{1,L}=A_1\stackrel{R_1}{\to }{A}_2\stackrel{R_2}{\to }\cdots \stackrel{R_{L-1}}{\to }{A}_L$ as an example, it indicates that there is a compound relationship ${(R}_1, R_2, \cdots , R_{L-1})$ between entity type $A_1$ and $A_L$. Where $A_i (i=1, 2, \dots , L)$ is the entity and $R_j (j=1, 2, \dots , L-1)$ is the relationship [25].

Definition 5.

The water process elements are the factors that affect the movement of water bodies.

2.2. Overall Structure

This paper elaborated the relevant definition of water process and the architecture of knowledge graph, as well as its application in water process. Figure 1 shows the overall structure and includes four compounds: water process knowledge, related definitions of water process and their relationships, water process knowledge extraction and fusion, and knowledge graph in water process and analysis.

2.2.1. Water Process Knowledge

For the construction of water process KG, this paper gives some definitions of water process (as shown in Section “Problem formalization”) and selects three categories of texts, including literature data, network data, and book data, which are collected. Considering the reliability and stability of knowledge, network data mainly includes some encyclopedia web pages whose contents are more reliable and stable; book data includes the authoritative books in the field of hydrology, such as Handbook of Hydrology [26]. All network data and literature data are chosen based on Chinese and English keywords about the water process; for example, they describe the three-state changes in water body, such as “evaporation”, “sublimation”, etc.; the movement of water body, such as “recharge”, etc.; and the factors that affect the water process, such as “evaporation factors”, etc.

2.2.2. Water Process Knowledge Extraction

Water process knowledge extraction is a process of identifying entity and relationship types from both structured and unstructured water process documents. In the construction of water process KG, we extract the carrier of water body as entities, such as rivers, lakes, etc., and extract the movement and change process of water bodies as relations, such as evaporation, infiltration, condensation, etc., and extract the factors that affect the movement of water bodies as attributes, such as temperature, air pressure and so on. In addition, according to the difference in water boundary and conceptual semantic relationship [27], this paper classifies the entities in the water process.

2.2.3. Knowledge Graph in Water Process and Analysis

As we are dealing with entities and their relationships, it only makes sense to store the results in a graph database. Graph databases are used as a warehouse to store KGs in terms of nodes, edges, and its properties of graphs [28]. In this paper, we use the Neo4j diagram database to describe the water process. The Neo4j diagram database is an opensource graph database with high performance, high reliability, and expandability. It contains two basic data types: nodes and edges, and both nodes and edges can contain various forms of attributes. Among them, the node is equivalent to the entity in the water process and is the vertex in the graph; the edge is equivalent to the relationship between the two entities in the water process and is the edge in the graph; Attributes are the characteristics of nodes and relationships, that is, elements that describe water processes, which are stored in neo4j in the form of key-value pairs. Moreover, we use the triplets of <entity-relation-entity> to form a semantic network to represent the water process.

In addition, based on graph theory search algorithm, Neo4j graph database provides a set of descriptive Cypher query language, which mainly includes three parts: MATCH sentence to realize graph matching, WHERE sentence to realize filter condition, and RETURN sentence to realize result return. Among them, path search refers to traversing along a specific edge type from a specified node and returns the terminating nodes and edges that can be reached. The query is translated into Cypher as “MATCH Paths = (p1: Entity {name: ‘Head Entity’})-[r*n]-> (p2: Entity {name: ‘Tail Entity’}) RETURN Paths”, where n represents the length between two entities, which will help us to study the water process path.

The application of knowledge graph in water process is mainly in two aspects: (1) water process path. The knowledge graph clearly shows the movement and change process of the water body and can provide the path between any two water bodies, which will help to find the source that affects the water quality. (2) water process element. Combined with the water process path, we will obtain all the influencing factors of the water process, which will provide a data basis for using artificial intelligence to study the water process.

3. Results

To achieve this work, this study employed graph database technologies using Neo4j (version 5.15) and Python 3.11 (with Scrapy 2.11 and Py2neo 2021.2.3) as the computational platform. The hardware environment consisted of a Windows 11 workstation equipped with an AMD Ryzen ™ 9 7945HX processor (5.4 GHz boost) and 32 GB RAM, Santa Clara, CA, USA. In addition, during the data-acquisition phase, the data collected in this study primarily draws from Handbook of Hydrology. Through systematic knowledge integration, we have constructed an extensible framework.

3.1. The Water Process KG

Figure 2 shows the water process KG, where arrows of different colors represent that two water bodies are connected in different states. For example, the purple arrow indicates that two water bodies are connected in a solid state. And

Table 1. Division results of water process entities.

No.	Concept I	Concept A	Concept B	Concept Entity
1	Water body	Sky	Water vapor	Water vapor
2			Cloud	Water cloud
3				Mixed cloud
4				Halicloud
5			Falling water	Rain
6				Snow
7				Hail
8		Surface	Surface water	River
9				Lake
10				Reservoir
11				Swamp
12				Ocean
13				Glacier
14				Perpetual snow
15			Biowater	Green plants
16			Human water use	Agricultural water
17				Industrial water
18				Domestic water
19		Underground	Unsaturated zone	Soil water zone
20				Intermediate hose
21				Capillary zone
22			Saturated zone	Phreatic water
23				Confined water

shows the results of the classification of water process entities according to conceptual semantic relations [27], which include “Water vapor”, “Water cloud”, “Mixed cloud”, “Halicloud”, “Rain”, “Snow” and other entities.

Table 2. Definition of the entity relation type.

Relation Type	Connotation	Head Entity	Tail Entity
Condensation_collision_growth	Indicating that water clouds form rain through condensation, collision and growth.	Water cloud	Rain
Condensation_collision_growth	Indicating that mixed clouds form snow through condensation, collision and growth.	Mixed cloud	Snow
Condensation_collision_growth	Indicating that halicloud forms hail through condensation, collision and growth.	Halicloud	Hail
Landing_L	Indicating that the rain falls from the sky to the river.	Rain	River
Landing_S	Indicating that the snow falls from the sky to the river.	Snow	River
Landing_S	Indicating that the hail falls from the sky to the river.	Hail	River
Surfacerunoff	Indicating that rain enters the river in the form of surface runoff.	Rain	River
Surfacerunoff	Indicating that rain enters the lake in the form of surface runoff.	Rain	Lake
Landing_L	Indicating that the rain falls from the sky to the lake.	Rain	Lake
Landing_S	Indicating that the snow falls from the sky to the lake.	Snow	Lake
Landing_S	Indicating that the hail falls from the sky to the lake.	Hail	Lake
Landing_L	Indicating that the rain falls from the sky to the reservoir.	Rain	Reservoir
Landing_S	Indicating that the snow falls from the sky to the reservoir.	Snow	Reservoir
Landing_S	Indicating that the hail falls from the sky to the reservoir.	Hail	Reservoir
Landing_L	Indicating that the rain falls from the sky to the swamp.	Rain	Swamp
Landing_S	Indicating that the snow falls from the sky to the swamp.	Snow	Swamp
Landing_S	Indicating that the hail falls from the sky to the swamp.	Hail	Swamp
Surfacerunoff	Indicating that rain enters the swamp in the form of surface runoff.	Rain	Swamp
Surfacerunoff	Indicating that rain enters the ocean in the form of surface runoff.	Rain	Ocean
Landing_L	Indicating that the rain falls from the sky to the ocean.	Rain	Ocean
Landing_S	Indicating that the snow falls from the sky to the ocean.	Snow	Ocean
Landing_S	Indicating that the hail falls from the sky to the ocean.	Hail	Ocean
Landing_L	Indicating that the rain falls from the sky to the glacier.	Rain	Glacier
Landing_S	Indicating that the snow falls from the sky to the glacier.	Snow	Glacier
Landing_S	Indicating that the hail falls from the sky to the glacier.	Hail	Glacier
Landing_L	Indicating that the rain falls from the sky to the perpetual snow.	Rain	Perpetual snow
Landing_S	Indicating that the snow falls from the sky to the perpetual snow.	Snow	Perpetual snow
Landing_S	Indicating that the hail falls from the sky to the perpetual snow.	Hail	Perpetual snow
Interception	Indicating that the parts of trees and plants cut off by the rain before it hits the ground.	Rain	Green plants
Interception	Indicating that the parts of trees and plants cut off by the snow before it hits the ground.	Snow	Green plants
Diversion_irrigation	Indicating that agriculture is irrigated with river water.	River	Agricultural water
Diversion_irrigation	Indicating that agriculture is irrigated with reservoir water.	Reservoir	Agricultural water
Diversion_irrigation	Indicating that agriculture is irrigated with lake water.	Lake	Agricultural water
Irrigation_reosmosis	Indicating that agricultural water is called soil water through irrigation.	Agricultural water	Soil water zone
Water_diversion	Indicating that the river water is diverted to domestic use.	River	Domestic water
Water_diversion	Indicating that the lake water is diverted to domestic water.	Lake	Domestic water
Water_diversion	Indicating that the reservoir water is diverted to domestic use.	Reservoir	Domestic water
Processing_discharge	Indicating that domestic water is processed and discharged into the river.	Domestic water	River
Water_diversion	Indicating that the river water is used for production.	River	Industrial water
Water_diversion	Indicating that the reservoir water is used for production.	Reservoir	Industrial water
Water_diversion	Indicating that the lake water is used for production.	Lake	Industrial water
Water_diversion	Indicating that the ocean water is used for production.	Ocean	Industrial water
Processing_discharge	Indicating that industrial water is processed and discharged into the river.	Industrial water	River
Intercept	Indicating that the river has been intercepted and entered the reservoir.	River	Reservoir
Flood-relief	Indicating that the reservoir water is discharged into the river through flood discharge.	Reservoir	River
Inflow	Indicating that the river flows into the lake.	River	Lake
Inflow	Indicating that river water flows into the ocean.	River	Ocean
Subsurfaceflow	Indicating that soil water enters the river in the form of soil flow.	Soil water zone	River
Stream_runoff	Indicating that glaciers melt into water and flow into rivers.	Glacier	River
Stream_runoff	Indicating that the snow melts into water and flows into the river.	Perpetual snow	River
Infiltration	Indicating that soil water can be changed into intermediate hose by infiltration.	Soil water zone	Intermediate hose
Infiltration	Indicating that intermediate hose water can be changed into capillary zone water by infiltration.	Intermediate hose	Capillary zone
Infiltration	Indicating that capillary zone can be changed into phreatic water by infiltration.	Capillary zone	Phreatic water
Infiltration	Indicating that phreatic water can be changed into confined water by infiltration.	Phreatic water	Confined water
Shallow_subsurface_runoff	Indicating that diving enters the river in the form of shallow subsurface runoff.	Phreatic water	River
Deep_subsurface_runoff	Indicating that the confined water enters the river in the form of deep underground runoff.	Confined water	River
Evaporation	Indicating that the river water changes from liquid to water vapor by evaporation.	River	Water vapor
Evaporation	Indicating that the lake water changes from liquid to water vapor by evaporation.	Lake	Water vapor
Evaporation	Indicating that the reservoir water changes from liquid to water vapor by evaporation.	Reservoir	Water vapor
Evaporation	Indicating that the ocean water changes from liquid to water vapor by evaporation.	Ocean	Water vapor
Evaporation	Indicating that the soil water changes from liquid to water vapor by evaporation.	Soil water zone	Water vapor
Evaporation	Indicating that the swamp water changes from liquid to water vapor by evaporation.	Swamp	Water vapor
Transpiration	Indicating that green plants turn water into water vapor by transpiration.	Green plants	Water vapor
Sublimation	Indicating that glaciers change solid water into vapor in the form of sublimation.	Glacier	Water vapor
Sublimation	Indicating that snow changes solid water into vapor in the form of sublimation.	Perpetual snow	Water vapor
Condensation	Indicating that water vapor changes into water and clouds through condensation.	Water vapor	Water cloud
Desublimation	Indicating that water vapor changes into mixed clouds through desublimation.	Water vapor	Mixed cloud
Condensation_desublimation	Indicating that water vapor turns into hail clouds by sublimation_condensation.	Water vapor	Halicloud

illustrates the definitions of corresponding entity types and relationships.

3.2. Water Process Path

Through the above analysis, we already know that the water process is very complicated. Therefore, to make clear the water source and its movement change process, we use the path to represent it. Due to the limited space, this paper takes the head entity “Water vapor”(id = 1) and the tail entity “River” (id = 8) as the research object and finds the path with length 3 based on the query language “match Paths= (p1: Entity {name: ‘Water vapor’})-[r*3]-> (p2: Entity {name: ‘River’}) return Paths”. The results are as follows:

$$\begin{gathered} {\mathrm{M}}_{\mathrm{1,8}}^1=\mathrm{W}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{r} \mathrm{v}\mathrm{a}\mathrm{p}\mathrm{o}\mathrm{r}\stackrel{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{n}\mathrm{s}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}}{\to }\mathrm{W}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{r} \mathrm{c}\mathrm{l}\mathrm{o}\mathrm{u}\mathrm{d}\stackrel{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{n}\mathrm{s}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\_\mathrm{c}\mathrm{o}\mathrm{l}\mathrm{l}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{o}\mathrm{n}\_\mathrm{g}\mathrm{r}\mathrm{o}\mathrm{w}\mathrm{t}\mathrm{h}\_\mathrm{L}}{\to }\mathrm{R}\mathrm{a}\mathrm{i}\mathrm{n}\stackrel{\mathrm{L}\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{n}\mathrm{g}\_\mathrm{L}}{\to }\mathrm{R}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{r}; \\ {\mathrm{M}}_{\mathrm{1,8}}^2=\mathrm{W}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{r} \mathrm{v}\mathrm{a}\mathrm{p}\mathrm{o}\mathrm{r}\stackrel{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{n}\mathrm{s}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}}{\to }\mathrm{W}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{r} \mathrm{c}\mathrm{l}\mathrm{o}\mathrm{u}\mathrm{d}\stackrel{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{n}\mathrm{s}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\_\mathrm{c}\mathrm{o}\mathrm{l}\mathrm{l}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{o}\mathrm{n}\_\mathrm{g}\mathrm{r}\mathrm{o}\mathrm{w}\mathrm{t}\mathrm{h}\_\mathrm{L}}{\to }\mathrm{R}\mathrm{a}\mathrm{i}\mathrm{n}\stackrel{\mathrm{S}\mathrm{u}\mathrm{r}\mathrm{f}\mathrm{a}\mathrm{c}\mathrm{e}\mathrm{r}\mathrm{u}\mathrm{n}\mathrm{o}\mathrm{f}\mathrm{f}}{\to }\mathrm{R}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{r}; \\ {\mathrm{M}}_{\mathrm{1,8}}^3=\mathrm{W}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{r} \mathrm{v}\mathrm{a}\mathrm{p}\mathrm{o}\mathrm{r}\stackrel{\mathrm{D}\mathrm{e}\mathrm{s}\mathrm{u}\mathrm{b}\mathrm{l}\mathrm{i}\mathrm{m}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}}{\to }\mathrm{M}\mathrm{i}\mathrm{x}\mathrm{e}\mathrm{d} \mathrm{c}\mathrm{l}\mathrm{o}\mathrm{u}\mathrm{d}\stackrel{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{n}\mathrm{s}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\_\mathrm{c}\mathrm{o}\mathrm{l}\mathrm{l}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{o}\mathrm{n}\_\mathrm{g}\mathrm{r}\mathrm{o}\mathrm{w}\mathrm{t}\mathrm{h}\_\mathrm{S}}{\to }\mathrm{S}\mathrm{n}\mathrm{o}\mathrm{w}\stackrel{\mathrm{L}\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{n}\mathrm{g}\_\mathrm{S}}{\to }\mathrm{R}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{r}; \\ {\mathrm{M}}_{\mathrm{1,8}}^4=\mathrm{W}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{r} \mathrm{v}\mathrm{a}\mathrm{p}\mathrm{o}\mathrm{r}\stackrel{\mathrm{D}\mathrm{e}\mathrm{s}\mathrm{u}\mathrm{b}\mathrm{l}\mathrm{i}\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{n}\mathrm{s}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\_\mathrm{d}\mathrm{e}\mathrm{s}\mathrm{u}\mathrm{b}\mathrm{l}\mathrm{i}\mathrm{m}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}}{\to }\mathrm{H}\mathrm{a}\mathrm{l}\mathrm{i}\mathrm{c}\mathrm{l}\mathrm{o}\mathrm{u}\mathrm{d}\stackrel{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{n}\mathrm{s}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\_\mathrm{c}\mathrm{o}\mathrm{l}\mathrm{l}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{o}\mathrm{n}\_\mathrm{g}\mathrm{r}\mathrm{o}\mathrm{w}\mathrm{t}\mathrm{h}\_\mathrm{S}}{\to }\mathrm{H}\mathrm{a}\mathrm{i}\mathrm{l}\stackrel{\mathrm{L}\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{n}\mathrm{g}\_\mathrm{S}}{\to }\mathrm{R}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{r}. \end{gathered}$$

Among them, ${\mathrm{M}}_{\mathrm{1,8}}^{\mathrm{i}}$ represents the $\mathrm{i}$-th (here $\mathrm{i}=4$) path from “Water vapor” to “River”. The above examples of different paths (there are four.) between “Water vapor” and “River” show that it is feasible for scholars to construct input data from multiple perspectives for research. The KG of sub-water process connecting water vapor and river is shown in Figure 3. As can be seen from Figure 3, the water entering the river is in the form of “rain”, “snow”, and “hail”. If the problem of river water quality is studied, it can be seen from the above forms that the source is likely to be surface runoff. Therefore, it is particularly important to study the path of water process.

Additionally, to figure out what happens when the path length continues to increase, we increase the path (connecting “Water vapor” to the “River”) length to 5, and the result is shown in Figure 4. Comparing Figure 3 and Figure 4, we can see that with the increase in the length of the connection between “Water vapor” and “River”, the more entities and relationships appear, the more the number of paths. And, to increase the persuasion of the results, we increase the path length of connecting different head entities and tail entities from 1 to 9, and count the number of paths, as shown in

Table 3. Definition of the tertiary relation type.

Head Entity (id)	Water Vapor (1)	Water Vapor (1)	River (8)
Tail Entity (id)	River (8)	Ocean (12)	Ocean (12)
Paths by length	1 → 8	1 → 12	8 → 12
1	0	0	1
2	0	0	0
3	4	4	3
4	6	4	9
5	30	6	24
6	46	30	48
7	165	103	151
8	348	205	355
9	1036	467	894

and Figure 5. As can be seen from Table 3 and Figure 5, the number of paths also increases as the path length between the two entities increases. In addition, when the path length exceeds 6, the number of paths from the head entity to the tail entity increases rapidly.

3.3. Water Process Elements

Due to the limited space, this section takes

$$“{\mathrm{M}}_{\mathrm{8,8}}=\mathrm{River}\stackrel{\mathrm{E}\mathrm{v}\mathrm{a}\mathrm{p}\mathrm{o}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{s}\mathrm{s}}{\to }\mathrm{Water} \mathrm{vapor}\stackrel{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{n}\mathrm{s}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}}{\to }\mathrm{W}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{r} \mathrm{c}\mathrm{l}\mathrm{o}\mathrm{u}\mathrm{d}\stackrel{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{n}\mathrm{s}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\_\mathrm{c}\mathrm{o}\mathrm{l}\mathrm{l}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{o}\mathrm{n}\_\mathrm{g}\mathrm{r}\mathrm{o}\mathrm{w}\mathrm{t}\mathrm{h}\_\mathrm{L}}{\to }\mathrm{Rain}\stackrel{\mathrm{L}\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{n}\mathrm{g}\_\mathrm{L}}{\to }\mathrm{River}”$$

a path from the “River” to “River” as an example, to give the influencing factors of the effluent process. To show the relationship between elements and water process, this section defines water body, process and influence factors as entities, between entities and processes, between entities and influence factors, and between processes and influence factors as relationships. Thus, the path ${\mathrm{M}}_{\mathrm{8,8}}$ becomes ${\mathrm{M}}_{\mathrm{8,8}}^{\prime }$, and the specific ${\mathrm{M}}_{\mathrm{8,8}}^{\prime }$ is

$$\begin{gathered} “{\mathrm{M}}_{\mathrm{8,8}}^{\prime }=\mathrm{River}\stackrel{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{c}\mathrm{e}\mathrm{s}\mathrm{s}}{\to }\mathrm{E}\mathrm{v}\mathrm{a}\mathrm{p}\mathrm{o}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\stackrel{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{c}\mathrm{e}\mathrm{s}\mathrm{s}}{\to }\mathrm{Water} \mathrm{vapor}\stackrel{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{c}\mathrm{e}\mathrm{s}\mathrm{s}}{\to }\mathrm{Condensation}\stackrel{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{c}\mathrm{e}\mathrm{s}\mathrm{s}}{\to }\mathrm{W}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{r} \mathrm{c}\mathrm{l}\mathrm{o}\mathrm{u}\mathrm{d} \\ \stackrel{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{c}\mathrm{e}\mathrm{s}\mathrm{s}}{\to }\mathrm{Condensation}\_\mathrm{collision}\_\mathrm{growth}\_\mathrm{L}\stackrel{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{c}\mathrm{e}\mathrm{s}\mathrm{s}}{\to }\mathrm{Rain}\stackrel{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{c}\mathrm{e}\mathrm{s}\mathrm{s}}{\to }\mathrm{Landing}\_\mathrm{L}\stackrel{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{c}\mathrm{e}\mathrm{s}\mathrm{s}}{\to }\mathrm{River}” \end{gathered}$$

Figure 6 shows the KG of water process with influence factors, in which “blue” represents the influence factors of entity and process, “green” is liquid water body, “red” is gaseous water body, and “curry” color is the movement and change process of water body. From Figure 6, we can see that it takes 25 elements, 4 processes, and 4 different bodies of water to start from the river and then return to the river. This means that to calculate the amount of rainfall that falls into a river in the form of rain, it requires at least 25 input factors, four processes and four different carriers. This also indicates that the water process can be regarded as the propagation chain of the water body, and if one of the processes is interrupted or some influencing factor cannot reach the value of the change, it will lead to the whole transmission. In such cases, it is necessary to study every component of the communication chain. Therefore, here we assume that the process of water is continuous. In addition, we found that processes and entities have the same influence factors, such as the “water area” is not only the factor (attribute) of the river, but also the factor affecting evaporation. This is very critical for us to study the location of the input factors later.

4. Discussion

We have constructed a knowledge graph of water processes and analyzed the paths between any two water bodies and the elements driving the movement and change in water bodies. Several studies in the literature analyze related physical water processes using knowledge graphs [21,22,23]. Their core similarity is that they both study water by using the method of knowledge graphs, but differ in scope: Literatures [21,22] mainly study the construction of the knowledge graph of water conservancy information; Literature [23] primarily studies water quality through knowledge graphs, which is a relatively specific water process. This paper investigates the dynamic processes of water bodies across different states, including: (1) transition paths between any two water bodies; (2) the elements that can drive the movement and change in water bodies based on these paths.

In addition, to demonstrate the practical applicability of our framework to water processes, we apply the concept-instance relational principle from Reference [21]—which states that relations between conceptual entities imply analogous instance-level linkages. Focusing on the Qinghai Lake Basin as a specific watershed, we construct a knowledge graph for the Qinghai Lake Basin. The Qinghai Lake Basin, also known as the Qinghai Lake Basin, is an independent and complete physical geography unit. Its geographical coordinates range from 97°4′55.70″ E to 101°11′24.88″ E and 36°17′43.58″ N to 38°19′16.20″N, with elevations spanning 3195 to 5291 m [29].

Firstly, within the Qinghai Lake Basin, conceptual entities were mapped to instance entities. Due to space limitations, the instance entities for the Qinghai Lake Basin were primarily referenced from Literature [29], with the mapping results presented in

Table 4. The Mapping Relationship between Conceptual Entities and Instance Entities within the Qinghai Lake Basin.

No.	Concept Entity	Instance Entity	No.	Concept Entity	Instance Entity
1	Water vapor	Water vapor	14	Perpetual snow	Zanbaohuaixiu Mountain Snowpack
2	Water cloud	Water cloud			Riyue Mountain Snowpack
3	Mixed cloud	Mixed cloud	15	Green plants	Splendid Achnatherum
4	Halicloud	Halicloud			Leymus
5	Rain	Rain			Short-flower Needlegrass
6	Snow	Snow			Purple Needlegrass
7	Hail	Hail			Drooping Wildrye
8	River	Buha River	16	Agricultural water	Agricultural Irrigation
		Shaliu River			Livestock Watering
		Quanji River	17	Industrial water	Agro-livestock Product Processing
		Haergai River			Clean Energy Facility Maintenance
		Ganzi River	18	Domestic water	Residential Water Use
		Daotang River			Tourism Service Water Use
		Heima River			Pastoral Settlement Water Supply
9	Lake	Qinghai Lake	19	Soil water zone	Gelic Leptosols
		Erhai Lake			Histic Cambisols
		Gahai Lake			Mollic Planosols
10	Reservoir	Dongwei Reservoir			Calcic Kastanozems
		Heima River Regulating Reservoir			Gleysols
		Tanggema Embankment Dam			Arenosols
11	Swamp	Daotang River Wetland			Chernozems
		Nymph Bay Wetland			Kastanozems
		Sand Island Wetland			Solonchaks
		Spring Bay Wetland	20	Intermediate hose	Intermediate hose
		Satchel Lake Wetland	21	Capillary zone	Capillary zone
		Hada Beach Wetland	22	Phreatic water	Overflow Spring
13	Glacier	Shule South Mountain Glacier	23	Confined water	Confined water
		Datong Mountain Glacier

. As shown in Table 4, marine entities (id = 12) are not involved here, as there are no oceans within the Qinghai Lake Basin.

Secondly, based on the conceptual entity relationships in Table 2 and the instance entities in Table 4, we constructed the Qinghai Lake Basin knowledge graph using Neo4j and Python. The visualization employs a color-coded schema where each conceptual entity category shares a unified hue. As shown in Figure 7: all lake instances (Qinghai Lake, Erhai Lake, and Gahai Lake) are rendered in #0000FF (blue), while wetland instances use #FF00FF (magenta). This approach enables intuitive recognition of the 22 entity types within the knowledge graph.

5. Conclusions

The water process plays an important role in obtaining influence factors and finding the source of water. Previous studies have used related factors to calculate and analyze water process but overlook the sources of factors affecting water process. Thus, we explored the construction and application of the top-down water process knowledge graph to clarify the changing process of water movement and the sources of influencing factors. The build process consists of four steps: (1) defining and classifying water bodies of water process; (2) identifying the knowledge of water process; (3) constructing the knowledge graph of water process and water process factors; (4) and analyzing the results.

The findings are summarized as follows: by comparing the entities and relationships involved in the water process, it is known that the water process is more complex. The path of water process and its influencing factors are extracted and analyzed. When the length of the connection between the two nodes is longer, the more paths of the water process, the more factors that affect the water process. Moreover, there are intersections between different paths, mainly because there are similarities in some parts of the water process. This may also be the basis for scholars to build model input from different angles. Furthermore, combined with the knowledge graph of influence factors of water process, we can not only clearly know the source of water process, but also know how many input factors are needed to calculate a certain amount. This will provide a data basis for scholars to adopt deep learning algorithms to study water processes.

We acknowledge that certain water processes were not explicitly considered in this study, such as those processes at different latitudes, basins, and a given basin. Moreover, quantification of those processes as well as scenarios (for example, how will the water process take place if the air temperature rises?) also were not taken into consideration. To address these limitations, future work will (1) establish composite entity concepts (e.g., watershed entities encapsulating vegetation, river, soil) and construct their knowledge graphs; (2) implement spatiotemporal graph neural networks (STGNNs) to quantify and analyze process dynamics based on spatial topologies and attribute evolution.