Examining and addressing the Wallacean shortfall : species distribution modelling and biodiversity patterns of hawkmoths in the old world

The geographic distribution of the species is one of the basic units of information in ecology and
biogeography, yet for the vast majority of the species in the world is quite unknown. It is a
phenomenon called the Wallacean shortfall. The knowledge available about the distribution is
biased to few certain taxa and regions. For insects, in particular, this information is even sparser
despite being very specious taxa and play an important ecological role. Species Distribution
Modelling (SDMs) offers a potentially powerful tool that might help to fill those gaps in our
knowledge especially for those species from which there is very little known about their ecology
and places where collecting has been very scarce. In this thesis, I used a database compiled from
museum and private collections, publications (including online databases) and fieldwork data
already assembled by my co-authors.
Such database contains over 109,880 distributional records of the global distribution for all the 982
non-American taxa of the Sphingidae family of Lepidoptera which then I combined with SDMs
algorithms to provide high-resolution distribution maps for all the taxa in the family and study
patterns of biodiversity. Since the purpose of this document is to provide stand alone manuscripts
that are at the point of submission are either submitted, in review or published, I will refer to “we”
throughout much of the text.
As a first step, we compared the performance of 8 commonly used SDM’s algorithms while
considering some intrinsic properties of the species and data with a representative sample of the
species in the family (Chapter 2). The algorithm that performed the best was Maxent followed by
Random Forest, however we could not confirm effects of species traits or data properties
influencing the modeling performance.
Subsequently, in Chapter 3 we assessed the value of different data sources, by comparing an
independent compilation of occurrence data vs GBIF database, and its contribution to different
aspects of the range of the species (i.e. range filling, range extent and climatic niche space). GBIF
provided more records than other sources though contributed with less information about the range,
so it is not yet an alternative to manual compilation of distributional data.
Species diversity patterns based on numerical estimators are studied in Chapter 4, in relation with
their main environmental correlates for a fraction of the study region. We also provided assessment
of inventory completeness in the same region. Variables describing vegetation emerged as
important predictors of species richness. Variables capturing heat, energy availability and
topographic heterogeneity were identified as further parameters influencing species richness.
Inventory completeness is positively associated with densely populated areas, accessibility,
protected areas and colonial history. We discussed how this approach sets the baseline to estimate
diversity patterns in under-studied taxa.
A detailed documentation of data acquisition, processing and modeling is compiled in Chapter 5.
We applied the modeling technique chosen in Chapter 2 in combination with environmental data
and vegetation cover data to the whole dataset. We could retrieve models for 789 taxa whereas we
provided expert drawn range maps for the remaining 193. In general, annual temperature range was
the factor contributing the most to shape species’ distributions followed by variables related to
precipitation. Variables related to vegetation did not highly contribute. In a next step, we
superimposed the resultant grids to study patterns of biodiversity at two spatial scales (a = 5 x 5 km
and ? = 200 x 200 km) and then used them to calculate ß-diversity. The a and ? diversity maps
exhibited a latitudinal gradient of species richness towards the tropics whereas ß-diversity patterns
revealed rather a altitudinal gradient, higher in mountainous regions and along biogeographical
boundaries. This set of maps is the result of a collaborative project that to the best of our knowledge
compiles the first distributional data set for a complete family of invertebrates at an almost global
scale. Achievements, challenges and limitations of the project are also reported and discussed.
A specific application of SDM is shown in Chapter 6. We predicted the potential the range of an
invasive species (Agrius cingulata), native to the American continent which have recently spread
and established populations in Africa. We used two types of SDM based on native range records
and environmental data. Our results showed that Agrius cingulata could find suitable habitat across
wide stretches across Sub-saharan Africa. Early monitoring programs might be valuable to evaluate
the status of the invasion.
In Chapter 7, a general discussion of the results plus an outlook to further research with this data is
presented. All the maps (i.e. from raw data, intermediate steps until the final map) together with
appendixes containing details of the models, list of species, literature and museum collection
sources are deposited on the network drive at the University Computing Centre of Basel. It is our
plan for the future to make this database available throughout the website facility: The map of life
(http://www.mappinglife.org/).