如何在ggplot时间序列图中添加TP变量的倒置柱状图

我想优化现有的ggplot时间序列图，给TP（总降水量）变量添加倒置柱状图来提升可读性，但参考方案后效果不佳。期望最终效果用线条替代点，既能减少杂乱感，又能展示变量间的关联。以下是数据结构、示例数据和当前代码，求帮助：

数据结构 #

> str(DFSub0716)
'data.frame':   1633 obs. of  4 variables:
 $ datetime: POSIXct, format: "2016-07-06 12:00:00" "2016-07-06 13:00:00" ...
 $ Q_m3.s  : num  0.396 0.396 ...
 $ No3_mg.l: num  1.72 1.5 ...
 $ TP      : num  0.09661 0.06568 ...

示例数据 #

datetime Q_m3.s No3_mg.l         TP
6703 2016-07-06 06:00:00 0.4430 1.566312 0.01118792
6704 2016-07-06 07:00:00 0.4195 1.609813 0.02158611
6705 2016-07-06 08:00:00 0.3960 1.628437 0.05844045
6706 2016-07-06 09:00:00 0.3960 1.683905 0.04606792
6707 2016-07-06 10:00:00 0.3960 1.617095 0.07120785
6708 2016-07-06 11:00:00 0.3960 1.634220 0.08015819
6709 2016-07-06 12:00:00 0.3960 1.724347 0.09661102
6710 2016-07-06 13:00:00 0.3960 1.495940 0.06567970
6711 2016-07-06 14:00:00 0.3960 1.488123 0.13109615
6712 2016-07-06 15:00:00 0.3960 1.494128 0.03369540
6713 2016-07-06 16:00:00 0.3960 1.482408 0.00210596
6714 2016-07-06 17:00:00 0.3960 1.509152 0.00052649
6715 2016-07-06 18:00:00 0.3960 1.483495 0.00000000

当前代码 #

#### 2016/07//06
DFSub060716 <- DFSub[DFSub$datetime >= "2016-07-06 06:00:00" & DFSub$datetime <= "2016-07-06 18:00:00", ]

# to long format
library(reshape2)
DFSubL2 <- melt(DFSub060716, id="datetime")
# x-axis breaks
library(lubridate)
desired_breaks2 <- seq.POSIXt(from = lubridate::ymd_hms("20160706 060000"), 
                              to   = lubridate::ymd_hms("20160706 180000"), by = "2 hours")
# scale and plot data
library(ggplot2)
library(dplyr)
NQ060716 <- DFSubL2 %>%
  group_by(variable) %>%
  mutate(
    value.scaled = (value - mean(value)) / sd(value),
    idx = 1:n()) %>%
  ggplot(aes(x = datetime, y = value.scaled, colour = variable, linetype = variable)) +
  theme_bw() +
  theme(
    axis.title.x = element_blank(),
    panel.grid.major.x = element_blank())+
  scale_x_datetime(date_labels = "%m.%d\n%H:%M", breaks = desired_breaks2)+
  labs(title = "HND low-flow period 2016-07-06")+
  geom_line(lwd = 0.9)+
  scale_linetype_manual(values=c("solid", "solid", "solid")) +
  scale_color_manual(values = c("Q_m3.s" = "#000000",
                                "No3_mg.l" = "#669933",
                                "TP" = "#3399FF"))
NQ060716

核心思路是分开处理TP与其他变量：不将TP转为long格式，单独绘制倒置柱状图；其余变量保留标准化后的线条，既突出TP变化，又避免线条杂乱重叠。

优化后的代码如下：

#### 2016/07/06 数据筛选
DFSub060716 <- DFSub[DFSub$datetime >= "2016-07-06 06:00:00" & DFSub$datetime <= "2016-07-06 18:00:00", ]

# 加载所需包
library(lubridate)
library(ggplot2)
library(dplyr)
library(tidyr)

# 生成x轴刻度
desired_breaks2 <- seq.POSIXt(from = ymd_hms("20160706 060000"), 
                              to = ymd_hms("20160706 180000"), by = "2 hours")

# 对Q和No3进行标准化处理（TP单独处理）
scaled_df <- DFSub060716 %>%
  select(datetime, Q_m3.s, No3_mg.l) %>%
  pivot_longer(cols = -datetime, names_to = "variable", values_to = "value") %>%
  group_by(variable) %>%
  mutate(value.scaled = (value - mean(value)) / sd(value))

# 绘制图形
ggplot() +
  # 绘制Q和No3的标准化线条
  geom_line(data = scaled_df, 
            aes(x = datetime, y = value.scaled, color = variable, linetype = variable), 
            linewidth = 0.9) +
  # 绘制TP的倒置柱状图（取负值实现倒置）
  geom_col(data = DFSub060716, 
           aes(x = datetime, y = -TP, fill = "TP"), 
           width = 3600/1.2) + # width设为略小于小时间隔，避免柱子重叠
  # 主题设置
  theme_bw() +
  theme(
    axis.title.x = element_blank(),
    panel.grid.major.x = element_blank(),
    legend.title = element_blank() # 去掉图例标题
  ) +
  # 坐标轴设置
  scale_x_datetime(date_labels = "%m.%d\n%H:%M", breaks = desired_breaks2) +
  scale_y_continuous(name = "标准化值 / 倒置TP") + # 统一y轴标题
  # 颜色和线型设置
  scale_color_manual(values = c("Q_m3.s" = "#000000", "No3_mg.l" = "#669933")) +
  scale_fill_manual(values = c("TP" = "#3399FF")) +
  scale_linetype_manual(values = c("Q_m3.s" = "solid", "No3_mg.l" = "solid")) +
  # 标题
  labs(title = "HND低流量时段 2016-07-06")

关键说明 #

• TP倒置实现：通过y = -TP让柱状图向下延伸，调整width参数适配小时级时间间隔，避免柱子重叠。
• 变量分离处理：Q和No3保持标准化线条，TP用柱状图，既保留变量间的关联对比，又避免原long格式下的线条杂乱问题。
• 视觉一致性：沿用你原有的颜色和线型设置，同时优化图例，让整体更简洁。

内容的提问来源于stack exchange，提问作者JoSche