statStr = keyLst[iS]
temp = list()
for k in range(len(statDictLst)):
data = statDictLst[k][statStr]
data = data[~np.isnan(data)]
temp.append(data)
dataBox.append(temp)
fig = plot.plotBoxFig(dataBox, keyLst, ['LSTM', 'DA-1', 'DA-7'], sharey=False)
fig.show()
# plot time series
t = utils.time.tRange2Array([20050101, 20150101])
fig, axes = plt.subplots(5, 1, figsize=(12, 8))
# iLst = [54, 219, 298, 325, 408]
for k in range(5):
iGrid = np.random.randint(0, 671)
# iGrid = iLst[k]
yPlot = list()
for y in yLst:
yPlot.append(y[iGrid, :])
if k == 0:
plot.plotTS(
t,
yPlot,
ax=axes[k],
cLst='kbrg',
markerLst='----',
legLst=['USGS', 'LSTM', 'DA-1', 'DA-7'])
else:
plot.plotTS(t, yPlot, ax=axes[k], cLst='kbrg', markerLst='----')
fig.show()
siteNoLst = dictData['siteNoLst']
varC = dictData['varC']
bb = True
while bb is True:
iS = randint(0, len(siteNoLst))
iC = randint(0, 20)
iS = 0
iC = 1
t = dfT[dfT['siteNo'] == siteNoLst[iS]]['date']
a = dfT[dfT['siteNo'] == siteNoLst[iS]][varC[iC]]
b = dfP[dfP['siteNo'] == siteNoLst[iS]][varC[iC]]
if not a.isna().all():
fig, ax = plt.subplots(1, 1)
fig, ax = plot.plotTS(t=t,
y=[a, b],
legLst=[varC[iC] + ' obs', varC[iC] + ' pred'])
fig.show()
bb = False
pass
# time series map
siteNoLst = dictData['siteNoLst']
varC = dictData['varC']
nP = len(siteNoLst)
nC = len(varC)
matRho1 = np.ndarray([nP, nC])
matRho2 = np.ndarray([nP, nC])
matRmse1 = np.ndarray([nP, nC])
matRmse2 = np.ndarray([nP, nC])
matN1 = np.ndarray([nP, nC])
ylabel = "Flow rate ($\mathregular{m^3}$/s)"
for k in range(len(gageindex)):
iGrid = gageindex[k]
yPlot = [obs[iGrid, :]]
for y in predLst[0:npred]:
yPlot.append(y[iGrid, :])
# get the NSE value of LSTM and DI(1) model
NSE_LSTM = str(round(statDictLst[0]["NSE"][iGrid], 2))
NSE_DI1 = str(round(statDictLst[1]["NSE"][iGrid], 2))
# plot time series
plot.plotTS(
t,
yPlot,
ax=axes[k],
cLst="kbrmg",
markerLst="---",
legLst=["USGS", "LSTM: " + NSE_LSTM, "DI(1): " + NSE_DI1],
title=subtitle[k],
linespec=["-", ":", ":"],
ylabel=ylabel,
)
# plot gage location
plot.plotlocmap(
plat,
plon,
ax=axes[-1],
baclat=gagelat,
baclon=gagelon,
title=subtitle[-1],
txtlabel=txt,
)
# plot
# training loss
fig, ax = plt.subplots(1, 1)
plt.plot(np.arange(nEpoch), lossEpLst)
ax.set_xlabel('epoch')
ax.set_ylabel('loss')
fig.show()
# forcing
fig, axes = plt.subplots(len(varFcLst) + 1, 1)
titleLst = ['discharge', 'precipitation', 'temperature']
cLst = 'gbr'
for i, title in enumerate(titleLst):
plot.plotTS(t,
x[:, i],
ax=axes[i],
cLst=cLst[i],
lsLst=['-'],
tBar=t[-1] - np.timedelta64(nTest, 'D'),
title=title)
if i + 1 < len(varWqLst): axes[i].set_xticks([])
fig.show()
# water quality
dfCode = usgs.readUsgsText(os.path.join(workDir, 'usgs_parameter_code'))
fig, axes = plt.subplots(len(varWqLst), 1)
for i, code in enumerate(varWqLst):
title = code + ' ' + \
dfCode['parameter_nm'].loc[dfCode['parameter_cd'] ==code].values[0]
plot.plotTS(t, [y[:, i], yOut[:, i]],
ax=axes[i],
cLst='kr',
lsLst=['-', '-'],
# pct = [10, 25, 50, 75, 90]
# indLst = list()
# diff = (statP['RMSE'] - statF['RMSE']) / statP['RMSE']
# for p in pct:
# indLst.append(abs(diff - np.percentile(diff, p)).argmin())
nts = len(indLst)
fig, axes = plt.subplots(nts, 1, figsize=[8, 6])
t = df.getT()
for k in range(nts):
ind = indLst[k]
ax = axes[k]
legLst = ['SMAP', 'project', 'forecast'] if k == 2 else None
plot.plotTS(
t, [obs[ind, :], yp[ind, :], yf[ind, :]],
ax=ax,
cLst='krb',
linewidth=1)
if k != nts - 1:
ax.set_xticklabels([])
plt.tight_layout()
plt.subplots_adjust(hspace=0)
plt.subplots_adjust(vspace=0)
fig.show()
fig.savefig(os.path.join(saveDir, 'ts_pixel.eps'))
fig.savefig(os.path.join(saveDir, 'ts_pixel'))
fig, ax = plt.subplots(1, 1, figsize=[8, 6])
plot.plotTS(
t, [obs[ind, :], yp[ind, :], yf[ind, :]],
ax=ax,
t = df.getT()
prcp = df.getDataTs('APCP_FORA').squeeze()
tBarLst = [20160401, [20160401, 20170401], 20170401]
for k in range(nts):
ind = indLst[k]
indY = indYLst[k]
ax = fig.add_subplot(gs[k * 3, 0])
tBar = utils.time.t2dt(tBarLst[indY])
if k == 0:
legLst1 = ['project', 'forecast', 'SMAP']
legLst2 = ['prcp']
else:
legLst1 = None
legLst2 = None
plot.plotTS(t, [ypLst[indY][ind, :], yfLst[indY][ind, :], obs[ind, :]],
ax=ax,
tBar=tBar,
legLst=legLst1,
linewidth=1)
ax.set_xticklabels([])
ax = fig.add_subplot(gs[k * 3 + 1, 0])
plot.plotTS(t, [prcp[ind, :]],
ax=ax,
cLst='c',
legLst=legLst2,
tBar=tBar,
linewidth=1)
fig.show()
fig.savefig(os.path.join(saveDir, 'ts_extreme.eps'))
fig.savefig(os.path.join(saveDir, 'ts_extreme'))
# plot time series
indLst = [1023]
strLst = ['east Texas']
tBar = [utils.time.t2dt(20160401)]
t = df.getT()
[lat, lon] = df.getGeo()
prcp = df.getDataTs('APCP_FORA')
tsNameLst = ['obs', 'prj', 'fore']
for k in range(len(indLst)):
fig, axes = plt.subplots(2, 1, figsize=[16, 6])
ind = indLst[k]
titleStr = 'typical {} pixel, lat {:.3}, lon {:.3}'.format(
strLst[k], lat[ind], lon[ind])
tsLst = [obs[ind, :], yp[ind, :], yf[ind, :]]
tsLst2 = [prcp[ind, :]]
plot.plotTS(t, tsLst, ax=axes[1], legLst=tsNameLst, cLst='krb', tBar=tBar)
plot.plotTS(t,
tsLst2,
ax=axes[0],
legLst=['prcp'],
title=titleStr,
cLst='c',
tBar=tBar)
axes[0].set_xticks([])
plt.tight_layout()
plt.subplots_adjust(wspace=0, hspace=0)
fig.show()
# maps for argriculture
trLst = [[20160801, 20161101], [20170801, 20171101]]
trLab = '1001-1201'
saveFile = os.path.join(workDir, 'data', 'dailyTS', siteNo)
usgs.downloadDaily(siteNo, nwisCode, saveFile)
saveFile = os.path.join(workDir, 'data', 'sample', siteNo)
usgs.downloadSample(siteNo, saveFile)
# read data
dfDaily = usgs.readUsgsText(os.path.join(workDir, 'data', 'dailyTS', siteNo),
dataType='dailyTS')
dfSample = usgs.readUsgsText(os.path.join(workDir, 'data', 'sample', siteNo),
dataType='sample')
# forcing data
# see extractForcing.py
# plot time series
dfCode = usgs.readUsgsText(os.path.join(workDir, 'usgs_parameter_code'))
fig, axes = plt.subplots(len(nwisCode)-1, 1)
for i, code in enumerate(nwisCode[1:]):
t1 = dfDaily['datetime'].values
y1 = dfDaily[code].values
t2 = dfSample['datetime'].values
y2 = dfSample[code].values
title = code + ' ' + \
dfCode['parameter_nm'].loc[dfCode['parameter_cd'] ==code].values[0]
plot.plotTS([t1, t2], [y1, y2],
ax=axes[i],
cLst='rb',
mLst=[None, '*'],
lsLst=['-', ':'],
title=title)
if i + 1 < len(nwisCode): axes[i].set_xticks([])
fig.show()
lon=crd[:, 1],
t=t,
colorMap=colorMap,
mapNameLst=mapNameLst,
tsNameLst=tsNameLst)
# plot time series
t = utils.time.tRange2Array([20100101, 20150101])
fig, axes = plt.subplots(5, 1, figsize=(12, 8))
iLst = [10, 52, 178, 404, 620]
for k in range(5):
iGrid = np.random.randint(0, 671)
iGrid = iLst[k]
yPlot = [yt2[iGrid, :], yp2[iGrid, :]]
gageId = camels.gageDict['id'][iGrid]
if k == 0:
plot.plotTS(t,
yPlot,
ax=axes[k],
cLst='br',
markerLst='--',
legLst=['USGS', 'LSTM'])
else:
plot.plotTS(t, yPlot, ax=axes[k], cLst='br', markerLst='--')
axes[k].set_ylabel('streamflow(cfs)')
if k == 4:
axes[k].set_xlabel('time')
plt.tight_layout()
fig.show()
