Greenhouse gas concentration monitoring method using remote sensing technology
By combining remote sensing and ground data into a machine learning model, the problems of insufficient spatiotemporal coverage and insufficient dynamic response of traditional monitoring methods have been solved, and high-precision, real-time greenhouse gas concentration monitoring has been achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-18
- Publication Date
- 2026-03-20
AI Technical Summary
Traditional ground-based measurement stations cannot achieve comprehensive, real-time, and high-frequency monitoring of greenhouse gas concentrations. Remote sensing technology is affected by the atmosphere, has difficulties in data fusion, and lacks dynamic response capabilities.
A machine learning model based on physical constraints is adopted, which combines remote sensing data and ground measurement data. Through a multimodal input architecture, feature alignment module, graph neural network, spatiotemporal dual-dimensional attention module and physical constraint layer of radiative transfer equation, the data is accurately fused and corrected, and the model is trained using a federated learning architecture.
It improves the accuracy and spatiotemporal consistency of greenhouse gas concentration retrieval, enhances dynamic response capabilities, reduces computational costs, and improves the practicality and adaptability of the monitoring system.
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of environmental monitoring, more particularly, to a greenhouse gas concentration monitoring method using remote sensing technology. BACKGROUND
[0002] With the increasingly serious global climate change problem, the emission monitoring of greenhouse gases has become a key research direction in the field of climate science and environmental protection. The concentration change of greenhouse gases directly affects the process of global climate change, therefore, accurately monitoring the concentration of greenhouse gases, identifying their sources and distribution characteristics, is of great significance for climate policy making, environmental protection and global warming mitigation strategies.
[0003] Traditional greenhouse gas monitoring methods usually rely on ground measurement stations, which are widely distributed in different regions and can provide local concentration data. However, ground measurement stations have certain limitations, mainly in the following aspects:
[0004] Ground measurement stations are usually sparsely distributed, especially in some remote areas, making it difficult to achieve comprehensive coverage. This leads to blind spots in the monitoring of greenhouse gas concentrations, and cannot fully and real-time reflect the concentration changes in a large area. The monitoring frequency of ground measurement stations is usually limited by equipment and manual intervention, and cannot achieve high-frequency and continuous monitoring. This makes it difficult to grasp the dynamic changes of greenhouse gas concentrations in real time, affecting the timeliness of pollution source tracking and environmental early warning. Due to the complex spatial distribution of greenhouse gas concentrations, a single ground station cannot provide multi-dimensional monitoring data, which limits the modeling accuracy and spatial resolution of concentration change trends.
[0005] In order to solve the above problems, remote sensing technology has emerged and gradually become an effective tool for monitoring the concentration of greenhouse gases. Remote sensing technology, through satellite, unmanned aerial vehicle and other platforms, combined with spectral detection technology, can realize real-time, large-scale and high-frequency monitoring of greenhouse gas concentrations on a global scale. Remote sensing data provides large-scale, high temporal and spatial resolution information, which can complement the shortcomings of traditional ground measurement stations. However, the application of remote sensing technology also faces some challenges, mainly in the following aspects:
[0006] The greenhouse gas concentration information obtained by remote sensing data is greatly affected by the atmosphere. The gases, meteorological conditions, and cloud layers in the atmosphere can interfere with the remote sensing signal, causing the inversion accuracy to decrease. Therefore, how to model and correct these influencing factors through reasonable physical mechanisms has become the key to improving the inversion accuracy of remote sensing data. The data provided by remote sensing technology is usually multi-source and multi-modal, such as satellite sensor data and ground meteorological observation data. These data have different collection frequencies, spatial resolutions, and physical properties. How to effectively fuse, align, and ensure the spatio-temporal consistency of data from different data sources is an important problem to improve the inversion accuracy. The traditional remote sensing data analysis method mainly focuses on static analysis and lacks dynamic response to the change of greenhouse gas concentration. With the change of greenhouse gas concentration, it is particularly important to adjust the monitoring strategy and response mechanism in real time. How to realize real-time monitoring and early warning has become the key to improving the practicality of the monitoring system.
[0007] To solve the above problems, the present application provides a machine learning model based on physical mechanism constraint, which combines remote sensing data with ground measurement data, improves the accuracy of greenhouse gas concentration inversion through multi-modal input architecture and feature alignment module. In addition, the present application also introduces advanced technologies such as graph neural network, spatio-temporal two-dimensional attention module, and radiation transfer equation physical constraint layer, and improves the spatial resolution, dynamic response capability and cross-source feature interaction capability of the model by deeply coupling machine learning model and chemical transport model. SUMMARY
[0008] 1. Technical problems to be solved
[0009] In view of the problems in the prior art, the purpose of the present application is to provide a greenhouse gas concentration monitoring method using remote sensing technology, which combines remote sensing data with ground measurement data, and improves the accuracy of greenhouse gas concentration inversion through multi-modal input architecture and feature alignment module.
[0010] 2. Technical solutions
[0011] To solve the above problems, the present application adopts the following technical solutions.
[0012] The greenhouse gas concentration monitoring method using remote sensing technology comprises collecting remote sensing data and ground measurement data, and modeling the nonlinear coupling relationship between atmospheric parameters and remote sensing data through a physical mechanism constraint machine learning model and implementing systematic atmospheric correction.
[0013] Further, the machine learning model adopts a multi-modal input architecture to simultaneously receive remote sensing data and ground meteorological variable sequences, and performs spatio-temporal registration through a feature alignment module.
[0014] Further, the machine learning model adopts a graph neural network, wherein nodes represent ground measurement stations within a geographical area, and directed edges are constructed based on atmospheric transmission paths, wind field vectors, and emission source distribution.
[0015] Further, the machine learning model integrates a space-time dual-dimensional attention module.
[0016] Further, the machine learning model embeds a physical constraint layer based on a radiation transfer equation in the short-wave infrared band.
[0017] Further, the machine learning model integrates a differentiable radiation transfer simulator as a physical constraint layer and fuses heterogeneous observation data from multiple remote sensing platforms.
[0018] Further, a federated learning architecture is adopted to weight and fuse model gradients of distributed nodes, and a global concentration estimation model is generated based on a Bayesian update strategy.
[0019] Further, the convolution kernel, receptive field, and channel weight of the local model are driven by MSI sensor data, and bidirectional coupling of multi-source remote sensing input and greenhouse gas concentration field is realized under the federated framework.
[0020] Further, the physically mechanism-constrained machine learning model is deeply coupled with a chemical transport model, and cross-source feature interaction between remote sensing data and atmospheric parameters is realized through cross-attention mechanism.
[0021] Further, a multi-dimensional criterion is constructed based on the concentration change slope within a sliding time window, the spatial gradient mutation index, and the deviation degree from the historical same period mean, and a hierarchical alarm mechanism is triggered according to the multi-dimensional criterion.
[0022] 3, Advantage
[0023] Compared with the prior art, the advantages of the present application are:
[0024] 1) Through the introduction of multi-modal input architecture and feature alignment module, remote sensing data and ground meteorological data can be accurately registered in space-time dimension, which improves the modeling accuracy of the nonlinear coupling relationship between atmospheric parameters and remote sensing data, thereby significantly improving the accuracy and spatiotemporal consistency of greenhouse gas concentration inversion.
[0025] 2) Graph neural network and space-time dual-dimensional attention module are introduced, nodes represent ground measurement stations, directed edges are constructed based on atmospheric transmission paths, wind field vectors, and emission source distribution, so that the model can effectively fuse spatial correlation and physical transmission mechanism, and improve the spatial resolution and dynamic response accuracy of greenhouse gas concentration inversion results.
[0026] 3) Embedding a physical constraint layer based on the radiative transfer equation in the machine learning model ensures that the data-driven method is consistent with the physical law during the greenhouse gas concentration inversion process, effectively reducing the prediction bias that may be generated by the data-driven method, and improving the generalization ability and accuracy of the model.
[0027] 4) By introducing a differentiable radiative transfer simulator and heterogeneous observation data from multiple remote sensing platforms, the model's adherence to atmospheric physical processes is enhanced, enabling it to better handle data from different sources and improving the accuracy and reliability of greenhouse gas concentration inversion.
[0028] 5) The present application couples the machine learning model with the chemical transport model, uses the cross-attention mechanism to realize cross-source feature interaction between remote sensing data and atmospheric parameters, so that the inversion of greenhouse gas concentration not only has spatial consistency, but also has stronger temporal continuity, and can dynamically respond to gas concentration changes at different time scales.
[0029] 6) By using a federated learning architecture, the model can be trained on multiple distributed nodes, avoiding single-point centralized processing of large data sets, reducing computational cost, and enabling cross-geographical area collaboration in meteorological station data from different regions, improving the adaptability and real-time monitoring capability of the system. DETAILED DESCRIPTION
[0030] The technical solutions in the embodiments of the present application will be described below; obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.
[0031] Embodiment 1:
[0032] The greenhouse gas concentration monitoring method using remote sensing technology includes the following determination steps:
[0033] Step S100, using the TROPOMI spectrometer carried by the European Space Agency Sentinel-5P satellite to collect atmospheric remote sensing data;
[0034] Step S200, using the fixed meteorological station of the World Meteorological Organization Global Atmospheric Monitoring Network to collect ground measurement data;
[0035] Step S300, modeling the nonlinear coupling relationship between atmospheric parameters and remote sensing data by integrating a long short-term memory network of atmospheric radiative transfer equation;
[0036] Step S400, implementing Rayleigh scattering correction and ozone absorption correction for systematic atmospheric correction.
[0037] Embodiment 2
[0038] The greenhouse gas concentration monitoring method using remote sensing technology comprises the following determination steps:
[0039] Step S100, collecting atmospheric remote sensing data using the TROPOMI spectrometer carried by the Sentinel-5P satellite of the European Space Agency;
[0040] Step S200, collecting ground measurement data using the fixed weather stations of the World Meteorological Organization Global Atmospheric Monitoring Network;
[0041] Step S300, explicitly modeling the nonlinear coupling relationship between atmospheric parameters and remote sensing data by integrating a long short-term memory network of atmospheric radiation transfer equations, wherein the long short-term memory network adopts a multi-modal input architecture, simultaneously receives remote sensing data from the TROPOMI spectrometer and ground meteorological variable sequences from the fixed weather stations, and performs temporal and spatial registration on the two types of input data through a feature alignment module;
[0042] Step S400, implementing Rayleigh scattering correction and ozone absorption correction for systematic atmospheric correction.
[0043] Compared with the original embodiment 1, the present embodiment realizes accurate matching of remote sensing data and ground meteorological variables in time and space dimensions by introducing a multi-modal input architecture and a feature alignment module, effectively improves the modeling accuracy of the long short-term memory network on the nonlinear coupling relationship between atmospheric parameters and observation data, and thus improves the accuracy and spatiotemporal consistency of the greenhouse gas concentration inversion results.
[0044] Embodiment 3
[0045] The greenhouse gas concentration monitoring method using remote sensing technology comprises the following determination steps:
[0046] Step S100, collecting atmospheric remote sensing data using the TROPOMI spectrometer carried by the Sentinel-5P satellite of the European Space Agency;
[0047] Step S200, collecting ground measurement data using the fixed weather stations of the World Meteorological Organization Global Atmospheric Monitoring Network;
[0048] Step S300, explicitly modeling the nonlinear coupling relationship between atmospheric parameters and remote sensing data by integrating a long short-term memory network of atmospheric radiation transfer equations, wherein the machine learning model further adopts a graph neural network, the nodes in the graph neural network represent ground measurement stations within a geographical region, and the directed edges are constructed based on atmospheric transmission paths, wind field vectors, and emission source distribution;
[0049] Step S400, implementing Rayleigh scattering correction and ozone absorption correction for systematic atmospheric correction.
[0050] Compared with the original embodiment 1, the new embodiment introduces a graph neural network structure, uses node representations of ground stations, and constructs directed edges according to atmospheric transmission paths, wind field vectors, and emission source distributions, which can effectively integrate spatial correlation and physical transmission mechanisms, and improve the spatial consistency and dynamic response accuracy of greenhouse gas concentration inversion.
[0051] Embodiment 4:
[0052] The greenhouse gas concentration monitoring method using remote sensing technology comprises the following determination steps:
[0053] Step S100, using the TROPOMI spectrometer carried by the Sentinel-5P satellite of the European Space Agency to collect atmospheric remote sensing data;
[0054] Step S200, using the fixed weather station of the World Meteorological Organization Global Atmospheric Monitoring Network to collect ground measurement data;
[0055] Step S300, modeling the nonlinear coupling relationship between atmospheric parameters and remote sensing data by integrating a long short-term memory network of atmospheric radiation transfer equation, and the long short-term memory network outputs a preliminary concentration estimate;
[0056] Step S400, implementing Rayleigh scattering correction and ozone absorption correction for systematic atmospheric correction.
[0057] Compared with the original embodiment 1, the embodiment explicitly outputs the preliminary concentration estimate by the long short-term memory network in step S300, so that the subsequent atmospheric correction step has a clear initial input benchmark, improving the convergence speed of the correction process and the accuracy of the final concentration inversion result.
[0058] Embodiment 5:
[0059] The greenhouse gas concentration monitoring method using remote sensing technology comprises the following determination steps:
[0060] Step S100, using the TROPOMI spectrometer carried by the Sentinel-5P satellite of the European Space Agency to collect atmospheric remote sensing data;
[0061] Step S200, using the fixed weather station of the World Meteorological Organization Global Atmospheric Monitoring Network to collect ground measurement data;
[0062] Step S300, modeling the nonlinear coupling relationship between atmospheric parameters and remote sensing data by integrating a long short-term memory network of atmospheric radiation transfer equation, and the long short-term memory network outputs a preliminary concentration estimate;
[0063] Step S400, Rayleigh scattering correction and ozone absorption correction are implemented to perform systematic atmospheric correction.
[0064] Compared with the original embodiment 1, the embodiment can respectively focus on key time steps and important spatial regions by introducing a space-time dual-dimensional attention module in the long short-term memory network, effectively improving the model's ability to capture the spatial and temporal distribution characteristics of greenhouse gas concentration, thereby improving the inversion accuracy and robustness.
[0065] Embodiment 6:
[0066] The greenhouse gas concentration monitoring method using remote sensing technology comprises the following determination steps:
[0067] Step S100, using the TROPOMI spectrometer carried by the Sentinel-5P satellite of the European Space Agency to collect atmospheric remote sensing data;
[0068] Step S200, using the fixed weather station of the World Meteorological Organization Global Atmospheric Monitoring Network to collect ground measurement data;
[0069] Step S300, the long short-term memory network integrating the atmospheric radiation transfer equation is used to explicitly model the nonlinear coupling relationship between the atmospheric parameters and the remote sensing data, wherein the long short-term memory network is embedded with a physical constraint layer based on the radiation transfer equation in the short-wave infrared band;
[0070] Step S400, Rayleigh scattering correction and ozone absorption correction are implemented to perform systematic atmospheric correction.
[0071] Compared with the original embodiment 1, the embodiment introduces a physical constraint layer based on the radiation transfer equation in the short-wave infrared band of the long short-term memory network, so that the model strictly follows the physical law of atmospheric radiation transfer during the learning process, effectively suppressing the prediction deviation caused by the lack of physical consistency in the short-wave infrared band of the pure data-driven method, and improving the accuracy and generalization ability of the greenhouse gas concentration inversion result.
[0072] Embodiment 7:
[0073] The greenhouse gas concentration monitoring method using remote sensing technology comprises the following determination steps:
[0074] Step S100, using the TROPOMI spectrometer carried by the Sentinel-5P satellite of the European Space Agency to collect atmospheric remote sensing data;
[0075] Step S200, using the fixed weather station of the World Meteorological Organization Global Atmospheric Monitoring Network to collect ground measurement data;
[0076] Step S300, explicitly modeling the nonlinear coupling relationship between atmospheric parameters and remote sensing data by integrating a long short-term memory network of atmospheric radiation transfer equations, wherein the long short-term memory network embeds a differentiable radiation transfer simulator as a physical constraint layer and fuses heterogeneous observation data from a Sentinel-5P satellite TROPOMI spectrometer and a World Meteorological Organization global atmospheric monitoring network fixed meteorological station;
[0077] Step S400, implementing Rayleigh scattering correction and ozone absorption correction for systematic atmospheric correction.
[0078] Compared with the original embodiment 1, the present embodiment enhances the compliance of the model to atmospheric physical processes and improves the precision and generalization performance of greenhouse gas concentration inversion by introducing a differentiable radiation transfer simulator as a physical constraint layer in the machine learning model and explicitly fusing heterogeneous observation data from multiple remote sensing platforms.
[0079] Embodiment 8:
[0080] The greenhouse gas concentration monitoring method using remote sensing technology comprises the following determination steps:
[0081] Step S100, using a TROPOMI spectrometer carried by a Sentinel-5P satellite of the European Space Agency to collect atmospheric remote sensing data;
[0082] Step S200, using a fixed meteorological station of a World Meteorological Organization global atmospheric monitoring network to collect ground measurement data;
[0083] Step S300, explicitly modeling the nonlinear coupling relationship between atmospheric parameters and remote sensing data by integrating a long short-term memory network of atmospheric radiation transfer equations;
[0084] Step S400, implementing Rayleigh scattering correction and ozone absorption correction for systematic atmospheric correction.
[0085] Step S500, adopting a federated learning architecture to weight and fuse model gradients of distributed nodes, and based on a Bayesian update strategy, aggregating to generate a global concentration estimation model.
[0086] Compared with the original embodiment 1, the newly added step S500 introduces a federated learning architecture and a Bayesian update strategy, so that each distributed node optimizes model parameters without sharing original observation data, effectively improving the generalization ability of the model in heterogeneous geographical regions, and enhancing the robustness and precision of global greenhouse gas concentration estimation.
[0087] Embodiment 9:
[0088] The greenhouse gas concentration monitoring method using remote sensing technology comprises the following determination steps:
[0089] Step S100, using the TROPOMI spectrometer carried by the Sentinel-5P satellite of the European Space Agency to collect atmospheric remote sensing data, and synchronously acquiring MSI sensor data carried by the Sentinel-2 satellite;
[0090] Step S200, using the fixed meteorological station of the World Meteorological Organization Global Atmospheric Monitoring Network to collect ground measurement data;
[0091] Step S300, driving the convolution kernel, receptive field and channel weight of the local model using the MSI sensor data, constructing a long short-term memory network based on the weight configuration, integrating the atmospheric radiation transfer equation to explicitly model the nonlinear coupling relationship between the atmospheric parameters and the remote sensing data, and realizing the multi-source input fusion of the TROPOMI remote sensing data, the MSI sensor data and the ground measurement data under the federated learning framework, and completing the bidirectional coupling between the greenhouse gas concentration field and the multi-source remote sensing input;
[0092] Step S400, implementing Rayleigh scattering correction and ozone absorption correction for systematic atmospheric correction.
[0093] Compared with Embodiment 1, the embodiment introduces MSI sensor data to dynamically drive the local model structure parameters, and realizes bidirectional coupling of multi-source remote sensing and concentration field under the federated framework, which significantly improves the spatial sensitivity of the model to surface heterogeneity and local emission sources, and enhances the spatial resolution and temporal consistency of greenhouse gas concentration inversion.
[0094] Embodiment 10:
[0095] The greenhouse gas concentration monitoring method using remote sensing technology comprises the following determination steps:
[0096] Step S100, using the TROPOMI spectrometer carried by the Sentinel-5P satellite of the European Space Agency to collect atmospheric remote sensing data;
[0097] Step S200, using the fixed meteorological station of the World Meteorological Organization Global Atmospheric Monitoring Network to collect ground measurement data;
[0098] Step S300, through the long short-term memory network integrating the atmospheric radiation transfer equation to explicitly model the nonlinear coupling relationship between the atmospheric parameters and the remote sensing data, wherein the long short-term memory network constitutes a physical mechanism constrained machine learning model, the model is deeply coupled with a chemical transport model, and cross-attention mechanism is used to realize cross-source feature interaction between remote sensing data and atmospheric parameters;
[0099] Step S400, implementing Rayleigh scattering correction and ozone absorption correction for systematic atmospheric correction.
[0100] Compared with the original embodiment 1, the embodiment significantly enhances the feature alignment capability between remote sensing data and ground atmospheric parameters by deeply coupling the physical mechanism constrained machine learning model with the chemical transmission model and introducing the cross attention mechanism, improves the spatial consistency and time continuity of greenhouse gas concentration inversion, and effectively suppresses the system deviation caused by the difference in observation scale.
[0101] Embodiment 11:
[0102] The greenhouse gas concentration monitoring method using remote sensing technology comprises the following determination steps:
[0103] Step S100, using the TROPOMI spectrometer carried by the European Space Agency Sentinel-5P satellite to collect atmospheric remote sensing data;
[0104] Step S200, using the fixed weather station of the World Meteorological Organization Global Atmospheric Monitoring Network to collect ground measurement data;
[0105] Step S300, modeling the nonlinear coupling relationship between atmospheric parameters and remote sensing data by integrating the long short-term memory network of atmospheric radiation transfer equation;
[0106] Step S400, implementing Rayleigh scattering correction and ozone absorption correction for systematic atmospheric correction;
[0107] Step S500, constructing a multi-dimensional criterion based on the concentration change slope, spatial gradient mutation index and deviation degree from the historical same period mean in the sliding time window, and activating a hierarchical alarm mechanism according to the multi-dimensional criterion.
[0108] Compared with the original embodiment 1, the newly added step S500 introduces a multi-dimensional criterion composed of concentration change slope, spatial gradient mutation index and historical same period mean deviation, and activates a hierarchical alarm mechanism accordingly, thereby improving the dynamic identification capability and early warning response accuracy of greenhouse gas abnormal emission events.
[0109] The above is only a preferred specific embodiment of the present application; however, the protection scope of the present application is not limited thereto. Any skilled person in the art can make equivalent replacements or changes to the technical range disclosed in the present application according to the technical solution and improvement concept of the present application, which should be covered within the protection scope of the present application.
Claims
1. A method for monitoring greenhouse gas concentrations using remote sensing technology, comprising collecting remote sensing data and ground measurement data, characterized in that: Machine learning models constrained by physical mechanisms explicitly model the nonlinear coupling relationship between atmospheric parameters and remote sensing data and implement systematic atmospheric correction.
2. The method for monitoring greenhouse gas concentrations using remote sensing technology according to claim 1, characterized in that: The machine learning model adopts a multimodal input architecture, simultaneously receiving remote sensing data and ground meteorological variable sequences, and performing spatiotemporal registration through a feature alignment module.
3. The method for monitoring greenhouse gas concentrations using remote sensing technology according to claim 1, characterized in that: The machine learning model employs a graph neural network, where nodes represent ground measurement stations within a geographical area, and directed edges are constructed based on atmospheric transport paths, wind field vectors, and emission source distributions.
4. The method for monitoring greenhouse gas concentrations using remote sensing technology according to claim 1, characterized in that: The machine learning model integrates a spatiotemporal dual-dimensional attention module.
5. The method for monitoring greenhouse gas concentrations using remote sensing technology according to claim 1, characterized in that: The machine learning model embeds a physical constraint layer based on the radiative transfer equation in the shortwave infrared band.
6. The method for monitoring greenhouse gas concentrations using remote sensing technology according to claim 1, characterized in that: The machine learning model integrates a differentiable radiative transfer simulator as a physical constraint layer and fuses heterogeneous observation data from a multi-source remote sensing platform.
7. The method for monitoring greenhouse gas concentrations using remote sensing technology according to claim 1, characterized in that: A federated learning architecture is used to weight and fuse the model gradients of distributed nodes, and a global concentration estimation model is generated by aggregating them based on a Bayesian update strategy.
8. The method for monitoring greenhouse gas concentrations using remote sensing technology according to claim 1, characterized in that: MSI sensor data is used to drive the convolution kernel, receptive field, and channel weights of the local model, and bidirectional coupling between multi-source remote sensing input and greenhouse gas concentration field is achieved under a federated framework.
9. The method for monitoring greenhouse gas concentrations using remote sensing technology according to claim 1, characterized in that: The machine learning model constrained by the physical mechanism is deeply coupled with the chemical transport model, and cross-source feature interaction between remote sensing data and atmospheric parameters is achieved through a cross-attention mechanism.
10. The method for monitoring greenhouse gas concentrations using remote sensing technology according to claim 1, characterized in that: It also includes constructing multidimensional criteria based on the slope of concentration change within a sliding time window, the spatial gradient mutation index, and the deviation from the historical average, and triggering a graded alarm mechanism based on the multidimensional criteria.