Main Drivers Of Carbon And Water Flux At Temperate Deciduous Forests And Evergreen Needleleaf Forests In North America

Terrestrial carbon and water flux is one of the key components of global climate change and regional sustainable development.The net ecosystem exchange of CO₂?NEE?between the biosphere and the atmosphere in forests accounts for over 90%of the total carbon exchange of the terrestrial ecosystem and atmospheric.Human activities and climate changes have strong impact on carbon and water cycles in global forests.At present,the research mainly focuses on the selection of factors affecting the carbon and water flux through the use of expert experience to establish predictions and simulation models,but not on in-depth analysis and mining of carbon and water flux drivers.Furthermore,most attribution studies focus only on the linear correlation between NEE or Evapotranspiration?ET?and one or two factors individually,and ignored the potential interactive effects among the multifactor combinations on NEE or ET,resulting in the existing model cannot accurately estimate the NEE or ET,affecting the process of carbon and water cycle mechanism research.Hence,in this study we took advantage of long-term flux records in North America and proposed a fuzzy rough set algorithm with binary shuffled frog leaping?BSFL-FRSA?to calculated and identified the contribution rate of both main driving factors individually and multifactor in combination,and then used correlation networks to evaluate interactions of multifactor combinations in influencing NEE and ET.The primary conclusions and contributions of the paper are listed as follows:?1?This study proposed a fuzzy rough set algorithm with binary shuffled frog leaping?BSFL-FRSA?to identify main driving factors and multiple-factor combinations,and define the contribution rate of them to NEE or ET.In order to fit non-linear relations between uncertainty factors and NEE and ET,BSFL-FRSA was applied to calculated and identified the contribution rate of both main driving factors individually and multifactor in combination to variations in NEE and ET at temperate deciduous forests and evergreen needleleaf forest in the North America.To evaluate the performance of the BSFL-FRSA for NEE and ET,the accuracy of BP?back-propagation?neural network prediction model with both all initial factors and the BSFL-FRSA selected ones was used.?2?This study calculated and identified the contribution rate of both main driving factors individually and multifactor in combination to variations in NEE at temperate deciduous forests and evergreen needleleaf forest sites across the North America.Temperature was the most influential factors for NEE at temperate deciduous forest and evergreen needleleaf forest.The importance of five main driving factors to NEE across temperate deciduous forests sites was in the order:TS?soil temperature?>TA?air temperature?>NETRAD?net radiation?>RH?relative humidity?>SWC?soil water content?with mean contribution rate of 59.6%,46.7%,22.1%,13.4%,12.9%and 7.6%,respectively.The importance of five main driving factors to NEE across all evergreen needleleaf forest sites was in the order:TA>TS>CCO₂?atmospheric CO₂ concentration?>NETRAD>SWC,with a mean contribution rate of 45.2%,20.6%,20.5%,12.0%and 9.8%,respectively.Of all two-factor combinations,[TS,NETRAD]was observed to be the most dominant at temperate deciduous forests?with an average contribution of 72.7%?and that of[TA,NETRAD]at evergreen needleleaf forest?with an average contribution of 64.5%?.It is indicated that temperature and radiation have a significant effect on NEE.Of three-factor combinations,[TS,NETRAD,SWC/RH]was observed to be the most dominant at temperate deciduous forests?with an average contribution of 79.2%?and that of[TA,NETRAD,CCO₂]at evergreen needleleaf forest?with an average contribution of 72.2%?.?3?This study calculated and identified the contribution rate of both main driving factors individually and multifactor in combination to variations in ET at temperate deciduous forests and evergreen needleleaf forest sites across the North America.Temperaturewas the most influential factors for ET at temperate deciduous forestand evergreen needleleaf forest.The importance of influential five factors to ET across temperate deciduous forests sites was in the order:TS>TA>NETRAD>RH>SWC,with a mean contribution rate of 48.36%,45.83%,23.82%,13.76%and 6.45%,respectively.The importance of influential five factors to ET across all evergreen needleleaf forest sites was in the order:TA>CCO₂>TS>SWC>NETRAD,with a mean contribution rate of 47.7%,36.2%,32.2%,17.8%and 13.1%,respectively.Of all two-factor combinations,[TS,NETRAD]was observed to be the most dominant at temperate deciduous forests?with an average contribution of 65.4%?and that of[TA,CCO₂]at evergreen needleleaf forest?with an average contribution of 70.7%?.Of three-factor combinations,[TS,NETRAD,SWC]was observed to be the most dominant at temperate deciduous forests?with an average contribution of 78.5%?and that of[TA,NETRAD,CCO₂]at evergreen needleleaf forest?with an average contribution of 81.7%?.It is indicated that temperature and radiation have a significant effect on ET.?4?This study calculated and identified the contribution rate of both main driving factors individually and multifactor in combination to variations in NEE and ET in evergreen needleleaf forest at different ages,emphasizing the ages variations of the main driving factors and multifactor in combination among different ages.For different age forests,the control factors on NEE and ET are different.Of all factors,SWC was the most influential factors for NEE at 5 year old forest,and that of TS at 72 year old forest,and that of air temperature?TA?at other.[TS,NETRAD]?Of all two-factor combinations,[TS,SWC]was observed to be the most dominant at 5 year old and 22 year old forest,and that of[TA,TS]at153 year old and 13 year old forest,and that of[TA,CCO₂]at 39 year old and73 year old forest.TS was observed to be the most dominant for variation in ET at 73a site,and that of TA at other.Of all two-factor combinations,[TS,SWC]was observed to be the most dominant at 5 year old and 153 year old forests,and that of[TA,TS]at 13 year old,39 year old and 73 year old forest,and that of[TA,CCO₂]at21 year old forest.This study proposed a fuzzy rough set algorithm with binary shuffled frog leaping?BSFL-FRSA?to calculated and identified the contribution rate of both main driving factors individually and multifactor in combination to variations in NEE and ET at temperate deciduous forests and evergreen needleleaf forest sites across the North America.This study helps understand the influence of global climate change on carbon and water cycles and processes in forest ecosystems,and provides scientific support for the efforts that improve production capacity and increase the carbon sequestration capacity in the forest.The study contributes to the investigation for alleviating water crisis in the forest,and improving the use efficiency of water resources,and supporting regional environmental sustainability.