merge - Connect points from two SpatialPointsDataFrames in R, considering the closest distance

I have two SpatialPointsDataFrames. In SpatialPointsDataFrame one (dat1) I have three points (A, B and C). In SpatialPointsDataFrame two (dat2), I have eight points with temperature measurements. I must connect each of those eight points to the closest point from dat1 and then calculate the average temperature for each point. How can I do that in R?

library(sp)
library(tmap)

# SpatialPointsDataFrame 1
dat_orig <- read.table(text="
point   lat   lng
A   15.6  80.9
B   20.9  10.9
C   1.2   81.8", header=TRUE, stringsAsFactors=FALSE)


dat1 <- dat_orig
coordinates(dat1) <- ~lng+lat

# SpatialPointsDataFrame 2
dat2_orig <- read.table(text = 
"temp   lat   lng
18   15.6  81.9
18.5   19.9  10.9
12.3   11.2   81.8
14.2   15.6  85.9

13.1   1.2  60.9
18.5   20.8  40.9
17   14.6  10.9
18.1   14.6  11.9", header=TRUE, stringsAsFactors=FALSE)

dat2 <- dat2_orig
coordinates(dat2) <- ~lng+lat

# tmap plot
tm_shape(dat1, ylim = c(-10, 35), xlim = c(-10,100)) + tm_dots(col = "red", size = 0.2) + tm_text("point") +

  tm_shape(dat2) + tm_dots(size = 0.1)

Answer

Use the FNN package:

> library(FNN)

this computes the first-nearest neighbours from dat1 for each of dat2:

> nn1 = get.knnx(coordinates(dat1), coordinates(dat2), 1)

The index into the three elements of dat1 for the 8 rows of dat2 is:

> ii = nn1$nn.index[,1]

> ii
[1] 1 2 1 1 3 2 2 2

(You can also get the distance out from nn1$nn.dist)

So we can then average the properties of dat2 grouped by that index with tapply:

> tapply(dat2$temp, list(ii), mean)
       1        2        3 
14.83333 18.02500 13.10000