Manipulating GeoPlots Effectively

Added sample data for performance testing -- see Update below.

The Manipulate controls for latitude and longitude allow the minimum values to be greater than the maximums. Is this the problem you want to fix in a more natural way?

I suggest a better method is to use latitude and longitude controls for the center of a region, and separate controls for the dimensions of the region. The dimensions can be a distance such as kilometers or miles. This way, the minimum and maximum values will always be valid (so there's no need to do checks[ ... ]). Here's a simple example that computes the values of minlat, maxlat, minlng, and maxlng.

Manipulate[
 Column[{
   {minlat, maxlat} = 
     GeoDestination[GeoPosition[{a, b}], 
       GeoDisplacement[{Quantity[c/2, "Kilometers"], {180, 0}}]][[1, All, 1]],
   {minlng, maxlng} = 
     GeoDestination[GeoPosition[{a, b}], 
       GeoDisplacement[{Quantity[d/2, "Kilometers"], {270, 90}}]][[1, All, 2]]
   }],
  {{a, 49.24, "Center Latitude"}, 49.18, 49.41, 0.00001}, 
  {{b, -123.15, "Center Longitude"}, -123.2, -123.0, 0.00001}, 
  {{c, 5, "Latitude Dimension"}, 1, 10, 0.1},
  {{d, 6, "Longitude Dimension"}, 1, 10, 0.1}
]

---

Update:

We need some sample data to do performance timing. Make random points centered on the Manipulate coordinate ranges.

{minlat, maxlat, minlng, maxlng} = {49.24, 49.292, -123.15, -123.05};
c = {Mean[{minlat, maxlat}], Mean[{minlng, maxlng}]}; (*center*)
ϕ = GeoDestination[c, {Quantity[10, "Kilometers"], {180, 0}}][[1, All, 1]];
λ = GeoDestination[c, {Quantity[10, "Kilometers"], {270, 90}}][[1, All, 2]];
SeedRandom[123];
pts = GeoPosition[Transpose[{RandomReal[ϕ, 5000], RandomReal[λ, 5000]}]];
ds = Dataset[<|"lat"->First@#, "lng"->Last@#|> &/@ First[pts]];

Set values for testing:

org = ds;
bounds = <|
  "lat" -> <|"min" -> minlat, "max" -> maxlat|>,
  "lng" -> <|"min" -> minlng, "max" -> maxlng|>|>;

We can improve performance of the filter (to about 40% of the original method) by combining the bounds tests, and by using flat variables.

Case #1 - original method

filtered1 = org[
  Select[
    #lat > bounds[["lat", "min"]] && 
    #lat < bounds[["lat", "max"]] &&
    #lng > bounds[["lng", "min"]] &&
    #lng < bounds[["lng", "max"]] &
  ]
]; // RepeatedTiming // First
(* 0.0199887 *)

Case #2 - Combine the bounds tests (faster)

filtered2 = org[
  Select[
    bounds["lat", "min"] < #lat < bounds["lat", "max"] &&
    bounds["lng", "min"] < #lng < bounds[["lng", "max"]] &
  ]
]; // RepeatedTiming // First
(* 0.0129947 *)

Case #3 - Use flat variables for bounds (fastest)

filtered3 = org[
  Select[
     minlat < #lat < maxlat && minlng < #lng < maxlng &
  ]
]; // RepeatedTiming // First
(* 0.00745205 *)

The filtered values are the same.

filtered1 === filtered2 === filtered3
(* True *)

With the sample dataset, we can test the replacement geoFilter (I'll try it when I get back to a desktop version of MMA).

---

Here's a drop-in replacement for geoFilter. It uses GeoWithinQ with a polygon computed from the latitude and longitude values from coords. Test this new method on your dataset to see if it's better.

geoFilter[org_, coords_List] := Module[{geopts, poly, filtered},
  geopts = Normal[GeoPosition[{#lat, #lng}] &/@ org];
  poly = Polygon[GeoPosition[
    Flatten[
      Table[{a, b}, {a, First@coords}, {b, Last@coords}], 1][[{1, 2, 4, 3}]]]];
  filtered = org[Position[GeoWithinQ[poly, geopts], True], All];
  If[Length@filtered >= 1, filtered,
    Throw["No Points in Selected Area"]]
]

The polygon coordinates are ordered beginning at the southwest corner, to the southeast, the northeast, and the northwest corners. GeoWithinQ finds the points that are inside poly. The positions of the selected points are used to extract the matching filtered values from the org dataset.