1 | % ================================================================ |
---|
2 | % Chapter Assimilation increments (ASM) |
---|
3 | % ================================================================ |
---|
4 | \chapter{Apply assimilation increments (ASM)} |
---|
5 | \label{ASM} |
---|
6 | |
---|
7 | Authors: D. Lea, M. Martin, K. Mogensen, A. Weaver, ... % do we keep |
---|
8 | |
---|
9 | \minitoc |
---|
10 | |
---|
11 | |
---|
12 | \newpage |
---|
13 | $\ $\newline % force a new line |
---|
14 | |
---|
15 | The ASM code adds the functionality to apply increments to the model variables: |
---|
16 | temperature, salinity, sea surface height, velocity and sea ice concentration. |
---|
17 | These are read into the model from a NetCDF file which may be produced by separate data |
---|
18 | assimilation code. The code can also output model background fields which are used |
---|
19 | as an input to data assimilation code. This is all controlled by the namelist |
---|
20 | \textit{\ngn{nam\_asminc} }. There is a brief description of all the namelist options |
---|
21 | provided. To build the ASM code \key{asminc} must be set. |
---|
22 | |
---|
23 | %=============================================================== |
---|
24 | |
---|
25 | \section{Direct initialization} |
---|
26 | \label{ASM_DI} |
---|
27 | |
---|
28 | Direct initialization (DI) refers to the instantaneous correction |
---|
29 | of the model background state using the analysis increment. |
---|
30 | DI is used when \np{ln\_asmdin} is set to true. |
---|
31 | |
---|
32 | \section{Incremental Analysis Updates} |
---|
33 | \label{ASM_IAU} |
---|
34 | |
---|
35 | Rather than updating the model state directly with the analysis increment, |
---|
36 | it may be preferable to introduce the increment gradually into the ocean |
---|
37 | model in order to minimize spurious adjustment processes. This technique |
---|
38 | is referred to as Incremental Analysis Updates (IAU) \citep{Bloom_al_MWR96}. |
---|
39 | IAU is a common technique used with 3D assimilation methods such as 3D-Var or OI. |
---|
40 | IAU is used when \np{ln\_asmiau} is set to true. |
---|
41 | |
---|
42 | With IAU, the model state trajectory ${\bf x}$ in the assimilation window |
---|
43 | ($t_{0} \leq t_{i} \leq t_{N}$) |
---|
44 | is corrected by adding the analysis increments for temperature, salinity, horizontal velocity and SSH |
---|
45 | as additional tendency terms to the prognostic equations: |
---|
46 | \begin{eqnarray} \label{eq:wa_traj_iau} |
---|
47 | {\bf x}^{a}(t_{i}) = M(t_{i}, t_{0})[{\bf x}^{b}(t_{0})] |
---|
48 | \; + \; F_{i} \delta \tilde{\bf x}^{a} |
---|
49 | \end{eqnarray} |
---|
50 | where $F_{i}$ is a weighting function for applying the increments $\delta |
---|
51 | \tilde{\bf x}^{a}$ defined such that $\sum_{i=1}^{N} F_{i}=1$. |
---|
52 | ${\bf x}^b$ denotes the model initial state and ${\bf x}^a$ is the model state |
---|
53 | after the increments are applied. |
---|
54 | To control the adjustment time of the model to the increment, |
---|
55 | the increment can be applied over an arbitrary sub-window, |
---|
56 | $t_{m} \leq t_{i} \leq t_{n}$, of the main assimilation window, |
---|
57 | where $t_{0} \leq t_{m} \leq t_{i}$ and $t_{i} \leq t_{n} \leq t_{N}$, |
---|
58 | Typically the increments are spread evenly over the full window. |
---|
59 | In addition, two different weighting functions have been implemented. |
---|
60 | The first function employs constant weights, |
---|
61 | \begin{eqnarray} \label{eq:F1_i} |
---|
62 | F^{(1)}_{i} |
---|
63 | =\left\{ \begin{array}{ll} |
---|
64 | 0 & {\rm if} \; \; \; t_{i} < t_{m} \\ |
---|
65 | 1/M & {\rm if} \; \; \; t_{m} < t_{i} \leq t_{n} \\ |
---|
66 | 0 & {\rm if} \; \; \; t_{i} > t_{n} |
---|
67 | \end{array} \right. |
---|
68 | \end{eqnarray} |
---|
69 | where $M = m-n$. |
---|
70 | The second function employs peaked hat-like weights in order to give maximum |
---|
71 | weight in the centre of the sub-window, with the weighting reduced |
---|
72 | linearly to a small value at the window end-points: |
---|
73 | \begin{eqnarray} \label{eq:F2_i} |
---|
74 | F^{(2)}_{i} |
---|
75 | =\left\{ \begin{array}{ll} |
---|
76 | 0 & {\rm if} \; \; \; t_{i} < t_{m} \\ |
---|
77 | \alpha \, i & {\rm if} \; \; \; t_{m} \leq t_{i} \leq t_{M/2} \\ |
---|
78 | \alpha \, (M - i +1) & {\rm if} \; \; \; t_{M/2} < t_{i} \leq t_{n} \\ |
---|
79 | 0 & {\rm if} \; \; \; t_{i} > t_{n} |
---|
80 | \end{array} \right. |
---|
81 | \end{eqnarray} |
---|
82 | where $\alpha^{-1} = \sum_{i=1}^{M/2} 2i$ and $M$ is assumed to be even. |
---|
83 | The weights described by \eqref{eq:F2_i} provide a |
---|
84 | smoother transition of the analysis trajectory from one assimilation cycle |
---|
85 | to the next than that described by \eqref{eq:F1_i}. |
---|
86 | |
---|
87 | %========================================================================== |
---|
88 | % Divergence damping description %%% |
---|
89 | \section{Divergence damping initialisation} |
---|
90 | \label{ASM_details} |
---|
91 | |
---|
92 | The velocity increments may be initialized by the iterative application of |
---|
93 | a divergence damping operator. In iteration step $n$ new estimates of |
---|
94 | velocity increments $u^{n}_I$ and $v^{n}_I$ are updated by: |
---|
95 | \begin{equation} \label{eq:asm_dmp} |
---|
96 | \left\{ \begin{aligned} |
---|
97 | u^{n}_I = u^{n-1}_I + \frac{1}{e_{1u} } \delta _{i+1/2} \left( {A_D |
---|
98 | \;\chi^{n-1}_I } \right) \\ |
---|
99 | \\ |
---|
100 | v^{n}_I = v^{n-1}_I + \frac{1}{e_{2v} } \delta _{j+1/2} \left( {A_D |
---|
101 | \;\chi^{n-1}_I } \right) \\ |
---|
102 | \end{aligned} \right., |
---|
103 | \end{equation} |
---|
104 | where |
---|
105 | \begin{equation} \label{eq:asm_div} |
---|
106 | \chi^{n-1}_I = \frac{1}{e_{1t}\,e_{2t}\,e_{3t} } |
---|
107 | \left( {\delta _i \left[ {e_{2u}\,e_{3u}\,u^{n-1}_I} \right] |
---|
108 | +\delta _j \left[ {e_{1v}\,e_{3v}\,v^{n-1}_I} \right]} \right). |
---|
109 | \end{equation} |
---|
110 | By the application of \eqref{eq:asm_dmp} and \eqref{eq:asm_dmp} the divergence is filtered |
---|
111 | in each iteration, and the vorticity is left unchanged. In the presence of coastal boundaries |
---|
112 | with zero velocity increments perpendicular to the coast the divergence is strongly damped. |
---|
113 | This type of the initialisation reduces the vertical velocity magnitude and alleviates the |
---|
114 | problem of the excessive unphysical vertical mixing in the first steps of the model |
---|
115 | integration \citep{Talagrand_JAS72, Dobricic_al_OS07}. Diffusion coefficients are defined as |
---|
116 | $A_D = \alpha e_{1t} e_{2t}$, where $\alpha = 0.2$. The divergence damping is activated by |
---|
117 | assigning to \np{nn\_divdmp} in the \textit{nam\_asminc} namelist a value greater than zero. |
---|
118 | By choosing this value to be of the order of 100 the increments in the vertical velocity will |
---|
119 | be significantly reduced. |
---|
120 | |
---|
121 | |
---|
122 | %========================================================================== |
---|
123 | |
---|
124 | \section{Implementation details} |
---|
125 | \label{ASM_details} |
---|
126 | |
---|
127 | Here we show an example \ngn{namasm} namelist and the header of an example assimilation |
---|
128 | increments file on the ORCA2 grid. |
---|
129 | |
---|
130 | %------------------------------------------namasm----------------------------------------------------- |
---|
131 | \namdisplay{namasm} |
---|
132 | %------------------------------------------------------------------------------------------------------------- |
---|
133 | |
---|
134 | The header of an assimilation increments file produced using the NetCDF tool |
---|
135 | \mbox{\textit{ncdump~-h}} is shown below |
---|
136 | |
---|
137 | \begin{alltt} |
---|
138 | \tiny |
---|
139 | \begin{verbatim} |
---|
140 | netcdf assim_background_increments { |
---|
141 | dimensions: |
---|
142 | x = 182 ; |
---|
143 | y = 149 ; |
---|
144 | z = 31 ; |
---|
145 | t = UNLIMITED ; // (1 currently) |
---|
146 | variables: |
---|
147 | float nav_lon(y, x) ; |
---|
148 | float nav_lat(y, x) ; |
---|
149 | float nav_lev(z) ; |
---|
150 | double time_counter(t) ; |
---|
151 | double time ; |
---|
152 | double z_inc_dateb ; |
---|
153 | double z_inc_datef ; |
---|
154 | double bckint(t, z, y, x) ; |
---|
155 | double bckins(t, z, y, x) ; |
---|
156 | double bckinu(t, z, y, x) ; |
---|
157 | double bckinv(t, z, y, x) ; |
---|
158 | double bckineta(t, y, x) ; |
---|
159 | |
---|
160 | // global attributes: |
---|
161 | :DOMAIN_number_total = 1 ; |
---|
162 | :DOMAIN_number = 0 ; |
---|
163 | :DOMAIN_dimensions_ids = 1, 2 ; |
---|
164 | :DOMAIN_size_global = 182, 149 ; |
---|
165 | :DOMAIN_size_local = 182, 149 ; |
---|
166 | :DOMAIN_position_first = 1, 1 ; |
---|
167 | :DOMAIN_position_last = 182, 149 ; |
---|
168 | :DOMAIN_halo_size_start = 0, 0 ; |
---|
169 | :DOMAIN_halo_size_end = 0, 0 ; |
---|
170 | :DOMAIN_type = "BOX" ; |
---|
171 | } |
---|
172 | \end{verbatim} |
---|
173 | \end{alltt} |
---|