我正在尝试使用geom_raster()或geom_tile()生成一个图，其中包含正值和负值，在负域和正域中具有非常不同的范围和数据分布。我是在一个以0为中心的发散尺度之后，但在正值和负值上都有良好的色阶。
我实际上可以通过生成两个图（下面的代码），一个正域和一个负面来生成我想要的输出，然后保存图并将它们覆盖在图像编辑程序中，尽管这种方法有它的自己的问题。下面的代码生成了这两个单独的图，以指示如果两个层可以堆叠，我希望看到的东西。不幸的是，当我尝试这个时，第二个图中的填充将替换第一个图中的填充。
我还发布了一个示例代码，它产生了一个建议的解决方案here，这显然与我希望在颜色渐变的分辨率方面实现的结果不一样。范围窄得多的范围（在这种情况下是负值）。
最后，抱歉大样本数据集（更大的df的一部分），但很难用较小的df来表达问题。
任何建议都非常感谢。
library(ggplot2)
library(scales)
df<-structure(list(x = c(1979, 1979, 1979, 1979, 1979, 1979, 1979,
1979, 1979, 1979, 1979, 1979, 1979, 1979, 1979, 1979, 1979, 1979,
1979, 1979, 1979, 1979, 1979, 1979, 1979, 1979, 1979, 1979, 1979,
1979, 1979, 1979, 1979, 1979, 1979, 1979, 1979, 1979, 1979, 1980,
1980, 1980, 1980, 1980, 1980, 1980, 1980, 1980, 1980, 1980, 1980,
1980, 1980, 1980, 1980, 1980, 1980, 1980, 1980, 1980, 1980, 1980,
1980, 1980, 1980, 1980, 1980, 1980, 1980, 1980, 1980, 1980, 1980,
1980, 1980, 1980, 1980, 1980, 1981, 1981, 1981, 1981, 1981, 1981,
1981, 1981, 1981, 1981, 1981, 1981, 1981, 1981, 1981, 1981, 1981,
1981, 1981, 1981, 1981, 1981, 1981, 1981, 1981, 1981, 1981, 1981,
1981, 1981, 1981, 1981, 1981, 1981, 1981, 1981, 1981, 1981, 1981
), y = c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16,
17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32,
33, 34, 35, 36, 37, 38, 39, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11,
12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27,
28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 1, 2, 3, 4, 5,
6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22,
23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38,
39), z = c(159.669, 179.469, 211.73, 217.466, 239.328, 353.454,
507.62, 656.972, 793.268, 1109.764, 1546.142, 1730.088, 1617.977,
1384.787, 1134.496, 906.544, 739.345, 614.798, 515.718, 428.918,
405.267, 466.252, 451.449, 346.907, 274.782, 220.826, 189.506,
139.73, 114.489, 101.309, 109.289, 114.041, 87.833, 65.061, 48.837,
45.621, 49.423, 198.88, 552.906, -219, -220, -221, -222, 88.223,
3303.874, 3861.382, 3540.399, 6716.934, 16477.35, 35683.23, 59065.06,
48825.8, 34849.34, 31240.76, 28330.3, 25030.27, 21862.9, 18641.04,
15372.67, 12525.73, 10198.47, 8496.241, 7267.41, 5876.377, 4155.166,
2666.032, 1786.76, 1246.569, 853.448, 624.179, 445.049, 353.794,
295.245, 262.17, 322.384, 415.967, 853.054, 1393.934, -177, -178,
-179, -180, -181, -182, -183, -184, -185, -186, -187, -188, -189,
-190, -191, -192, -193, -194, -195, 508.171, 4042.008, 6805.848,
9286.832, 13477.52, 16515.75, 17776.59, 17434.65, 15664.74, 13071.11,
10933.25, 10161.01, 11719.8, 18530.01, 33172.12, 59649.19, 62212.69,
51674, 39258.99, 32972.9)), .Names = c("x", "y", "z"), row.names = c(4384L,
4385L, 4386L, 4387L, 4388L, 4389L, 4390L, 4391L, 4392L, 4393L,
4394L, 4395L, 4396L, 4397L, 4398L, 4399L, 4400L, 4401L, 4402L,
4403L, 4404L, 4405L, 4406L, 4407L, 4408L, 4409L, 4410L, 4411L,
4412L, 4413L, 4414L, 4415L, 4416L, 4417L, 4418L, 4419L, 4420L,
4421L, 4422L, 4749L, 4750L, 4751L, 4752L, 4753L, 4754L, 4755L,
4756L, 4757L, 4758L, 4759L, 4760L, 4761L, 4762L, 4763L, 4764L,
4765L, 4766L, 4767L, 4768L, 4769L, 4770L, 4771L, 4772L, 4773L,
4774L, 4775L, 4776L, 4777L, 4778L, 4779L, 4780L, 4781L, 4782L,
4783L, 4784L, 4785L, 4786L, 4787L, 5115L, 5116L, 5117L, 5118L,
5119L, 5120L, 5121L, 5122L, 5123L, 5124L, 5125L, 5126L, 5127L,
5128L, 5129L, 5130L, 5131L, 5132L, 5133L, 5134L, 5135L, 5136L,
5137L, 5138L, 5139L, 5140L, 5141L, 5142L, 5143L, 5144L, 5145L,
5146L, 5147L, 5148L, 5149L, 5150L, 5151L, 5152L, 5153L), class = "data.frame")
#produce negative value plot and associated sensible scale and colour range
ggplot() +
geom_tile(data = subset(df, z <= 0), aes(x = x,y = y, fill = z)) +
scale_fill_gradient(low = "red", high = "white") +
scale_x_continuous(limits = c(1979, 1982)) +
scale_y_continuous(limits = c(0, 40))
#produce positive value plot and associated sensible scale and colour range
ggplot() +
geom_tile(data = subset(df, z > 0), aes(x = x, y = y, fill = z)) +
scale_fill_gradient(low = "white", high = "blue", trans = "log10") +
scale_x_continuous(limits = c(1979, 1982)) +
scale_y_continuous(limits = c(0, 40))
#one solution, but one that's inferior to what could be achieved overlaying the above plots.
ggplot(df, aes(x = x, y = y, fill = z)) + geom_tile() +
scale_fill_gradientn(colours = c("red", "white", "blue"),
values = rescale(c(min(df$z) ,0 , max(df$z))),
guide = "colorbar", limits = c(min(df$z), max(df$z))) +
scale_x_continuous(limits = c(1979, 1982)) + scale_y_continuous(limits = c(0, 40))