Ecological intelligent partitioning method integrating ecological niche theory and machine learning

Abstract

The invention belongs to the technical field of resources and environment, and discloses an ecological niche theory and machine learning integrated ecological intelligent zoning method, which comprises the steps of environmental factor and phytoplankton community data measurement, sampling point ecological niche width and overlap calculation, key environmental factor identification and ecological environment zoning. According to the method, data such as hydrochemical element phytoplankton communities, ecological niche overlapping and width can be effectively integrated, and a machine learning algorithm is applied to perform analysis, so that factors which have remarkable influence on an ecological system are identified, and scientific and reasonable ecological zoning is performed on a research area according to the key influence factors. The method not only can accurately reflect regional ecological characteristics, but also can provide important data support for ecological protection, species distribution prediction and water resource management. Meanwhile, the method can provide an effective basis for ecological environment monitoring, ecological restoration and regional environment planning in practical application, and has high application value and wide popularization prospect.

Description

Technical Field

[0001] The present invention belongs to the technical field of resources and environment, and particularly relates to an ecological intelligent zoning method integrating niche theory and machine learning. Background Art

[0002] At present, traditional habitat assessment techniques are mostly used in ecological zoning methods, but these methods are often difficult to adapt to complex and changing environmental conditions. Traditional methods often rely on experience and manual selection of environmental factors, and cannot comprehensively and objectively reflect the dynamic changes of the ecological environment. Currently, ecological zoning methods mainly rely on traditional environmental factor measurement and habitat assessment, but existing methods have many limitations. Especially when facing complex and changing ecological environments, it is often difficult to comprehensively reflect the relationship between the environment and biological communities. Traditional ecological zoning methods usually rely on empirical selection of environmental factors, and often ignore the interaction between various environmental factors in the ecosystem, lacking a systematic and quantitative ecological environment analysis method. In addition, the characterization of regional ecological characteristics by traditional methods is often too simplified, unable to effectively reveal the key factors affecting the ecosystem, resulting in the lack of scientificity and accuracy of ecological zoning results.

[0003] In recent years, as an important tool for studying the interaction between biological species and environmental factors, niche theory has been applied to ecological research, but its application still faces many challenges. Traditional niche analysis methods rely on artificially selected environmental factors and simple mathematical models, and cannot comprehensively analyze large-scale and multi-dimensional ecological environments. At the same time, existing research often ignores the impact of niche overlap and width on ecosystem functions, and fails to deeply explore the application of these factors in ecological environment zoning.

[0004] In summary, there are obvious deficiencies in existing ecological zoning methods, and it is urgent to develop new technologies and methods to solve these problems in order to achieve more scientific and accurate ecological zoning. Summary of the Invention

[0005] In view of the above existing technical bottlenecks, the present invention proposes an ecological intelligent zoning method integrating niche theory and machine learning, breaking through the data dependence limitation of traditional methods. By effectively integrating environmental factor data and phytoplankton community data and using machine learning algorithms for analysis, it realizes the automatic identification of key factors of the ecosystem and spatial heterogeneity analysis under complex environments, provides a scientific basis for ecological protection and resource management in data-scarce areas, thereby scientifically identifying the environmental factors that have a significant impact on the ecosystem, and reasonably zoning the ecological environment based on these factors.

[0006] The object of the present invention is achieved as follows:

[0007] The present invention provides an ecological intelligent zoning method integrating the niche theory and machine learning, comprising the following steps:

[0008] Step 1, measurement of environmental factors and phytoplankton community data:

[0009] S1.1: Determine the study area and sampling points, collect water chemical element samples at each sampling point, and send them to the laboratory for measurement;

[0010] S1.2: Collect phytoplankton samples at each sampling point in the study area and send them to the laboratory for measurement;

[0011] S1.3: Based on the phytoplankton sample data measured in the laboratory, calculate the biomass, species richness or species diversity index of phytoplankton at each sampling point;

[0012] S1.4: Calculate the dominance of each species, and select the species with a dominance greater than 0.02 as representative species.

[0013] Step 2, calculation of niche breadth and overlap of sampling points:

[0014] S2.1: For all sampling points in the study area, divide them into different concentration ranges based on all water chemical element indexes;

[0015] S2.2: Calculate the niche breadth of representative species under different water chemical element concentration gradients through the Levins formula;

[0016] S2.3: Calculate the niche overlap of representative species under different water chemical element concentration gradients through the Pianka overlap model;

[0017] S2.4: Calculate the niche breadth of each index at each sampling point through the biomass proportion and niche breadth of the representative species at each sampling point;

[0018] S2.5: Calculate the niche overlap of each index at each sampling point through the biomass proportion and niche overlap of the representative species at each sampling point.

[0019] Step 3, identification of key environmental factors:

[0020] S3.1: Calculate the importance of each index through the random forest algorithm for the water chemical element indexes of each sampling point measured in Step 1 and the calculated species richness or species diversity index, and the niche breadth and niche overlap of each sampling point calculated in Step 2;

[0021] S3.2: Based on the importance of each index, screen out the key environmental factors.

[0022] Step 4, ecological environment zoning:

[0023] S4.1: Input the key environmental factors identified in step 3 and the species richness or species diversity index as feature vectors, calculate the silhouette coefficients for different numbers of clusters, and select the number of clusters with the highest silhouette coefficient as the optimal number of clusters for the best clustering effect.

[0024] S4.2: Perform clustering through the unsupervised learning K-means algorithm or hierarchical clustering method to complete the ecological environment zoning of the study area.

[0025] Furthermore, in S1.1 of step 1, the number of sampling points set is greater than 15. The water chemical element indicators measured in each water chemical element sample include: total nitrogen, total phosphorus, total carbon, total silicon, water temperature, pH, dissolved oxygen, dissolved organic carbon, dissolved total nitrogen, dissolved total phosphorus, and dissolved silicon. The indicators measured in the phytoplankton sample include: the number of phytoplankton species, cell density, and biomass.

[0026] Furthermore, in S2.1 of step 2, the concentration intervals of the water chemical element indicators are divided based on the percentile method.

[0027] Furthermore, in S3.1 of step 3, the specific steps of calculating by the random forest algorithm include:

[0028] Input the water chemical element indicators, niche breadth, and niche overlap of each sampling point as feature vectors, and at the same time input the species richness or species diversity index as the target variable. Use the random forest algorithm to calculate the importance of each indicator through the increase in mean squared error.

[0029] Furthermore, in 3.2 of step 3, if the input target variable is even, then take the environmental indicators ranked in the top n / 2 in terms of importance as the key environmental factors; if the input target variable is odd, then take the environmental indicators ranked in the top (n + 1) / 2 in terms of importance as the key environmental factors.

[0030] The advantages and beneficial effects of the present invention are:

[0031] 1. Compared with traditional methods, the present invention combines the niche theory with machine learning algorithms to scientifically zone the ecological environment of the study area. Based on the analysis of phytoplankton niche characteristics and environmental factors, it can identify the key ecological factors in the area and provide reasonable criteria for regional ecological zoning based on these factors. On the one hand, this criterion can improve the accuracy of traditional ecological zoning methods and enable them to better reflect the characteristics of the ecosystem in a changing ecological environment;

[0032] 2. The present invention can achieve a comprehensive and systematic analysis of complex ecosystems, help quickly identify key areas for ecological protection, and provide scientific guidance for ecological environment monitoring, restoration, and regional planning. The application of this method can more objectively and accurately reflect the ecological characteristics of the region, provide important data support for ecological protection, species distribution prediction, and water resource management, not only contribute to improving the efficiency of ecological protection work, but also provide important data support for water resource management and species distribution prediction, and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The present invention will be further described below in conjunction with the drawings and embodiments.

[0034] Figure 1 is a flowchart of the ecological intelligent zoning method integrating the niche theory and machine learning according to an embodiment of the present invention;

[0035] Figure 2 shows the example area and sampling points selected in the embodiment of the present invention;

[0036] Figure 3 is the weighted value of the niche breadth and niche overlap of the sampling points calculated in the embodiment of the present invention;

[0037] Figure 4 is the importance degree of the key factors calculated in the embodiment of the present invention, and the key environmental factors (water temperature, niche breadth, niche overlap, pH, DTN, TN, TP) are selected;

[0038] Figure 5 is the silhouette coefficient of different clustering trees calculated in the embodiment of the present invention, and the clustering effect is the best when the number of clusters is 3;

[0039] Figure 6 is the ecological environment zoning result of the example area described in the present invention and the geographical distribution of different zones. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0040] Embodiment 1:

[0041] As Figure 1 shown, this embodiment provides an ecological intelligent zoning method integrating the niche theory and machine learning, including the following steps:

[0042] Step 1, measurement of environmental factors and phytoplankton community data:

[0043] S1.1: Determine the study area and sampling points, collect water chemical element samples at each sampling point, and send them to the laboratory for measurement. The number of sampling points set is greater than 15. The water chemical element indicators measured in each water chemical element sample include: total nitrogen, total phosphorus, total carbon, total silicon, water temperature, pH, dissolved oxygen, dissolved organic carbon, dissolved total nitrogen, dissolved total phosphorus, and dissolved silicon; the indicators measured in the phytoplankton sample include: the number of phytoplankton species, cell density, and biomass.

[0044] S1.2: Collect phytoplankton samples at each sampling point in the study area and send them to the laboratory for measurement;

[0045] S1.3: Based on the data of the phytoplankton samples measured in the laboratory, calculate the biomass, species richness, or species diversity index of phytoplankton at each sampling point;

[0046] S1.4: Calculate the dominance of each species, and select the species with a dominance greater than 0.02 as the representative species.

[0047] Step 2, calculation of the niche breadth and overlap of sampling points:

[0048] S2.1: For all sampling points in the study area, divide them into different concentration intervals based on all water chemical element indicators; among them, the concentration intervals of water chemical element indicators are divided based on the percentile method;

[0049] S2.2: Calculate the niche breadth of the representative species at different water chemical element concentration gradients through the Levins formula;

[0050] S2.3: Calculate the niche overlap of the representative species at different water chemical element concentration gradients through the Pianka overlap model;

[0051] S2.4: Calculate the niche breadth of each indicator at each sampling point through the biomass proportion and niche breadth of the representative species at each sampling point;

[0052] S2.5: Calculate the niche overlap of each indicator at each sampling point through the biomass proportion and niche overlap of the representative species at each sampling point.

[0053] Step 3, identify key environmental factors:

[0054] S3.1: Use the random forest algorithm to calculate the importance of each indicator for the water chemical element indicators measured at each sampling point in Step 1 and the calculated species richness or species diversity index, and the niche breadth and niche overlap calculated at each sampling point in Step 2.

[0055] The specific steps of the calculation by the random forest algorithm include:

[0056] Input the water chemical element indexes, niche breadth, and niche overlap of each sampling point as feature vectors, and at the same time input the species richness or species diversity index as the target variable. Use the random forest algorithm to calculate the importance of each index through the increase in mean squared error.

[0057] S3.2: Based on the importance of each index, screen out the key environmental factors; specifically, if the input target variable is even, then take the environmental indicators ranked in the top n / 2 (including n / 2) in terms of importance as the key environmental factors; if the input target variable is odd, then take the environmental indicators ranked in the top (n + 1) / 2 (including (n + 1) / 2) in terms of importance as the key environmental factors.

[0058] Step 4, ecological environment zoning:

[0059] S4.1: Input the key environmental factors identified in step 3 and the species richness or species diversity index as feature vectors, calculate the silhouette coefficients for different numbers of clusters, and select the number of clusters with the highest silhouette coefficient as the number of clusters with the best clustering effect.

[0060] S4.2: Perform clustering through the unsupervised learning K-means algorithm or hierarchical clustering method to complete the ecological environment zoning of the study area.

[0061] Application example:

[0062] The upper reaches of the Yangtze River Basin is located at 90°13′ - 111°30′E, 24°37′ - 35°54′N, from Geladandong on the Qinghai-Tibet Plateau to Yichang, Hubei Province. The area is 1.006×10 6 km 2 , accounting for about 55.9% of the total area of the Yangtze River Basin. The total length is about 4500 km, accounting for about 71.4% of the total length of the whole basin. The total drop exceeds 5100 meters, accounting for 95% of the total drop of the Yangtze River. The upper reaches of the Yangtze River Basin has significant ecological diversity and uniqueness, with complex hydrographic and geographical characteristics and relatively large ecological environment protection needs. Taking the upper reaches of the Yangtze River as an example to elaborate the present invention has good representativeness.

[0063] The first step: Measure environmental factors and phytoplankton community data in the upper reaches of the Yangtze River:

[0064] (1) The basin survey was mainly carried out on the main stream in the upper reaches. Sampling points were set under the key power station dams. Considering that the tributaries are also important sources of nutrients, some larger tributaries in the upper reaches of the Yangtze River Basin were also surveyed. Sampling points were also set at the tributaries of Dingqu, Yalong River, Hengjiang River, Minjiang River, and Chishui River and their confluence points with the main stream ( Figure 2 ).

[0065] (2) At each selected sampling point, water samples were collected on-site and sent to the laboratory for measurement and analysis. The measurement indicators included dissolved oxygen (DO), pH, water temperature (WT), total carbon (TC), dissolved organic carbon (DOC), total nitrogen (TN), dissolved total nitrogen (DTN), total phosphorus (TP), dissolved total phosphorus (DTP), total silicon (TSi), dissolved silicon (DSi), and the number of phytoplankton species, cell density, biomass, etc.

[0066] (3) Through the Mcnaughton dominance index (Y), the dominance of each species was calculated, as shown in Table 1. Species with a dominance greater than 0.02 were selected as representative species. A total of 10 dominant species with a dominance greater than 0.02 were selected, as shown in Table 2.

[0067] Table 1 Dominance of phytoplankton in the sample area for accounting

[0068]

[0069]

[0070] Table 2 Representative species selected

[0071]

[0072] Step 2: Calculation of niche breadth and overlap of sampling points

[0073] (1) For all sampling points in the study area, all water chemical element indicators were divided into five concentration intervals based on percentiles;

[0074] (2) Through the Levins formula, the niche breadths of 10 representative species based on the concentration gradients of 11 water chemical indicators were calculated for different concentration gradients of water chemical elements, as shown in Table 3.

[0075] Table 3 Results of niche breadths of representative species for 11 water chemical indicators

[0076]

[0077] (3) Through the Pianka overlap model, the niche overlaps of 10 representative species with other species were calculated based on the concentration gradients of different indicators for different concentration gradients of water chemical elements, as shown in Table 4.

[0078] Table 4 Results of niche overlaps of representative species for 11 water chemical indicators

[0079]

[0080] (4) Based on the biomass proportion of each phytoplankton at each sampling point, multiply the proportion of each representative phytoplankton species in all representative species by the proportion of that species at each sampling point and then sum them up. Finally, calculate the weighted niche width and niche overlap of the 26 sampling points ( Figure 3 ).

[0081] Step 3: Identify the key environmental factors in the upper reaches of the Yangtze River

[0082] (1) Take the 11 water chemical indicators (dissolved oxygen (DO), pH, water temperature (WT), total carbon (TC), dissolved organic carbon (DOC), total nitrogen (TN), dissolved total nitrogen (DTN), total phosphorus (TP), dissolved total phosphorus (DTP), total silicon (TSi), and dissolved silicon (DSi)) of the 26 sampling points, as well as the niche width and niche overlap as feature vectors, and take the species richness as the target vector and input them into the random forest model.

[0083] (2) The number of input target variables is 13, which is an odd number. Take the top seven environmental indicators ranked by importance as the key environmental factors, namely water temperature (47.09), niche width (39.33), niche overlap (21.47), pH (17.90), dissolved total nitrogen (9.33), total nitrogen (8.05), and total phosphorus (6.41)( Figure 4 ).

[0084] Step 4: Ecological environment zoning in the upper reaches of the Yangtze River

[0085] (1) Through the hierarchical clustering of unsupervised machine learning, calculate the clustering results for different numbers of clusters. When the number of clusters is 3, the clustering effect is the best ( Figure 5 ).

[0086] (2) Based on the setting of the number of clusters being 3, use the hierarchical clustering of unsupervised machine learning to cluster the 26 sampling points in the upper reaches of the Yangtze River. The results show that the 26 sampling points are divided into three clusters, containing 5, 6, and 15 sampling points respectively. The points in Cluster 1 are all distributed in the upper reaches of the Jinsha River. Most of the points in Cluster 2 are located in the middle and lower reaches of the Jinsha River, while the points in Cluster 3 are mainly concentrated in other main and tributary areas of the upper reaches of the Yangtze River connected to the lower reaches of the Jinsha River( Figure 6)。Clusters 1, 2, and 3 are geographically located in the upper, middle, and lower reaches of the example area, respectively. The characteristics of the upper reaches include high pH, low water temperature (WT), medium to high concentrations of total nitrogen (TN), dissolved total nitrogen (DTN), and total phosphorus (TP), and low levels of niche breadth, niche overlap, and species richness. The characteristics of the middle reaches include medium pH and water temperature (WT), low concentrations of total nitrogen (TN), dissolved total nitrogen (DTN), and total phosphorus (TP), medium to high levels of niche breadth and overlap, and relatively low species richness. The lower reaches exhibit low pH, high water temperature (WT), high concentrations of total nitrogen (TN) and dissolved total nitrogen (DTN), medium to high concentrations of total phosphorus (TP), relatively high levels of niche breadth and niche overlap, and medium levels of species richness. These conclusions provide a theoretical basis for the protection of phytoplankton communities in the upper reaches of the Yangtze River and the management of aquatic ecosystems, especially having important reference value when formulating targeted protection and restoration measures.

[0087] Finally, it should be noted that the above is only used to illustrate the technical solution of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred arrangement, those of ordinary skill in the art should understand that the technical solution of the present invention (such as the application of certain formulas, the sequence of steps, etc.) can be modified or equivalently replaced without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. An ecological intelligent zoning method integrating ecological niche theory and machine learning, characterized in that: The method comprises the following steps: Step 1: Measurement of environmental factors and phytoplankton community data: S1.1: Determine the study area and sampling points, collect water chemical element samples at each sampling point, and send them to the laboratory for measurement; S1.2: Collect phytoplankton samples from each sampling point in the study area and send them to the laboratory for measurement; S1.3: Based on the phytoplankton sample data measured in the laboratory, calculate the phytoplankton biomass, species richness or species diversity index at each sampling point; S1.4: Calculate the dominance of each species and select species with a dominance greater than 0.02 as representative species; Step 2, calculation of sampling point niche width and overlap: S2.1: For all sampling points in the study area, different concentration ranges are divided based on all water chemical element indicators; S2.2: Calculate the niche width of representative species under different water chemical element concentration gradients using the Levins formula; S2.3: Calculate the niche overlap of representative species under different water chemical element concentration gradients using the Pianka overlap model; S2.4: Calculate the niche width of each indicator at each sampling point through the biomass proportion and niche width of the representative species at each sampling point; S2.5: Calculate the niche overlap of each indicator at each sampling point through the biomass proportion and niche overlap of the representative species at each sampling point; Step 3, identify key environmental factors: S3.1: The water chemical element indicators of each sampling point measured in step 1 and the calculated species richness or diversity index, and the niche width and niche overlap of each sampling point calculated in step 2 are used to calculate the importance of each indicator through the random forest algorithm; S3.2: Screen out key environmental factors based on the importance of each indicator; Step 4, ecological environment zoning: S4.1: Use the key environmental factors and species richness or species diversity index identified in step 3 as feature vector inputs, calculate the silhouette coefficients of different cluster numbers, and select the cluster number with the highest silhouette coefficient as the cluster number with the best clustering effect; S4.2: Clustering is performed through the unsupervised learning K-means algorithm or hierarchical clustering method to complete the ecological environment zoning of the study area.

2. The method according to claim 1, characterized in that In step 1, S1.1, the number of sampling points set is greater than 15, and the water chemical element indicators measured in each water chemical element sample include: total nitrogen, total phosphorus, total carbon, total silicon, water temperature, pH, dissolved oxygen, dissolved organic carbon, dissolved total nitrogen, dissolved total phosphorus and dissolved silicon; the indicators measured for phytoplankton samples include: the number of phytoplankton species, cell density and biomass.

3. The method according to claim 1, characterized in that In step 2, S2.1, the concentration ranges of water chemical element indicators are divided based on percentiles.

4. The method according to claim 1, characterized in that: In step 3, S3.1, the specific steps of the random forest algorithm calculation include: The water chemical element indicators, niche width and niche overlap of each sampling point were taken as feature vector input data, and the species richness or species diversity index was input as the target variable. The random forest algorithm was used to calculate the importance of each indicator through the mean square error increase.

5. The method according to claim 1, characterized in that In step 3, 3.2, if the input target variable is an even number, the environmental indicators ranked in the top n / 2 in terms of importance are taken as key environmental factors; if the input target variable is an odd number, the environmental indicators ranked in the top (n+1) / 2 in terms of importance are taken as key environmental factors.

Images