模型单次运行结果¶

来自服务器试验收集的单次运行。

In [1]:

%load_ext autoreload
%autoreload 2
%matplotlib inline
%config InlineBackend.figure_format = 'retina'
from IPython.core.interactiveshell import InteractiveShell

InteractiveShell.ast_node_interactivity = "all"

from pathlib import Path
import pandas as pd
import numpy as np

%matplotlib inline
import matplotlib.pyplot as plt
import seaborn as sns

%load_ext autoreload %autoreload 2 %matplotlib inline %config InlineBackend.figure_format = 'retina' from IPython.core.interactiveshell import InteractiveShell InteractiveShell.ast_node_interactivity = "all" from pathlib import Path import pandas as pd import numpy as np %matplotlib inline import matplotlib.pyplot as plt import seaborn as sns

In [2]:

from src.ci import ResultAnalyzer
from hydra import compose, initialize
import os

# 加载项目层面的配置
with initialize(version_base=None, config_path="../../config"):
    cfg = compose(config_name="config")
os.chdir(cfg.root)


path = "outputs/2024-06-07/20-50-15/"
res = ResultAnalyzer(path)
res

from src.ci import ResultAnalyzer from hydra import compose, initialize import os # 加载项目层面的配置 with initialize(version_base=None, config_path="../../config"): cfg = compose(config_name="config") os.chdir(cfg.root) path = "outputs/2024-06-07/20-50-15/" res = ResultAnalyzer(path) res

Out[2]:

<src.ci.res_analysis.ResultAnalyzer at 0x320a622c0>

已经确认了 irr_area, quota 按省/年加总起来和输入数据是一致的。

In [3]:

data = res.get_data(level="province", water_unit="m3", agg_crops=True)
data.head()

data = res.get_data(level="province", water_unit="m3", agg_crops=True) data.head()

Out[3]:

	province	Time	irr_area	surface	ground	capital_yield	potential_yield	quota	net_irr	outdo	breach
0	Gansu	1981	346468.886321	7.693569e+08	1.471008e+09	1.556021e+08	1.562797e+08	2.234000e+09	7.104225e+08	0.0	False
1	Gansu	1982	353486.754695	1.201210e+09	1.091025e+09	1.448273e+08	1.482066e+08	2.256023e+09	1.039762e+09	0.0	False
2	Gansu	1983	357603.158990	1.425108e+09	8.800641e+08	1.308234e+08	1.313946e+08	2.300229e+09	4.477846e+08	0.0	False
3	Gansu	1984	362902.212791	1.408108e+09	9.331364e+08	1.285231e+08	1.290931e+08	2.317527e+09	3.554214e+08	0.0	False
4	Gansu	1985	343963.999287	1.218044e+09	1.016483e+09	1.240710e+08	1.245561e+08	2.174206e+09	4.695289e+08	0.0	False

In [4]:

after_data = data[data["Time"].isin(range(1987, 2011))].copy()
after_data["diff"] = after_data["surface"] - after_data["quota"]
after_data.head()

after_data = data[data["Time"].isin(range(1987, 2011))].copy() after_data["diff"] = after_data["surface"] - after_data["quota"] after_data.head()

Out[4]:

	province	Time	irr_area	surface	ground	capital_yield	potential_yield	quota	net_irr	outdo	breach	diff
6	Gansu	1987	324157.217516	1.257736e+09	8.459186e+08	1.313431e+08	1.321325e+08	1.085224e+09	4.835167e+08	1.725121e+08	True	1.725121e+08
7	Gansu	1988	323380.382180	1.219656e+09	9.734573e+08	1.399424e+08	1.413493e+08	1.068571e+09	4.060630e+08	1.510854e+08	True	1.510854e+08
8	Gansu	1989	335736.085515	1.341481e+09	9.070143e+08	1.494974e+08	1.501398e+08	1.086347e+09	4.833427e+08	2.551336e+08	True	2.551336e+08
9	Gansu	1990	332481.013807	1.090509e+09	1.197130e+09	1.478879e+08	1.490107e+08	1.078931e+09	5.005715e+08	1.157757e+07	True	1.157757e+07
10	Gansu	1991	343899.154883	1.349977e+09	1.027042e+09	1.686480e+08	1.704521e+08	1.067993e+09	6.022914e+08	2.819840e+08	True	2.819840e+08

In [5]:

fig, ax = plt.subplots(figsize=(4, 3))
ax = sns.stripplot(
    data=after_data,
    y="province",
    x="diff",
    dodge=True,
    alpha=0.25,
    zorder=1,
    legend=False,
    ax=ax,
    orient="h",
)
_, max_ = ax.get_xlim()
sns.despine(ax=ax, offset=5, trim=True)
ax.axvspan(0.0, max_, color="red", alpha=0.2)
plt.show();

fig, ax = plt.subplots(figsize=(4, 3)) ax = sns.stripplot( data=after_data, y="province", x="diff", dodge=True, alpha=0.25, zorder=1, legend=False, ax=ax, orient="h", ) _, max_ = ax.get_xlim() sns.despine(ax=ax, offset=5, trim=True) ax.axvspan(0.0, max_, color="red", alpha=0.2) plt.show();

In [6]:

farmers_data = res.get_data(level="village", water_unit="m3", agg_crops=False)

cmap = sns.cubehelix_palette(rot=-0.2, as_cmap=True)
g = sns.relplot(
    farmers_data,
    x="surface",
    y="quota",
    hue="crop",
    # col='crop',
    size="irr_area",
    sizes=(10, 200),
    alpha=0.8,
)
g.set(xscale="log", yscale="log")

farmers_data = res.get_data(level="village", water_unit="m3", agg_crops=False) cmap = sns.cubehelix_palette(rot=-0.2, as_cmap=True) g = sns.relplot( farmers_data, x="surface", y="quota", hue="crop", # col='crop', size="irr_area", sizes=(10, 200), alpha=0.8, ) g.set(xscale="log", yscale="log")

Out[6]:

<seaborn.axisgrid.FacetGrid at 0x177cca7a0>

In [7]:

sns.jointplot(
    farmers_data,
    x="surface",
    y="quota",
    hue="crop",
    # size='irr_area',
    # kind='kde'
)

sns.jointplot( farmers_data, x="surface", y="quota", hue="crop", # size='irr_area', # kind='kde' )

Out[7]:

<seaborn.axisgrid.JointGrid at 0x177af43a0>

超出水资源配额的用水基本都由水稻和小麦两种作物造成。

种植水稻的主体更倾向于违背水资源配额（$35\%$）

In [8]:

from matplotlib import pyplot as plt

_, ax = plt.subplots(figsize=(4, 3))
ax2 = ax.twinx()
ax = sns.lineplot(farmers_data, x="Time", y="breach", hue="crop", ax=ax)
ax.set_xlim(1987, 1997)

from matplotlib import pyplot as plt _, ax = plt.subplots(figsize=(4, 3)) ax2 = ax.twinx() ax = sns.lineplot(farmers_data, x="Time", y="breach", hue="crop", ax=ax) ax.set_xlim(1987, 1997)

Out[8]:

(1987.0, 1997.0)

在1998年前后，超出水资源配额制度的主体数量达到顶峰，在这之后逐步下降。

In [9]:

breach_year = farmers_data.loc[farmers_data["breach"], ["Time", "crop"]]
sns.displot(breach_year, kde=True)

breach_year = farmers_data.loc[farmers_data["breach"], ["Time", "crop"]] sns.displot(breach_year, kde=True)

Out[9]:

<seaborn.axisgrid.FacetGrid at 0x33b0c37c0>

灌溉效率¶

净灌溉量 net_irr 列，是由 AquaCrop 模型计算的作物需求的净灌溉量。

而 surface + ground 两列代表的是地表和地下水的总灌溉量，包括了灌溉途中的损失（运输和田间应用）。

所以我们可以评估不同地区农田水分利用系数的变化。

In [10]:

data["WU"] = data["surface"] + data["ground"]
data["efficiency"] = data["net_irr"] / data["WU"]

data["WU"] = data["surface"] + data["ground"] data["efficiency"] = data["net_irr"] / data["WU"]

In [11]:

sns.lineplot(
    data,
    x="Time",
    y="efficiency",
    # hue='province',
)

sns.lineplot( data, x="Time", y="efficiency", # hue='province', )

Out[11]:

<Axes: xlabel='Time', ylabel='efficiency'>

根据一张各省有效利用系数的2006年至2010年调查结果可以看到，这个农田水分利用效率系数还是很符合实际情况的：

In [12]:

data.pivot_table(index="Time", columns="province", values="efficiency")

data.pivot_table(index="Time", columns="province", values="efficiency")

Out[12]:

province	Gansu	Henan	Neimeng	Ningxia	Qinghai	Shaanxi	Shandong	Shanxi
Time
1981	0.317101	0.378489	0.144443	0.090189	0.116208	0.480455	0.643349	0.500403
1982	0.453602	0.357144	0.167440	0.113904	0.127955	0.564716	0.581372	0.598908
1983	0.194252	0.191684	0.171096	0.063398	0.067759	0.054824	0.562848	0.335834
1984	0.151809	0.160959	0.134339	0.057355	0.065868	0.068383	0.273500	0.227663
1985	0.210125	0.302077	0.142592	0.064554	0.046064	0.277030	0.488423	0.499501
1986	0.230327	0.375606	0.194123	0.069088	0.071852	0.365065	0.648532	0.630905
1987	0.229846	0.232363	0.200166	0.062539	0.050767	0.136028	0.390057	0.432704
1988	0.185154	0.288652	0.147949	0.055269	0.066928	0.334466	0.500149	0.449097
1989	0.214963	0.206589	0.184226	0.057626	0.056131	0.285017	0.529377	0.482779
1990	0.218816	0.118863	0.147302	0.053839	0.096193	0.317303	0.222897	0.415577
1991	0.253381	0.210634	0.147523	0.060184	0.118273	0.417826	0.240890	0.542815
1992	0.233781	0.346039	0.160050	0.050072	0.067094	0.401803	0.705646	0.675743
1993	0.261415	0.249420	0.161700	0.076573	0.075084	0.251024	0.450569	0.679599
1994	0.275737	0.268157	0.169313	0.061319	0.098391	0.602986	0.475359	0.523079
1995	0.382534	0.403511	0.196464	0.074671	0.098985	0.853839	0.450253	0.878708
1996	0.216703	0.233799	0.176899	0.062390	0.070600	0.263211	0.397500	0.308363
1997	0.415876	0.447553	0.174170	0.108770	0.068498	0.844457	0.741947	0.810111
1998	0.246646	0.117319	0.141015	0.075899	0.086318	0.279233	0.371215	0.299127
1999	0.207839	0.382264	0.222395	0.073118	0.098032	0.188212	0.535720	0.633229
2000	0.512614	0.554926	0.198109	0.116016	0.218233	0.550371	0.764036	0.738561
2001	0.437066	0.585482	0.236198	0.124254	0.155002	0.799161	0.910239	1.076134
2002	0.296331	0.402862	0.140096	0.095567	0.104067	0.610554	0.676216	0.483047
2003	0.328881	0.298391	0.160440	0.159231	0.081194	0.561968	0.534575	0.373131
2004	0.466659	0.332127	0.185518	0.146455	0.083318	0.843098	0.266426	0.462450
2005	0.409881	0.581141	0.291079	0.159733	0.079310	0.553654	0.652481	0.794051
2006	0.473087	0.423634	0.285925	0.132477	0.090203	0.552490	0.488319	0.590589
2007	0.445827	0.638833	0.273157	0.102106	0.094492	0.650283	0.677219	0.910394
2008	0.567308	0.374479	0.276586	0.163628	0.140244	0.582036	0.403661	0.725791
2009	0.511396	0.446279	0.350456	0.141233	0.100509	0.445963	0.684429	0.912813
2010	0.364736	0.536524	0.326634	0.115277	0.140426	0.436617	0.778261	0.779348

粮食产量¶

• capital_yield 代表的是单位面积的产量，
• potential_yield 是粮食生产潜力。

两者相除，反映的是水资源对粮食产量的限制。

In [13]:

data.head()

data.head()

Out[13]:

	province	Time	irr_area	surface	ground	capital_yield	potential_yield	quota	net_irr	outdo	breach	WU	efficiency
0	Gansu	1981	346468.886321	7.693569e+08	1.471008e+09	1.556021e+08	1.562797e+08	2.234000e+09	7.104225e+08	0.0	False	2.240364e+09	0.317101
1	Gansu	1982	353486.754695	1.201210e+09	1.091025e+09	1.448273e+08	1.482066e+08	2.256023e+09	1.039762e+09	0.0	False	2.292235e+09	0.453602
2	Gansu	1983	357603.158990	1.425108e+09	8.800641e+08	1.308234e+08	1.313946e+08	2.300229e+09	4.477846e+08	0.0	False	2.305172e+09	0.194252
3	Gansu	1984	362902.212791	1.408108e+09	9.331364e+08	1.285231e+08	1.290931e+08	2.317527e+09	3.554214e+08	0.0	False	2.341245e+09	0.151809
4	Gansu	1985	343963.999287	1.218044e+09	1.016483e+09	1.240710e+08	1.245561e+08	2.174206e+09	4.695289e+08	0.0	False	2.234526e+09	0.210125

加快推进灌区现代化改造 促进灌区高质量发展

保障国家粮食安全提出了新要求。尽管近年来我国粮食产量连续增长，但农产品“总量基本平衡、结构性紧缺”的状况依然存在。在农业种植结构不断调整、粮食生产效益持续走低的背景下，大中型灌区将承担更多的粮食和农产品生产任务，在保障国家粮食安全，优化农业区域布局，以及落实“藏粮于地、藏粮于技”战略中发挥更为重要的作用。

In [14]:

data["yield_ratio"] = data["capital_yield"] / data["potential_yield"]
sns.lineplot(
    data,
    x="Time",
    y="yield_ratio",
    # hue='province',
)

data["yield_ratio"] = data["capital_yield"] / data["potential_yield"] sns.lineplot( data, x="Time", y="yield_ratio", # hue='province', )

Out[14]:

<Axes: xlabel='Time', ylabel='yield_ratio'>

大型灌区节水改造让粮食增产

陈小江说，我国现有耕地１８．２６亿亩，其中有效灌溉面积８．６７亿亩，占耕地面积４７％，却生产着全国７５％的粮食和９０％以上的经济作物，且产量相对稳定。灌区平均亩产达５００多公斤，是全国粮食平均亩产的１．８倍，是旱田亩产的２倍至４倍。