Question

我有一个带有（很多）横断面的数据集，其中土地高度（x是水平距离，z是海拔，以米为单位）和植被结构。 b_ml，b_kl1，b_kl2和b_s1是苔藓层，草本层1，草本层2和灌木层的覆盖率％和h_ ??是同一层的高度（苔藓层总是3厘米）。我想在有吸引力的图表中呈现这些数据。这是一个数据样本（1个横断面）：

#sample data
structure(list(X = c(432529.4846, 432530.4562, 432531.3492, 432532.3046, 
432533.3252, 432533.3419, 432534.3361, 432535.2709, 432536.3843, 
432537.4198, 432538.4336, 432539.3736, 432540.448, 432541.3967, 
432542.2748, 432543.0681, 432544.2508, 432545.2269, 432546.0911, 
432547.0195, 432548.0396, 432549.0209, 432549.9539, 432550.9391, 
432551.8999, 432552.8647, 432553.8258, 432554.7889, 432555.7722, 
432556.7435, 432557.7456, 432558.714, 432559.6824, 432560.6456, 
432561.6306, 432562.6059, 432563.5442, 432564.5688, 432565.5408, 
432566.5742), Z = c(10.6399, 10.5674, 10.4631, 10.3553, 10.2856, 
10.2392, 10.1234, 10.0792, 10.037, 9.9068, 9.834, 9.7207, 9.6024, 
9.5322, 9.3343, 9.2266, 9.0705, 8.9673, 8.9716, 8.8292, 8.7495, 
8.6541, 8.5429, 8.4306, 8.2933, 8.1877, 7.9679, 7.7799, 7.6801, 
7.5073, 7.3754, 7.2414, 7.0254, 7.0095, 6.8553, 6.7976, 6.7528, 
6.7151, 6.5225, 6.1952), b_ml = c(0.1, 0.05, NA, 0.1, 0.05, NA, 
NA, NA, 0.05, NA, 0.1, 0.2, 0.1, 0.05, NA, NA, NA, NA, NA, NA, 
NA, 0.01, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, 
NA, NA, NA, NA, NA), h_kl1 = c(75, 75, 70, 80, 80, 70, NA, 30, 
30, 50, 60, 60, 60, 60, 40, 30, 30, 70, 10, 10, 10, 20, 30, 30, 
30, 15, 15, 7, 5, 5, 5, 5, 3, 5, 5, 5, 5, 5, 3, 3), b_kl1 = c(0.9, 
0.95, 0.85, 0.95, 0.95, 0.9, NA, 0.4, 0.8, 0.9, 0.9, 0.9, 0.4, 
0.05, 0.6, 0.1, 0.05, 0.1, 0.1, 0.05, 0.1, 0.1, 0.98, 1, 1, 0.98, 
0.98, 0.9, 0.95, 0.95, 0.9, 0.8, 0.4, 0.95, 0.98, 0.95, 0.5, 
0.25, 0.05, 0.01), h_kl2 = c(110, 110, 110, 100, NA, 110, NA, 
110, 110, 100, 110, 120, 110, 110, 110, 70, 70, 120, 130, 120, 
110, 50, 80, NA, 90, NA, NA, NA, NA, 20, NA, NA, NA, NA, NA, 
45, 45, 25, NA, NA), b_kl2 = c(0.05, 0.05, 0.03, 0.02, NA, 0.1, 
NA, 0.05, 0.05, 0.05, 0.05, 0.1, 0.05, 0.1, 0.05, 0.95, 0.95, 
0.95, 0.95, 0.95, 0.95, 0.95, 0.01, NA, 0.01, NA, NA, NA, NA, 
0.05, NA, NA, NA, NA, NA, 0.1, 0.6, 0.1, NA, NA), h_s1 = c(NA, 
NA, NA, NA, NA, NA, NA, 300, 250, NA, NA, NA, 270, 270, 250, 
NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, 
NA, NA, NA, NA, NA, NA, NA, NA, NA), b_s1 = c(NA, NA, NA, NA, 
NA, NA, NA, 0.7, 0.6, NA, NA, NA, 0.75, 0.75, 0.75, NA, NA, NA, 
NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, 
NA, NA, NA, NA, NA, NA)), .Names = c("X", "Z", "b_ml", "h_kl1", 
"b_kl1", "h_kl2", "b_kl2", "h_s1", "b_s1"), class = "data.frame", row.names = 40:79)

我尝试使用geom_line和geom_pointrange绘制它们，使用密度作为alpha，但这看起来并不好看：

#set the terrein heigth
ld <- 2
p <- ggplot(df, aes(x=X,y=Z)) + geom_line()

#add the vegetation layers
p + geom_line(aes(x=X,y=Z+0.02, alpha = b_ml), size = ld, color = "darkgreen") + 
    geom_line(aes(x=X,y=Z+h_kl1/100, alpha = b_kl1), size = ld, color = "green") +  #divide by 100 because z is in meters and h in cm
    geom_line(aes(x=X,y=Z+h_kl2/100, alpha = b_kl2), size = ld, color = "green") +
    geom_line(aes(x=X,y=Z+h_s1/100, alpha = b_s1), size = ld, color = "brown") +
  theme_bw()

#add the vegetation layers with pointrange
p + geom_pointrange(aes(x = X, ymin = Z, ymax=Z+0.03, alpha = b_ml), color = "darkgreen") + 
  geom_pointrange(aes(x = X, ymin = Z+0.03, ymax=Z+h_kl1/100, alpha = b_kl1), color = "green") + 
  geom_pointrange(aes(x = X, ymin = Z+h_kl1/100, ymax=Z+h_kl2/100, alpha = b_kl2), color = "green") +
  geom_pointrange(aes(x = X, ymin = Z+h_kl2/100, ymax=Z+h_s1/100, alpha = b_s1), color = "brown") +
  theme_bw()

我想我想用一个代表层密度的点密度制作一个更好看的图。就像这张来自大自然的照片：

基本上，两层之间的区域用点填充，点数取决于密度/覆盖率但我有点坚持如何到达那里。我还想过'geom_point'和'geom_dotplot'，但这包括手工计算点数。

有关如何实现这一目标的任何建议？例如。哪个geom要使用（抖动？），或者如何重构我的数据？

Answer 1

我认为你没有实际的数据来创建一个情节作为你的例子，仅仅是因为他们似乎有一定数量的数据点（个人观察？某种扫描技术？），而你只有密度估计。从中创造点似乎充其量令人困惑，并且可能具有误导性。如果你有计算这些密度的原始数据，那么这可能是另一回事。以下是您可以制作的图表的一些示例：

首先，我以整齐的格式重新排列数据，以便更有意义：

library(dplyr)
library(tidyr)

densities <- df %>% 
  select(-h_kl1, -h_kl2, -h_s1) %>% 
  gather('type', 'density', b_ml, b_kl1, b_kl2, b_s1) %>% 
  mutate(type = substring(type, 3))
heights <- df %>% 
  select(-b_ml, -b_kl1, -b_kl2, -b_s1) %>% 
  gather('type', 'height', h_kl1, h_kl2, h_s1) %>% 
  mutate(type = substring(type, 3))
df2 <- left_join(densities, heights) %>% 
  mutate(height = ifelse(type == 'ml', 0.03, height / 100),
         type = factor(type, levels = c('s1', 'kl2', 'kl1', 'ml')))

尝试1：

ggplot(df2, aes(X, Z)) +
  geom_line() +
  geom_linerange(aes(ymin = Z, ymax = Z + height, alpha = density, col = type), size = 4) +
  scale_alpha_continuous(range = c(0, 0.7))

一个问题是ml几乎是不可见的，因为它与其他人的规模大不相同。我们可以试着躲闪：

ggplot(df2, aes(X, Z)) +
  geom_line() +
  geom_linerange(aes(ymin = Z, ymax = Z + height, alpha = density, col = type), 
                 size = 2, position = position_dodge(1))

这没什么用。也许是分面：

ggplot(df2, aes(X, Z)) +
  geom_line() +
  geom_linerange(aes(ymin = Z, ymax = Z + height, alpha = density, col = type), size = 4) +
  facet_wrap(~type)

最后一个选项：

ggplot(df2, aes(X, Z)) +
  geom_line() +
  geom_point(aes(y = Z + height, size = density, col = type), alpha = 0.6)

我认为这一点非常清楚地显示了植被的趋势。我喜欢这是同等重视不同的类别。

包括细分：

ggplot(df2, aes(X, Z)) +
  geom_line() +
  geom_segment(aes(xend = X, yend = Z + height), alpha = 0.1) +
  geom_point(aes(y = Z + height, size = density, col = type), alpha = 0.8)

Answer 2

我自己使用geom_jitter来接近解决方案。为了表明我想做什么（并且反对完全毫无根据地指责'召唤数据'和'将错误的想法置于观察者心中'[原文如此]），我将在这里发布我的尝试：

#make a tidy data.frame (taken from @Axeman)
library(dplyr)
library(tidyr)

densities <- df %>% 
  select(-h_kl1, -h_kl2, -h_s1) %>% 
  gather('type', 'density', b_ml, b_kl1, b_kl2, b_s1) %>% 
  mutate(type = substring(type, 3))
heights <- df %>% 
  select(-b_ml, -b_kl1, -b_kl2, -b_s1) %>% 
  gather('type', 'height', h_kl1, h_kl2, h_s1) %>% 
  mutate(type = substring(type, 3))
df2 <- left_join(densities, heights) %>% 
  mutate(height = ifelse(type == 'ml', 0.03, height / 100),
         type = factor(type, levels = c('s1', 'kl2', 'kl1', 'ml')))

# repeat rows according to density
# multiply density by 100: 1% will be 10 dots, 100%, 1000 dots and NA -> 1 (jitter can't handle NA)
df2$repli <- df2$density*1000
df2$repli[is.na(df2$repli)] <- 1

df3 <- df2[rep(rownames(df2), df2$repli), ]

require(ggplot2)

p <- ggplot(df3, aes(x=X,y=Z)) + geom_line(size = 1.5, colour = "gray") + theme_bw()

p + #geom_line(aes(y = Z + height, col = type)) +
  geom_jitter(aes(y = Z + height/2, col = type), position = position_jitter(width = 1, height = df3$height), 
                size = 0.0001, alpha = 0.1 ) +
  scale_color_manual(values = c("brown", "#CCFF00", "#33CC00","#666600"))

它仍然不完美，因为有些警告信息我不明白，“紧张”和一些较小的问题之间存在重叠。但它接近我最初的想法：植被层的视觉表示，获得高度和密度。

从审美角度看，Axemans的“最后一个选择”我认为更具吸引力。

绘制带点的植被样带

2 个答案: