83
Similar to the suite of open ocean Fe fertilization experiments that have been
undertaken in High Nutrient Low Chlorophyll (HNLC) regions of the world (Boyd
et al. 2007; Yoon et al. 2018), attempts have been made to stimulate natural populations of marine diazotrophs in situ. A PO 4
−3
release experiment termed CYCLOPS
in the eastern Mediterranean (Thingstad et al. 2005) did indicate some stimulation
of water column N 2 fixation (Rees et al. 2006). A similar field experiment in the
tropical N. Atlantic supplied both Fe and P (Rees et al. 2007) to surface waters but
saw no response.
5.11 Mathematical Models
Mathematical models provide a way to approach the large scale of the oceans as
well as complex interactions at the cellular scale, and are used to address many
oceanographic, biogeochemical and physiological questions. They are particularly
important in the age of “big data” on both the environmental and molecular biological sides in helping us organize our knowledge, to test what we think we know
(simulation modelling), help identify what we don’t know and develop new
hypotheses.
Many different types of models are used in marine N 2 fixation research including
forward running analytical (prognostic) models which are based on series of simultaneous and interacting equations which attempt to mimic and predict observations
(Hood and Christian 2008) (Table 5.1). Complementing traditional forward running
models are “inverse” models which work backwards from observations (Wang et al.
2019). New modelling approaches entering the field include the use of machine
learning (Tang et al. 2019).
Biogeochemical models that use dissolved nutrient distributions and their ratios,
N* and P*, and isotopic signatures coupled to known ocean circulation and flow are
used to predict when, where and how much N 2 fixation has or will occur (Deutsch
et al. 2007). Early modeling efforts often assumed fixed Redfield ratios because of
computational limitations (Hood and Christian 2008). However, computing power
has increased greatly and the current generation of models are much more sophisticate, provide inferences at higher temporal and spatial resolution and can allow for
non-Redfieldian ratios (Tang et al. 2020; Wang et al. 2019).
Models based on organism’s ecological characteristics are also used to forecast
distributions of this process (Hood et al. 2004; Inomura et al. 2018, 2019; Marconi
et al. 2017; Wang et al. 2019).
Trait based models were developed to forecast the distributions and dynamics of
phytoplankton in the oceans (Coles et al. 2017; Follows and Dutkiewicz 2011). This
principle has been applied to predict the large scale distributions of diazotrophs and
N 2 fixation (Dutkiewicz et al. 2012; Monteiro et al. 2010) (Fig. 7.5).
5.11 Mathematical Models
Similar to the suite of open ocean Fe fertilization experiments that have been
undertaken in High Nutrient Low Chlorophyll (HNLC) regions of the world (Boyd
et al. 2007; Yoon et al. 2018), attempts have been made to stimulate natural populations of marine diazotrophs in situ. A PO 4
−3
release experiment termed CYCLOPS
in the eastern Mediterranean (Thingstad et al. 2005) did indicate some stimulation
of water column N 2 fixation (Rees et al. 2006). A similar field experiment in the
tropical N. Atlantic supplied both Fe and P (Rees et al. 2007) to surface waters but
saw no response.
5.11 Mathematical Models
Mathematical models provide a way to approach the large scale of the oceans as
well as complex interactions at the cellular scale, and are used to address many
oceanographic, biogeochemical and physiological questions. They are particularly
important in the age of “big data” on both the environmental and molecular biological sides in helping us organize our knowledge, to test what we think we know
(simulation modelling), help identify what we don’t know and develop new
hypotheses.
Many different types of models are used in marine N 2 fixation research including
forward running analytical (prognostic) models which are based on series of simultaneous and interacting equations which attempt to mimic and predict observations
(Hood and Christian 2008) (Table 5.1). Complementing traditional forward running
models are “inverse” models which work backwards from observations (Wang et al.
2019). New modelling approaches entering the field include the use of machine
learning (Tang et al. 2019).
Biogeochemical models that use dissolved nutrient distributions and their ratios,
N* and P*, and isotopic signatures coupled to known ocean circulation and flow are
used to predict when, where and how much N 2 fixation has or will occur (Deutsch
et al. 2007). Early modeling efforts often assumed fixed Redfield ratios because of
computational limitations (Hood and Christian 2008). However, computing power
has increased greatly and the current generation of models are much more sophisticate, provide inferences at higher temporal and spatial resolution and can allow for
non-Redfieldian ratios (Tang et al. 2020; Wang et al. 2019).
Models based on organism’s ecological characteristics are also used to forecast
distributions of this process (Hood et al. 2004; Inomura et al. 2018, 2019; Marconi
et al. 2017; Wang et al. 2019).
Trait based models were developed to forecast the distributions and dynamics of
phytoplankton in the oceans (Coles et al. 2017; Follows and Dutkiewicz 2011). This
principle has been applied to predict the large scale distributions of diazotrophs and
N 2 fixation (Dutkiewicz et al. 2012; Monteiro et al. 2010) (Fig. 7.5).
5.11 Mathematical Models