如何在R中按最近日期填充NA值（优先locf处理等距情况）

解决方案 #

通过计算每个NA值到前后最近非NA值的日期距离，动态选择符合规则的填充值，替代固定的两次locf操作。以下是调整后的代码：

调整后代码 #

library(tidyverse)
library(zoo)

read.csv("path/to/merged_data.csv",
         colClasses = c("Date", "numeric", "numeric", "numeric", "character")) %>%
  group_by(field_id) %>%
  arrange(date) %>%
  # 处理NDVI_l8列的填充
  mutate(
    # 生成正向、反向填充的基础值
    l8_locf = na.locf(NDVI_l8, na.rm = FALSE),
    l8_locb = na.locf(NDVI_l8, fromLast = TRUE, na.rm = FALSE),
    # 获取前后最近非NA值的日期
    l8_prev_date = na.locf(date[!is.na(NDVI_l8)], na.rm = FALSE),
    l8_next_date = na.locf(date[!is.na(NDVI_l8)], fromLast = TRUE, na.rm = FALSE),
    # 计算到前后非NA值的日期差
    dist_prev_l8 = date - l8_prev_date,
    dist_next_l8 = l8_next_date - date,
    # 按规则选择填充值：距离相等时优先向前填充
    NDVI_l8_filled = case_when(
      !is.na(NDVI_l8) ~ NDVI_l8,
      dist_prev_l8 <= dist_next_l8 ~ l8_locf,
      TRUE ~ l8_locb
    )
  ) %>%
  # 处理NDVI_s2列，逻辑与NDVI_l8一致
  mutate(
    s2_locf = na.locf(NDVI_s2, na.rm = FALSE),
    s2_locb = na.locf(NDVI_s2, fromLast = TRUE, na.rm = FALSE),
    s2_prev_date = na.locf(date[!is.na(NDVI_s2)], na.rm = FALSE),
    s2_next_date = na.locf(date[!is.na(NDVI_s2)], fromLast = TRUE, na.rm = FALSE),
    dist_prev_s2 = date - s2_prev_date,
    dist_next_s2 = s2_next_date - date,
    NDVI_s2_filled = case_when(
      !is.na(NDVI_s2) ~ NDVI_s2,
      dist_prev_s2 <= dist_next_s2 ~ s2_locf,
      TRUE ~ s2_locb
    )
  ) %>%
  # 保留NDVI_l7非NA的行，清理中间变量
  filter(!is.na(NDVI_l7)) %>%
  select(-starts_with("l8_"), -starts_with("s2_"), -NDVI_l8, -NDVI_s2) %>%
  relocate(field_id, .after = last_col()) %>%
  write_csv(file.path(results, "merged_data_interpolated.csv"))

代码说明 #

• 先对NDVI_l8和NDVI_s2分别生成正向、反向填充值，同时计算每个NA值到前后最近非NA值的日期差
• 通过case_when判断：若当前值非NA则保留原数值；若到前一个非NA值的距离小于等于后一个，使用正向填充值，否则使用反向填充值
• 保留了原有代码的分组、排序、过滤逻辑，风格统一且规则实现更精准

内容的提问来源于stack exchange，提问作者Barbara Perez de Araújo