以前的帖子
Cleaning up a map using geom_tile
Get boundaries to come through on states
问题/疑问
我正在尝试平滑一些数据以使用ggplot2进行映射。感谢@MrFlick和@hrbrmstr,我取得了许多进步,但是在获得我需要列出的状态的“渐变”效果时遇到了问题。
这是一个示例,可让您大致了解我要查找的内容:
****这正是我要实现的目标。
http://nrelscience.org/2013/05/30/this-is-how-i-did-it-mapping-in-r-with-ggplot2/
(1)如何利用数据充分利用ggplot2?
(2)有没有更好的方法来达到渐变效果?
目标
我要从这个赏金中实现的目标是:
(1)内插数据以构建栅格对象,然后使用ggplot2进行绘制
(或者,如果可以对当前绘图执行更多操作,并且栅格对象不是一个好的策略)
(2)使用ggplot2构建更好的地图
当前结果
我一直在玩很多这些不同的情节,但由于两个原因,我仍然对结果不满意:(1)梯度的显示不尽如人意; (2)尽管我不确定该如何做,但可以改进演示文稿。
正如@hrbrmstr所指出的,如果我对数据进行插值以生成更多数据,然后将其拟合到栅格对象中并使用ggplot2进行绘制,则可能会提供更好的结果。我认为这是我现在应该追求的目标,但是鉴于我拥有的数据,我不确定如何做到这一点。
我在下面列出了到目前为止我对结果所做的代码。我真的很感激此事。谢谢。
数据集
这是两个数据集:
(1)完整数据集(175 mb):PRISM_1895_db_all.csv(不可用)
https://www.dropbox.com/s/uglvwufcr6e9oo6/PRISM_1895_db_all.csv?dl=0
(2)部分数据集(14 mb):PRISM_1895_db.csv(不可用)
https://www.dropbox.com/s/0evuvrlm49ab9up/PRISM_1895_db.csv?dl=0
***编辑:对于那些感兴趣的人,数据集不可用,但是我在我的网站上发布了一篇文章,将该代码与http://johnwoodill.com/pages/r-code.html处的加利福尼亚数据子集连接起来
剧情1
PRISM_1895_db <- read.csv("/.../PRISM_1895_db.csv")
regions<- c("north dakota","south dakota","nebraska","kansas","oklahoma","texas","minnesota","iowa","missouri","arkansas", "illinois", "indiana", "wisconsin")
ggplot() +
geom_polygon(data=subset(map_data("state"), region %in% regions), aes(x=long, y=lat, group=group)) +
geom_point(data = PRISM_1895_db, aes(x = longitude, y = latitude, color = APPT), alpha = .5, size = 5) +
geom_polygon(data=subset(map_data("state"), region %in% regions), aes(x=long, y=lat, group=group), color="white", fill=NA) +
coord_equal()
剧情2
PRISM_1895_db
regions<- c("north dakota","south dakota","nebraska","kansas","oklahoma","texas","minnesota","iowa","missouri","arkansas", "illinois", "indiana", "wisconsin")
ggplot() +
geom_polygon(data=subset(map_data("state"), region %in% regions), aes(x=long, y=lat, group=group)) +
geom_point(data = PRISM_1895_db, aes(x = longitude, y = latitude, color = APPT), alpha = .5, size = 5, shape = 15) +
geom_polygon(data=subset(map_data("state"), region %in% regions), aes(x=long, y=lat, group=group), color="white", fill=NA) +
coord_equal()
情节3
PRISM_1895_db <- read.csv("/.../PRISM_1895_db.csv")
regions<- c("north dakota","south dakota","nebraska","kansas","oklahoma","texas","minnesota","iowa","missouri","arkansas", "illinois", "indiana", "wisconsin")
ggplot() +
geom_polygon(data=subset(map_data("state"), region %in% regions), aes(x=long, y=lat, group=group)) +
stat_summary2d(data=PRISM_1895_db, aes(x = longitude, y = latitude, z = APPT)) +
geom_polygon(data=subset(map_data("state"), region %in% regions), aes(x=long, y=lat, group=group), color="white", fill=NA)
最佳答案
CRAN spatial view让我开始学习“克里金”。下面的代码需要大约7分钟才能在笔记本电脑上运行。您可以尝试使用更简单的插值(例如某种样条线)。您也可以从高密度区域中删除一些位置。您不需要所有这些点即可获得相同的热图。据我所知,用ggplot2
创建真正的渐变没有简单的方法(gridSVG
有几个选项,但没有像花哨的SVG编辑器中那样的“网格渐变”)。
根据要求,这是使用样条线进行插值的速度(快得多)。许多代码取自Plotting contours on an irregular grid。
克里金法则代码:
library(data.table)
library(ggplot2)
library(automap)
# Data munging
states=c("AR","IL","MO")
regions=c("arkansas","illinois","missouri")
PRISM_1895_db = as.data.frame(fread("./Downloads/PRISM_1895_db.csv"))
sub_data = PRISM_1895_db[PRISM_1895_db$state %in% states,c("latitude","longitude","APPT")]
coord_vars = c("latitude","longitude")
data_vars = setdiff(colnames(sub_data), coord_vars)
sp_points = SpatialPoints(sub_data[,coord_vars])
sp_df = SpatialPointsDataFrame(sp_points, sub_data[,data_vars,drop=FALSE])
# Create a fine grid
pixels_per_side = 200
bottom.left = apply(sp_points@coords,2,min)
top.right = apply(sp_points@coords,2,max)
margin = abs((top.right-bottom.left))/10
bottom.left = bottom.left-margin
top.right = top.right+margin
pixel.size = abs(top.right-bottom.left)/pixels_per_side
g = GridTopology(cellcentre.offset=bottom.left,
cellsize=pixel.size,
cells.dim=c(pixels_per_side,pixels_per_side))
# Clip the grid to the state regions
map_base_data = subset(map_data("state"), region %in% regions)
colnames(map_base_data)[match(c("long","lat"),colnames(map_base_data))] = c("longitude","latitude")
foo = function(x) {
state = unique(x$region)
print(state)
Polygons(list(Polygon(x[,c("latitude","longitude")])),ID=state)
}
state_pg = SpatialPolygons(dlply(map_base_data, .(region), foo))
grid_points = SpatialPoints(g)
in_points = !is.na(over(grid_points,state_pg))
fit_points = SpatialPoints(as.data.frame(grid_points)[in_points,])
# Do kriging
krig = autoKrige(APPT~1, sp_df, new_data=fit_points)
interp_data = as.data.frame(krig$krige_output)
colnames(interp_data) = c("latitude","longitude","APPT_pred","APPT_var","APPT_stdev")
# Set up map plot
map_base_aesthetics = aes(x=longitude, y=latitude, group=group)
map_base = geom_polygon(data=map_base_data, map_base_aesthetics)
borders = geom_polygon(data=map_base_data, map_base_aesthetics, color="black", fill=NA)
nbin=20
ggplot(data=interp_data, aes(x=longitude, y=latitude)) +
geom_tile(aes(fill=APPT_pred),color=NA) +
stat_contour(aes(z=APPT_pred), bins=nbin, color="#999999") +
scale_fill_gradient2(low="blue",mid="white",high="red", midpoint=mean(interp_data$APPT_pred)) +
borders +
coord_equal() +
geom_point(data=sub_data,color="black",size=0.3)
样条插补代码:
library(data.table)
library(ggplot2)
library(automap)
library(plyr)
library(akima)
# Data munging
sub_data = as.data.frame(fread("./Downloads/PRISM_1895_db_all.csv"))
coord_vars = c("latitude","longitude")
data_vars = setdiff(colnames(sub_data), coord_vars)
sp_points = SpatialPoints(sub_data[,coord_vars])
sp_df = SpatialPointsDataFrame(sp_points, sub_data[,data_vars,drop=FALSE])
# Clip the grid to the state regions
regions<- c("north dakota","south dakota","nebraska","kansas","oklahoma","texas",
"minnesota","iowa","missouri","arkansas", "illinois", "indiana", "wisconsin")
map_base_data = subset(map_data("state"), region %in% regions)
colnames(map_base_data)[match(c("long","lat"),colnames(map_base_data))] = c("longitude","latitude")
foo = function(x) {
state = unique(x$region)
print(state)
Polygons(list(Polygon(x[,c("latitude","longitude")])),ID=state)
}
state_pg = SpatialPolygons(dlply(map_base_data, .(region), foo))
# Set up map plot
map_base_aesthetics = aes(x=longitude, y=latitude, group=group)
map_base = geom_polygon(data=map_base_data, map_base_aesthetics)
borders = geom_polygon(data=map_base_data, map_base_aesthetics, color="black", fill=NA)
# Do spline interpolation with the akima package
fld = with(sub_data, interp(x = longitude, y = latitude, z = APPT, duplicate="median",
xo=seq(min(map_base_data$longitude), max(map_base_data$longitude), length = 100),
yo=seq(min(map_base_data$latitude), max(map_base_data$latitude), length = 100),
extrap=TRUE, linear=FALSE))
melt_x = rep(fld$x, times=length(fld$y))
melt_y = rep(fld$y, each=length(fld$x))
melt_z = as.vector(fld$z)
level_data = data.frame(longitude=melt_x, latitude=melt_y, APPT=melt_z)
interp_data = na.omit(level_data)
grid_points = SpatialPoints(interp_data[,2:1])
in_points = !is.na(over(grid_points,state_pg))
inside_points = interp_data[in_points, ]
ggplot(data=inside_points, aes(x=longitude, y=latitude)) +
geom_tile(aes(fill=APPT)) +
stat_contour(aes(z=APPT)) +
coord_equal() +
scale_fill_gradient2(low="blue",mid="white",high="red", midpoint=mean(inside_points$APPT)) +
borders
关于r - 平滑ggplot2 map ,我们在Stack Overflow上找到一个类似的问题:https://stackoverflow.com/questions/28059182/