One way of doing I can think of quickly is:

p = aqi[All /* Catenate /* (Partition[#, 2] &) /* GeoPosition, {"Latitude", "Longitude"}]

which will create GeoPosition arrays. You can generate map:

GeoListPlot[p]

pos = Values[Normal[aqi[All, {"Latitude", "Longitude", "AQI"}]]] /. {lat_, long_, aqi_} :> Thread[GeoPosition[{lat, long}] -> aqi];

GeoSmoothHistogram[Association[pos], ColorFunction -> "Rainbow", 
Mesh -> 20, MeshStyle -> Directive[Opacity[0.085], Blue], 
GeoBackground -> "StreetMap"]

You need to generate GeoElevationData :

   ged= GeoElevationData /@(Values[Normal[aqi[All, {"Latitude", "Longitude"}]]] /. {lat_, long_} :> GeoPosition[{lat, long}])

Check with GeoBackground -> "ReliefMap" option in the above plot. Play with GeoElevationData too. Have been just throwing quick solutions while commuting...

It's amazing! Different instructions produce different data structures! I am an environmental epidemiology student and mathematica beginner. I would like to use public data to observe the distribution of air pollution concentrations but plagued by a large number of monitoring station coordinates and observations (CSV file). The coordinates and observations of these coordinates and observations are converted to GeoPosition or GeoShothistogram every day. Now that the CSV file has converted to Dataset, and establish a long-term database and the workload reduced. The next step is to convert the station coordinates to GeoPosition. The coordinate data is already known as aqi[[;;,{-2,-1}]]. How to put coordinates into GeoPosition?

   aqi = Import[
          "http://opendata.epa.gov.tw/ws/Data/AQI/?$format=csv", {"CSV", 
           "Dataset"}, HeaderLines -> 1];

I have tried the following instructions but it does not work

p={#Latitude, #Longitude} & /@ aqi;GeoListPlot[p]

These two approaches give the same result:

AssociationThread[keys -> #] & /@ values
(*{<|"a" -> 1, "b" -> 2, "c" -> 3|>, <|"a" -> 4, "b" -> 5, "c" -> 6|>, <|"a" -> 7, "b" -> 8, "c" -> 9|>}*)

Association /@ (MapThread[Rule, {keys, #}] & /@ values)
(*{<|"a" -> 1, "b" -> 2, "c" -> 3|>, <|"a" -> 4, "b" -> 5, "c" -> 6|>, <|"a" -> 7, "b" -> 8, "c" -> 9|>}*)

Three approaches:

SemanticImportString[ExportString[temp, "CSV"]]

If your data is starting in a file (ie CSV or TSV file), you can go direct from file to dataset with SemanticImport[].

Regards

temp=AssociationThread[ToString /@ {a, b, c} -> #] & /@ {{1, 2, 3}, {4, 5, 6}, {7, 8, 9}};
Dataset@temp

@Henrik Schachner : Not at all - the more the merrier!