如何绘制基于特定离散分类的geom_raster？代码调试求助

需求说明 #

希望绘制geom_raster，让填充色基于SPI的特定分段分类，尝试使用scale_fill_gradient的代码未达到预期效果。

用户尝试的代码片段 #

scale_fill_gradient(colours = c( "brown","burlywood","bisque","aliceblue","cadetblue2","blue"), 
                      breaks=c(-2,-1.5,-1,1,1.5,2), labels = format(c("-2","-1.5","-1","1","1.5","2")))

完整示例代码 #

year <- seq(1977,2021,1)
jan = runif(45, min=-4, max=4)
feb = runif(45, min=-4, max=4)
mar = runif(45, min=-4, max=4)
apr = runif(45, min=-4, max=4)
may = runif(45, min=-4, max=4)
jun = runif(45, min=-4, max=4)
jul = runif(45, min=-4, max=4)
aug = runif(45, min=-4, max=4)
sep = runif(45, min=-4, max=4)
oct = runif(45, min=-4, max=4)
nov = runif(45, min=-4, max=4)
dec = runif(45, min=-4, max=4)

df = data.frame(year,jan,feb,mar,apr,may,jun,jul,aug,sep,oct,nov,dec)
df <- reshape2::melt(df, id.vars = "year")

df$year <- factor(df$year, levels = (unique(df$year)))
df$variable <- factor(df$variable, levels = (unique(df$variable)))

library(ggplot2)
e1 <- ggplot(df, aes(x = variable, y = year, fill = value)) +
  geom_raster()+
  guides(fill=guide_legend(title="Bohicon"))+
  scale_fill_gradient(colours = c( "brown","burlywood","bisque","aliceblue","cadetblue2","blue"), 
                      breaks=c(-2,-1.5,-1,1,1.5,2), labels = format(c("-2","-1.5","-1","1","1.5","2")))+
  theme(legend.position="bottom")

e1

问题原因与解决方案 #

核心问题 #

scale_fill_gradient是连续渐变比例尺，但需求是分段式的分类颜色映射，且原代码的断点设置跳过了-1到1的区间，导致连续渐变无法匹配离散的分类逻辑，最终色阶不符合预期。

修正方案 #

提供两种实现方式，分别对应渐变过渡的分段色阶和严格离散的分类色块：

方式1：分段渐变（贴合参考图效果）

使用scale_fill_stepsn实现带过渡的分段色阶，匹配你的断点和颜色需求：

library(ggplot2)
library(reshape2)

# 生成数据
year <- seq(1977,2021,1)
months <- replicate(12, runif(45, min=-4, max=4))
df <- data.frame(year, months)
colnames(df)[-1] <- month.abb
df <- melt(df, id.vars = "year")

df$year <- factor(df$year, levels = unique(df$year))
df$variable <- factor(df$variable, levels = month.abb)

# 定义断点和对应颜色
breaks <- c(-Inf, -2, -1.5, -1, 1, 1.5, 2, Inf)
colors <- c("brown", "burlywood", "bisque", "aliceblue", "cadetblue2", "blue", "darkblue")

e1 <- ggplot(df, aes(x = variable, y = year, fill = value)) +
  geom_raster()+
  guides(fill=guide_colorbar(title="Bohicon", barwidth = 15))+
  scale_fill_stepsn(
    colors = colors,
    breaks = breaks,
    labels = c("≤-2", "-2~-1.5", "-1.5~-1", "-1~1", "1~1.5", "1.5~2", "≥2"),
    limits = c(-4, 4)
  )+
  theme(legend.position="bottom")

e1

方式2：离散分类色块

先将连续的value变量转成离散分类，再用scale_fill_manual指定颜色：

# 给value分箱生成分类变量
df$spi_cat <- cut(
  df$value,
  breaks = c(-Inf, -2, -1.5, -1, 1, 1.5, 2, Inf),
  labels = c("≤-2", "-2~-1.5", "-1.5~-1", "-1~1", "1~1.5", "1.5~2", "≥2")
)

e2 <- ggplot(df, aes(x = variable, y = year, fill = spi_cat)) +
  geom_raster()+
  guides(fill=guide_legend(title="Bohicon"))+
  scale_fill_manual(
    values = c("brown", "burlywood", "bisque", "aliceblue", "cadetblue2", "blue", "darkblue")
  )+
  theme(legend.position="bottom")

e2

关键说明 #

• scale_fill_stepsn适合需要渐变过渡的分段色阶，和参考图效果更匹配；
• 若需要无过渡的离散色块，优先选择先分箱再用scale_fill_manual的方式；
• 原代码中guide_legend搭配连续比例尺会冲突，渐变类图例应使用guide_colorbar。

内容的提问来源于stack exchange，提问作者Marcel