Biology Reference
In-Depth Information
damage to the crop. Predictive models are usually constructed with
many variables, such as, in the case of crop-damage models, daily aver-
age temperature, sunlight, soil moisture, age of the crop, and reproduc-
tive and developmental biology of insects. Which of these variables are
necessary or sui cient is ot en unknown and irrelevant to the ability of
the model to make reasonable predictions. Alan Gelfand and Crayton
Walker called these models “Mulligan stew” models in their topic
En-
semble Modeling
at er a culinary dish reputed to have been made by
unemployed, homeless hoboes in the United States at the turn of the
twentieth century. It was a stew concocted from whatever ingredients
were available. h e individual ingredients were not important, only the
taste and nutrition of the end product.
Explanatory models are designed to understand how something
works and to seek out those variables that are necessary and sui cient to
explain the system. Gelfand and Walker call models of this type “stone-
soup” models at er a contested story of European origins in which hun-
gry travelers came to a village and asked for food. h e villagers were
unwilling to share their food, so the travelers set up a cooking pot over
a i re in the central square of the village and started boiling a stone. A
curious villager asked them what they were doing, and they explained
that they were making stone soup but were missing a few ingredients
needed to improve the l avor. h e villager shared some small item of
food. Another curious villager came by, asked the same question, and
shared some other small item of food. Eventually, the travelers had a
soup that was tasty and nutritious. Stone-soup models start with the
minimum parameters and add more only as needed to explain the sys-
tem studied.
Explanatory models can be built and then tested by comparing a
specii c outcome with the model's predictions based on hypothesized
relationships among a set of variables. h e models are modii ed and
sequentially retested until a good i t is achieved, as in the case of opti-
mality models in evolutionary biology that attempt to predict an opti-
mal phenotype (the observable traits of an individual, or a colony in the
case of social insects) for a given environment given a specii c set of
parameters and constraints. Or groups of models can be constructed
