A wet expression can either be an underwater city $\wet w$, an expanse of water with nothing in it $\wet 1$, a pair of wet regions next to each other $\wet{W_1 \ox W_2}$, or an island $\I(\dry{D})$, containing a dry expression.
A dry expression can either be a dry-land city $\dry d$, an expanse of land with nothing in it $\dry 1$, a pair of dry regions next to each other $\dry{D_1 \ox D_2}$, or a lake $\L(\wet{W})$, containing a wet expression.
For example if $\wet{W_1}$ denotes the map labelled ① above, and $\wet{W_2}$ denotes the map labelled ② above, that is, \[ \wet{W_1} = \wet I(\dry L(\wet{o \ox n\ox I}(\dry{a \ox L}\wet m))\dry{{}\ox b \ox c \ox L}\wet 1) \wet{{}\ox p\ox I}\dry{1} \wet{\ox I}(\dry{d \ox L}\wet{r}\dry{ \ox L}\wet q)\] \[ \wet{W_2} = \wet I(\dry L(\wet{o \ox n\ox I}\dry a\wet{\ox m})\dry{\ox c \ox L}\wet I\dry 1) \wet{{}\ox I{\dry b} \ox p} \wet{\ox I}(\dry{d \ox L}(\wet{r \ox q}))\] then \[ \wet{W_1} \le \wet{W_2}\]
$ \wet{I} \dry 1 \le \wet 1 $ | (delete an island) |
$ \dry 1 \le \dry{L} \wet 1 $ | (create a lake) |
$ \wet{I} (\dry{D_1 \ox D_2}) \le \wet I\dry{D_1} \wet{\ox} \wet I\dry{D_2} $ | (split a landmass) |
$ \dry L\wet{W_1} \dry{\ox} \dry L\wet{W_2} \le \dry{L} (\wet{W_1 \ox W_2} ) $ | (merge two bodies of water) |
$ \wet I\dry L\wet W \le \wet{ W} $ | (delete ring island) |
$ \dry{D} \le \dry L\wet I \dry{ D} $ | (create a moat) |
For the first-order structure, we suppose we have two sorts of first-order terms, wet terms $\wet \omega$ and dry terms $\dry \delta$, and a binary relation symbol (that is, an atomic proposition taking two arguments, one wet, one dry) which we will write infix as $\wet \omega \tri \dry \delta$. For each named 'city' in the source language, we assume one unary relation symbol named after that city, which takes a wet or dry term argument according to whether the city is underwater or on land.
I want to define a translation function that takes a wet 'proposition' $\wet W$ and wet term $\wet \omega$ to a first-order linear logic proposition $\wet{W \wat \omega}$, and a function that takes a dry 'proposition' $\dry D$ and a dry term $\dry \delta$ and outputs a first-order linear logic proposition $\dry{D \wat \delta}$.
The definition proceeds by induction on expressions, like so: \[ (\wet{W_1 \ox W_2})\wet{ \wat \omega} = (\wet{W_1 \wat \omega}) \ox (\wet{W_2 \wat \omega}) \qquad \wet {1 \wat \omega} = 1\] \[ (\dry{D_1 \ox D_2})\dry{ \wat \delta} = (\dry{D_1 \wat \delta}) \ox (\dry{D_2 \wat \delta}) \qquad \dry{1\wat \delta} = 1\] \[ {\wet I}{\dry D} \wet{\wat \omega} = \exists \dry \delta . !(\wet\omega\tri\dry \delta ) \ox (\dry {D \wat \delta}) \] \[ {\dry L}{\wet W} \dry{\wat \delta} = \forall \wet \omega . !(\wet\omega\tri\dry \delta ) \lol (\wet {W \wat \omega}) \] \[ \wet {w \wat\omega} = \wet w(\wet \omega) \] \[ \dry {d \wat\delta} = \dry d(\dry \delta) \] Conjecture: for any wet expressions $\wet{W_1}$ and $\wet{W_2}$, we have $\wet{W_1} \le \wet{W_2}$ if and only if $\wet{W_1 \wat \omega} \prov \wet{W_2 \wat \omega}$ in first-order intuitionistic linear logic.