Research On Risk Assessment Of Cyanobacterial Bloom And Spatio-Temporal Law Mining On Diurnal Bloom Changes In Lake Taihu Based On Multi-Source Data

As the rapid economic development and the intensification of human activities in China,the increasing eutrophication has been caused due to the discharge of massive pollutants into lake.Under this circumstance,cyanobacterial blooms in lakes have been catalyzed,which degrade the water quality and threaten the aquatic ecological safety.A large number of research institutions have carried out the risk management,pre-control and mitigation of cyanobacteria bloom in lakes.Among these,the risk assessment of cyanobacterial blooms and monitoring of bloom dynamics are critical work in guiding the formulation and implementation of disaster prevention and mitigation policies.However,it still has the problems need to be solved in the research on risk assessment of cyanobacterial bloom and spatio-temporal law mining on diurnal bloom changes.On the one hand,current researches on cyanobacterial bloom risk assessment are often conducted at the specific site and require a variety of input data(i.e.,ground station data and satellite remote sensing data).And thus,it is difficult to take into account both the richness of assessment indicators and the spatial continuity of assessment results.On the other hand,advantage of remote sensing technology on large-scale continuous monitoring provides the possibility of spatio-temporal law mining research.Despite that,the spatio-temporal law mining of bloom change patterns and its continuous process in a day is still not done adequately.First,the characteristics of bloom dynamics in the adjacent hours are mainly identified by the visual interpretation of RGB maps of cyanobacterial distribution,which is an inefficient way for the classification of diurnal bloom change patterns.Consequently,it difficult to realize a large-scale and automatic classification for its diurnal bloom change pattern.Second,the observation frequency and transit time of satellite are fixed(i.e.,GOCI observes eight time a day),and the satellite data may be invalid when the weather is cloudy or rainy.Therefore,it is difficult to realize all-weather continuous monitoring on the diurnal bloom process only relying on the effective observations of high-frequency satellite.Herein,a series of methods for the risk assessment of cyanobacterial blooms,and the bloom dynamics monitoring/forecasting have been proposed to solve the mentioned problems,which are then used for the spatio-temporal law mining on cyanobacterial bloom risks and diurnal bloom changes.Subsequently,Lake Taihu has been selected as the main study area due to its severe cyanobacterial blooms.With the use of multi-source data(i.e.,satellite remote sensing data,reanalysis meteorological data,remote sensing inversion product),the spatial continuous risk assessment and spatio-temporal continuous of bloom dynamics monitoring/forecasting have been realized.Based on these,the spatio-temporal mining of cyanobacteria bloom risk and diurnal bloom change have been conducted in Lake Taihu.The research contents are as follows.(1)In this study,a multivariable integrated risk assessment method was proposed for pixel-scale risk level of cyanobacterial blooms,which solves the problems of single evaluation index and spatial discontinuity of evaluation results by existing methods.In this method,the key indicators for cyanobacterial bloom risk assessment were selected according to the prior knowledge on driving force.Besides,the availability of the data in spatial and temporal resolution is also considered.The possibility and potential consequences of the five risk factors inducing the bloom were comprehensively evaluated with a clearly defined formula of cyanobacterial bloom risk index.The pixel-scale risk level of cyanobacterial blooms in three typical eutrophication lakes(Lake Taihu,Lake Chaohu and Dianchi Lake Chaohu)in China during from 2002 to 2020were estimated by using this method.Compared with other risk assessment results,the effectiveness of this method was verified(R~2=0.43,P-value<0.0001).Based on this,the spatio-temporal characteristics and the potential impact factors of cyanobacterial bloom risk among lakes(or in a lake)were systematically analyzed.In conclusion,the method proposed in this study realizes the spatial continuous risk assessment of cyanobacterial blooms with the consideration on the impact of multiple factors.It broadens the risk assessment efficiency and its implementation spatial scale,and can assist with the regional risk management of cyanobacterial blooms in a lake.(2)This study designed a diurnal change patterns classification(DCPC)method for cyanobacterial blooms,which solves the current manual interpretation method cannot identify the characteristics of bloom dynamics in the adjacent hours in a day.Based on the systematic analysis on the hourly bloom characteristics in a day,the classification criterion was clearly defined by formulas in this method.With eight GOCI observations(from～08:00 h to～15:00 h,per hour,UTC+8),this method can identify the types of diurnal bloom change patterns automatically.These types are the decreasing(Type1),decreasing first and then increasing(Type2),increasing(Type3),increasing first and then decreasing(Type4).And then,a pixel-scale classification was conducted for Lake Taihu from 2011 to 2020 using the mentioned method.The method reliability was verified by comparing with other interpretation results(the classification accuracy is more than 80%).Subsequently,the intensity of diurnal bloom changes was calculated.The spatio-temporal characteristics of diurnal bloom changes in Lake Taihu were analyzed systematically,including the hot patterns,seasons and hotspots.And then,the meteorological impacts(i.e.,temperature and wind speed)was discussed.In conclusion,compared with the existing manual interpretation method,the method designed in this study realizes a spatial continuous classification on the types of diurnal bloom change pattern,which improves classification efficiency and broadens identification spatial-scale.It can provide reference information for bloom dynamic management in a lake.(3)A prediction model of cyanobacterial blooms driven by meteorological data was constructed at hourly scale in this study,which solves the problem that the current studies cannot realize all-weather and spatio-temporal continuous monitoring of diurnal bloom process in a lake only relying on the effective observation of high-frequency satellites.Based on the areal coverage of cyanobacterial bloom and metrological data at eight moments in a day,the model was constructed under pre-defined spatiotemporal scenarios using the Light Gradient Boosting machine(Light GBM)method.The model with the best accuracy was selected after model comparison.Then,all-weather gridded meteorological data was inputted to this model,and the area coverage of cyanobacterial blooms at 24 moments in a day was predicted for Lake Taihu with the spatial resolution of 9 km(from 2011 to 2020).Based on this,the spatio-temporal characteristics on the areal coverage of cyanobacterial blooms in Lake Taihu were analyzed,and the environmental impacts were discussed.In conclusion,the model constructed in this study realizes the all-weather and spatio-temporal continuous prediction on the area coverage of cyanobacteria blooms in Lake Taihu,which can assist in the generation of cyanobacterial bloom data during periods when satellite data are missing or unavailable.It is one of the key technologies for the risk assessment and diurnal monitoring of bloom dynamics in a lake.And this study provides a new data source for the spatio-temporal mining of cyanobacterial bloom risk and its diurnal dynamics.Finally,this study selected Lake Taihu as a typical research area as its severe bloom problem.Combining with the cyanobacterial bloom data generated in research contents(3),all-weather and spatio-temporal risk levels of cyanobacterial blooms in Lake Taihu was estimated using the risk assessment method proposed in research contents(1).The diurnal change process(from～00:00 to～23:00,per hour)and regional characteristics on risk level and areal coverage of cyanobacterial blooms were compared and analyzed.Then,combining with the classification results of diurnal bloom change patterns in research contents(2),the diurnal change process and regional characteristics on risk level and areal coverage of cyanobacterial blooms risks were then summarized for four types of diurnal bloom change pattern in Lake Taihu.In conclusion,the methods proposed in this study can assist with risk management and diurnal dynamic monitoring of cyanobacterial blooms in other similar eutrophic lakes.