Inversion Of Water Quality Concentration And Denitrification Rate In Small Water Bodies Based On Smartphone And Artificial Neural Network

Water body is an indispensable resource for human production and life,with the increase of the intensity of human activities,a large number of nitrogen-containing pollutants enter the water body,resulting in increasingly serious nitrogen pollution problems in the water body.Nitrogen management in watersheds is the main method to control nitrogen pollution in water bodies,but it is limited by the accurate cognition of timely acquisition and absorption of water quality data.For water quality monitoring,although remote sensing and roadbed hyperspectral inversion rapid monitoring methods have been developed,they are limited by weather conditions and research scales.For absorption capacity monitoring,denitrification,and the main product of denitrification process is N₂,the N₂background content in the atmosphere is as high as 78%,the determination of denitrification rate at such a high background N₂concentration is very difficult,the general method is difficult to achieve,although a series of monitoring methods represented by membrane injection mass spectrometry have been developed,but most of them are indoor analysis methods,and the process is cumbersome,and it is urgent to develop a fast and accurate inversion method.Environmental monitoring through smartphone inversion has attracted more and more attention,and the research mainly focuses on the inversion of optical active parameters,but the inversion of non-optical active parameters is less,and this study intends to include the inversion of optical parameters and non-optical parameters of water.Since chlorophyll a and turbidity in water are optical parameters,different concentrations have significant differences in optical absorption wavelengths,which can be directly inverted by optical activity parameters.According to the previous work,denitrification and its main product N₂O are non-optical activity parameters,and because they are closely related to the optical activity parameters of water,indirect inversion monitoring can also be carried out.Based on smart phone photography,this paper obtains water image information through polarizer,mobile phone telescope,filter of different specifications,and 24-color standard color card,and combines stepwise regression and artificial neural network methods to carry out the inversion monitoring of water quality and denitrification rate of typical small and micro water bodies in southern Jiangsu.The main conclusions of this study are as follows:（1）The study found that the TN variation in the southern Jiangsu region was relatively large,with a TN concentration range of 0.24-10.59 mg/L and an average of 1.36 mg/L.The TN concentration in artificial lakes was the lowest,with an average of 0.55 mg/L;while the TN concentration in natural rivers was the highest,with an average of 2.41 mg/L.The TN concentration at the fixed sampling points in Changshu generally decreased from June to August and increased from September.The difference in chlorophyll a content in water bodies in southern Jiangsu was relatively large,with a range of 0.36-245.47 ug/L and an average of8.77 ug/L.The chlorophyll content in aquaculture ponds was the highest,with an average of89.26 ug/L;while the chlorophyll concentration in natural rivers was the lowest,with an average of 17.18 ug/L.The variation of denitrification rate in Changshu and Jurong cities was relatively strong,ranging from 16.42 to 418.91μmol N₂-N·m^-2·h^-1,and the average total denitrification rate was 98.48μmol N₂-N·m^-2·h^-1.The variation of natural rivers was the largest,ranging from 25.93 to 418.1μmol N₂-N·m^-2·h^-1,while that of natural ponds was the smallest,ranging from 16.42 to 116.11μmol N₂-N·m^-2·h^-1.The overall N₂O supersaturation degree in southern Jiangsu region was between 97%and 522.55%,with an average supersaturation degree of 175.11%.The analysis of N₂O supersaturation degree in Changshu area found that the N₂O supersaturation degree of fish ponds 1 and 2 was the highest in July,reaching 522.55%and 448.64%,respectively;while the lowest average N₂O supersaturation degree was 109.95%in June and the highest average N₂O supersaturation degree was242.49%in November.（2）In this study,the water quality inversion was further improved by further mining the data by the neural network model,and the accuracy of the overall sample value of chlorophyll was further improved to 0.70 on the basis of stepwise regression R²of 0.51,and the accuracy of turbidity was further improved to 0.72 on the basis of R²of 0.51.（3）In this study,three image feature parameters B/R2’,B/G2’,B/R10"of water denitrification rate were obtained by stepwise regression,an artificial neural network model was constructed,and the optimal topology was found to be 3-14-1（that is,the number of neurons in the input layer was set to 3,the number of neurons in the hidden layer was 14,and the number of neurons in the output layer was set to 1）,so as to realize the inversion monitoring of the denitrification rate of water based on smartphones.The final inversion result was 0.69 overall R²,0.87 for test set R²,and 34.13 for test set RMSE.（4）In this study,N₂O supersaturation and image feature parameters are gradually regressed,and finally 5 image special parameters R4",R5",B/G7’,R10’,G10",the optimal structure of N₂O supersaturation is the artificial neural network topology of 5-29-1,and the final inversion result is 0.51 overall R²,test set R²is 0.69,and test set RMSE is 0.43.This study explored for the first time the feasibility of using smartphones to invert the water quality and denitrification rate of small and micro water bodies.This provides a foundation and basis for the development of APP applications and inversion of other non-optical active parameters of water bodies.