test/data-prepare/处理记录.md

In kongdd/PhenoAsync: Extract Remote Sensing Vegetation Phenology

1. GPP_NT与GPP_DT的差别

GPP_DT: (daytime partition) much better, no negative value GPP_NT: (nighttime partition), has negative value, 含有负值，因此本文主要根据GPP_DT进行

重大修改

growing season dividing, 调整生长季周期的划分策略。

https://en.wikipedia.org/wiki/Moderate_Resolution_Imaging_Spectroradiometer

1、Terra（1999年12月8日）早于Aqua（2002年5月4日）发射，获取的数据，时间序列更为完整； 2、Terra是上午星，由北向南经过赤道，上午十点半过境；Aqua是下午星，由南向北经过赤道，下午一点半过境；咱们国家位于北半球，从太阳角度考虑，Terra卫星获取的卫星数据相比Aqua的光谱效果要好。

## 2.2 MOD09A1
## 2.2.1 tidy MOD09A1
d_mod09a1 <- readRDS("data_test/flux212_MOD09A1_VI.RDS")
d_mod09a1 <- d_mod09a1[scale == "0m", .(site, t, date, year, EVI, EVI2, NDVI, LSWI, w, StateQA, QC_flag)]

d_mod09a1[QC_flag %in% c("cloud", "snow"), EVI := EVI2] # fix bright EVI
# (a) make sure values in a reasonable range
d_mod09a1[ EVI > 1 | EVI < -0.1, EVI := NA]
# (b) remove outliers: abs(y - mean) > 3sd
d_mod09a1[!is.na(EVI), `:=`(mean = mean(EVI), sd = sd(EVI)), .(site)]
d_mod09a1[abs(EVI - mean) >= 3*sd & QC_flag != "good", EVI := NA_real_, .(site)]
d_mod09a1[, c("mean", "sd") := NULL]

d_mod09a1 <- merge(d_mod09a1, st[, .(site, lat)])
d_mod09a1[, year2 := year + ((month(date)>=7)-1)*(lat<0)]

kongdd/PhenoAsync documentation built on April 21, 2024, 2:36 a.m.