如何在0.05分辨率网格上聚合并绘制大型渔业作业数据集？

我正在处理一个包含坐标和渔业作业强度（HF）的数据集，共2985412行，样本数据如下：

Year  Month Type             SI_LONG  SI_LATI  HF
2008  04    OTB_DEF_>=70_0   -4.88    46.84    0.617
2008  04    OTB_DEF_>=70_0   -4.91    46.83    0.4  
2008  04    OTB_DEF_>=70_0   -4.96    46.85    1    
2008  04    OTB_DEF_>=70_0   -5.01    46.84    1.02 
2008  04    OTB_DEF_>=70_0   -5.01    46.92    1    
2008  04    OTB_DEF_>=70_0   -4.95    46.98    1    
2008  04    OTB_DEF_>=70_0   -4.92    46.96    1.02 
2008  04    OTB_DEF_>=70_0   -4.97    46.93    1.02 
2008  04    OTB_DEF_>=70_0   -4.96    46.97    0.167
2008  04    OTB_DEF_>=70_0   -5       46.94    0.833 

我需要将数据聚合到0.05分辨率的网格上，每个网格像素代表对应区域内所有点的HF均值，最终生成规整的可视化图。但目前用R代码生成的图存在像素大小不一、重叠的问题，代码如下：

# Map and grid limits
grid_xmin <- -6
grid_xmax <- 0
grid_ymin <- 42
grid_ymax <- 48
grid_limit <- extent(c(grid_xmin,grid_xmax,grid_ymin,grid_ymax))

# Resolution of grid
grid <- raster(grid_limit)
res(grid) <- 0.05
crs(grid) <- "+proj=longlat +datum=WGS84"
# Create grid polygon
gridpolygon <- rasterToPolygons(grid)
gridpolygon$layer <- c(1:length(gridpolygon$layer))
gridpolygon_sf <- st_as_sf(gridpolygon)

#Transform to sf object
points_sf <- st_as_sf(HF_filter[,c("SI_LONG","SI_LATI","HF","LE_MET_level6","Year","Month")],
                      coords = c("SI_LONG","SI_LATI"),crs="+proj=longlat +datum=WGS84")
points_sf <- points_sf[st_intersects(points_sf,gridpolygon_sf) %>% lengths > 0,]
points_df <- st_join(points_sf,gridpolygon_sf) %>% 
  as.data.frame %>%
  dplyr::select(-geometry)

## Aggregate at the cell level
polyg_sf <- inner_join(gridpolygon_sf,points_df)%>% 
  group_by(layer) %>% 
  dplyr::summarise(fishing_effort = mean(HF))

#Plot
ggplot() +
  geom_sf(data=polyg_sf,aes(fill = fishing_effort,col=fishing_effort))+
  #geom_sf(data=points_sf,size=0.1)+
  geom_sf(data=mapBase)+
  coord_sf(xlim = c(-6,0), ylim = c(43,48), expand = FALSE)+
  scale_fill_distiller(palette = "Spectral")+
  scale_color_distiller(palette = "Spectral")

问题根源 #

原方法通过rasterToPolygons生成网格多边形后进行空间连接聚合，不仅处理百万级数据效率低下，还可能因坐标系转换、多边形边缘计算误差，导致绘图时出现不规则重叠。更高效的方式是直接基于栅格填充值，或用sf生成规则网格后聚合。

方法一：栅格包直接聚合（高效处理大数据） #

直接将点数据的HF均值赋值到对应栅格单元，避免多边形转换的额外开销：

library(raster)
library(dplyr)
library(ggplot2)
library(sf)

# 1. 准备栅格模板
grid_xmin <- -6
grid_xmax <- 0
grid_ymin <- 42
grid_ymax <- 48
grid <- raster(extent(grid_xmin, grid_xmax, grid_ymin, grid_ymax), 
               res = 0.05, 
               crs = "+proj=longlat +datum=WGS84")

# 2. 将点数据聚合到栅格（计算每个单元的HF均值）
points_data <- HF_filter %>% select(SI_LONG, SI_LATI, HF)
raster_hf <- rasterize(points_data[,c("SI_LONG", "SI_LATI")], 
                       grid, 
                       field = points_data$HF, 
                       fun = mean)

# 3. 转换为sf多边形用于ggplot绘图
raster_sf <- rasterToPolygons(raster_hf, na.rm = TRUE) %>% st_as_sf()
names(raster_sf)[1] <- "fishing_effort"

# 4. 绘图
ggplot() +
  geom_sf(data = raster_sf, aes(fill = fishing_effort), color = NA) + # 去掉边框避免重叠
  geom_sf(data = mapBase) +
  coord_sf(xlim = c(-6, 0), ylim = c(43, 48), expand = FALSE) +
  scale_fill_distiller(palette = "Spectral", na.value = "transparent") +
  theme_minimal()

方法二：sf生成规则网格聚合（灵活适配后续分析） #

如果需要保留网格的sf对象用于后续操作，用st_make_grid生成规则网格，再进行空间连接聚合：

library(sf)
library(dplyr)
library(ggplot2)

# 1. 生成规则网格
grid_bbox <- st_bbox(c(xmin = -6, xmax = 0, ymin = 42, ymax = 48)) %>% st_as_sfc()
grid_sf <- st_make_grid(grid_bbox, cellsize = 0.05, what = "polygons") %>% st_as_sf()
grid_sf$cell_id <- 1:nrow(grid_sf)

# 2. 转换点数据为sf
points_sf <- st_as_sf(HF_filter, coords = c("SI_LONG", "SI_LATI"), 
                      crs = "+proj=longlat +datum=WGS84")

# 3. 空间连接并聚合（计算每个网格的HF均值）
polyg_sf <- st_join(grid_sf, points_sf, join = st_contains) %>%
  group_by(cell_id) %>%
  summarise(fishing_effort = mean(HF, na.rm = TRUE)) %>%
  filter(!is.na(fishing_effort)) # 过滤无数据的网格

# 4. 绘图
ggplot() +
  geom_sf(data = polyg_sf, aes(fill = fishing_effort), color = NA) +
  geom_sf(data = mapBase) +
  coord_sf(xlim = c(-6, 0), ylim = c(43, 48), expand = FALSE) +
  scale_fill_distiller(palette = "Spectral") +
  theme_minimal()

关键优化点 #

• 移除网格边框：绘图时设置color = NA，避免相邻网格边框重叠导致视觉不规则。
• 高效聚合逻辑：用rasterize或st_make_grid+st_join替代原方法的多次转换，减少计算误差和性能消耗。
• 过滤空值单元：移除无数据的网格，避免空值区域干扰可视化效果。

内容的提问来源于stack exchange，提问作者JulietteTC