personalizing.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Personalizing Sustainable Agriculture with Causal Machine Learning\n",
    "\n",
    "### *Tackling Climate Change with Machine Learning Workshop @ NeurIPS'22* | **Best Proposal Paper**\n",
    "\n",
    "#### Authors: Georgios Giannarakis, Vasileios Sitokonstantinou, Roxanne Suzette Lorilla, Charalampos Kontoes | National Observatory of Athens, Greece"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This jupyter notebook contains the code required to implement and reproduce the \"Personalizing Sustainable Agriculture with Causal Machine Learning\" paper.\n",
    "\n",
    "**NB**: Due to privacy reasons, some data cannot be open-sourced. This includes crop declaration data (found in ```IACS``` dataframes in the script) and soil organic carbon content values (found in the ```SOC``` dataframe). For retrieving crop declaration data via an Earth Observation (EO) based service such as [EuroCrops](https://www.eurocrops.tum.de/). To derive parcel-level SOC content values, we again refer the interested reader to the EO literature (https://doi.org/10.1016/j.rse.2020.111793), which is what was also used in the paper."
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Abbreviations\n",
    "\n",
    "- SOC = Soil Organic Carbon\n",
    "- IACS = Integrated Administration and Control System\n",
    "- LPIS = Land Parcel Identification System\n",
    "- ECO = LPIS with an indicator for fields practicing eco-friendly practices that year\n",
    "- ERA5 = The fifth generation ECMWF reanalysis for the global climate and weather\n",
    "\n",
    "Note that IACS and LPIS are used interchangeably. They simply refer to the crop declarations (aka crop type map) of a single year. These are not inferred via remote sensing: it is simply what the farmer declared (the ground truth)."
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### High-level flow:\n",
    "\n",
    "1. Create dataframe for 2021 with SOC, IACS, ECO info\n",
    "2. Create dataframe for 2020 with IACS, ECO info\n",
    "3. Match geometries and filter for parcels whose geometry remained constant\n",
    "4. Extend matched dataframe with ERA5 info\n",
    "5. Causal analysis with EconML\n",
    "6. Visualizing the results"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Imports"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Refer to ```requirements.txt``` file."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "from pandas import read_excel\n",
    "import numpy as np\n",
    "import os\n",
    "import geopandas as gpd\n",
    "import shapely\n",
    "import netCDF4\n",
    "import xarray as xr \n",
    "import rioxarray as rio\n",
    "import matplotlib\n",
    "import glob\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.linear_model import LassoCV, LogisticRegression, LinearRegression, Lasso\n",
    "from sklearn.ensemble import RandomForestClassifier, RandomForestRegressor, GradientBoostingClassifier, GradientBoostingRegressor\n",
    "from econml.dml import SparseLinearDML, LinearDML, CausalForestDML"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1. Data from 2021"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Read IACS/SOC 2021\n",
    "file_dir = '/nfs/data2/EIFFEL/'\n",
    "soc21 = gpd.read_file(os.path.join(file_dir, 'Lithuania_SOC_parcel/lithuania_soc_fields.shp')).set_crs('EPSG:3346').to_crs('EPSG:3857')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Read English crop codes\n",
    "file_dir = '/nfs/data2/EIFFEL/'\n",
    "my_sheet = 'Sheet1'\n",
    "file_name = 'paseliu2020_kodai_ENG.xls'\n",
    "crop_codes = pd.read_excel(os.path.join(file_dir, file_name), sheet_name = my_sheet)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Augment SOC21 with English info\n",
    "soc21 = soc21.rename(columns={'PSL_KODAS': 'code'}).merge(crop_codes, on='code', how='left')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Read ECO 2021\n",
    "file_dir = '/nfs/data2/EIFFEL/'\n",
    "eco = gpd.read_file(os.path.join(file_dir, 'eco_shp/eco_21.shp')).set_crs('EPSG:3346').to_crs('EPSG:3857')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create id and drop duplicates on ECO21 fields to ensure 1-1 outcome when merging on id\n",
    "eco[\"id\"] = eco[\"BLK_NR\"] + \"-\" + eco[\"LAUKO_NR\"] + \"-\" + eco[\"PSL_KODAS\"]\n",
    "eco[\"id\"] = eco.id.astype(str)\n",
    "eco = eco.drop_duplicates(subset=['id'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create id and drop duplicates on SOC21 to ensure 1-1 outcome when merging on id\n",
    "soc21[\"id\"] = soc21[\"BLK_NR\"] + \"-\" + soc21[\"LAUKO_NR\"] + \"-\" + soc21[\"code\"]\n",
    "soc21 = soc21.drop_duplicates(subset=['id'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# left join SOC21 with ECO21 on id and drop irrelevant columns\n",
    "soc_eco21 = soc21.merge(eco[[\"Programa\", \"RGR\", \"id\"]], on='id', how='left')\n",
    "soc_eco21 = soc_eco21.drop(columns=[\"BLK_NR\", \"LAUKO_NR\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create dummies for the 8 crop groups found in the dataset\n",
    "soc_eco21 = pd.get_dummies(soc_eco21, columns=['group'], prefix='', prefix_sep='')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create is_eco column (binary)\n",
    "soc_eco21['Programa'] = soc_eco21['Programa'].fillna(0)\n",
    "soc_eco21['is_eco'] = np.where(soc_eco21.Programa == 0, 0, 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2. Data from 2020"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Read IACS20 (henceforth SOC20 for symmetry)\n",
    "os.chdir('/nfs/data2/EIFFEL/IACS/inc_387887_2020')\n",
    "soc20 = gpd.read_file('inc_387887_2020.shp').set_crs('EPSG:3346').to_crs('EPSG:3857')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Augment SOC20 with English info\n",
    "soc20 = soc20.rename(columns={'PSL_KODAS': 'code'}).merge(crop_codes, on='code', how='left')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Read ECO20\n",
    "file_dir = '/nfs/data2/EIFFEL/'\n",
    "eco20 = gpd.read_file(os.path.join(file_dir, 'eco_shp/eco_20.shp')).set_crs('EPSG:3346').to_crs('EPSG:3857')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create id and drop duplicates on ECO20 fields to ensure 1-1 outcome when merging on id\n",
    "eco20[\"id\"] = eco20[\"BLK_NR\"] + \"-\" + eco20[\"LAUKO_NR\"] + \"-\" + eco20[\"PSL_KODAS\"]\n",
    "eco20[\"id\"] = eco20.id.astype(str)\n",
    "eco20 = eco20.drop_duplicates(subset=['id'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create id and drop duplicates on SOC20 fields to ensure 1-1 outcome when merging on id\n",
    "soc20[\"id\"] = soc20[\"BLK_NR\"] + \"-\" + soc20[\"LAUKO_NR\"] + \"-\" + soc20[\"code\"]\n",
    "soc20 = soc20.drop_duplicates(subset=['id'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "# left join SOC20 with ECO20 on id and drop irrelevant columns\n",
    "soc_eco20 = soc20.merge(eco20[[\"Programa\", \"RGR\", \"id\"]], on='id', how='left')\n",
    "soc_eco20 = soc_eco20.drop(columns=[\"BLK_NR\", \"LAUKO_NR\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create dummies for the 8 crop groups found in the dataset\n",
    "soc_eco20 = pd.get_dummies(soc_eco20, columns=['group'], prefix='', prefix_sep='')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create is_eco column (binary)\n",
    "soc_eco20['Programa'] = soc_eco20['Programa'].fillna(0)\n",
    "soc_eco20['is_eco'] = np.where(soc_eco20.Programa == 0, 0, 1)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3. Match 2021-2020 geometries to find parcels constant in space and time"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Matching is done as follows: First, initialize a spatial index on the relevant geometries. Then, apply the ```geopandas.sindex.SpatialIndex.nearest``` to the index with the second geometry as argument. Finally, calculate the centroid distance between each 2021 parcel and its 2020 matched counterpart, and only keep the parcels whose centroids were \"close enough\" (based on an arbitrary threshold). These are the \"constant\" (over the period of reference) parcels."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "s21 = soc_eco21.geometry.sindex # initialize spatial index\n",
    "s20 = soc_eco20.geometry.sindex # initialize spatial index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# see https://geopandas.org/en/stable/docs/reference/api/geopandas.sindex.SpatialIndex.nearest.html\n",
    "matched = s20.nearest(soc_eco21.geometry, max_distance=0.5, return_all = False) # quite fast"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "# extract the centroids for second-level distance-based matching\n",
    "a = soc_eco21.iloc[matched[0]].geometry.centroid\n",
    "b = soc_eco20.iloc[matched[1]].geometry.centroid"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# calculate 1-1 distance between elements of a and b\n",
    "dis_array = a.distance(b, align=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "# select matched 2021 rows and augment with distance from their 2020 match to filter later\n",
    "soc_eco21 = soc_eco21.iloc[matched[0]].copy()\n",
    "soc_eco21['distance'] = dis_array"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "# select matched 2020 rows and drop obsolete 2020 geometry\n",
    "soc_eco20 = soc_eco20.iloc[matched[1]].copy()\n",
    "soc_eco20 = soc_eco20.drop(columns=['geometry'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "# horizontal concatenation of 2021 and 2020 data\n",
    "final = pd.concat([soc_eco21.reset_index(drop=True), soc_eco20.reset_index(drop=True)], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "# only keep 2021 geometries whose distance from their 2020 match is \"small\"\n",
    "final = final[final['distance'] < 3]"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4. Augment data with ERA5 climate indices\n",
    "\n",
    "ERA5 data are open-source. See ```cds-api-data-retrieval.ipynb``` for the API call used to retrieve them in ```xarray``` format."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "os.chdir('/home/noa2/EIFFEL-git/data/raw')\n",
    "era = xr.open_dataset('era').rio.write_crs(\"EPSG:4326\").rio.reproject(\"EPSG:3857\") # ERA5 uses 4326"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "# select Jan 2020 - Dec 2021 data, and average over the months to get one value per ERA5 cell\n",
    "era = era.sel(time=slice('2020-01-01T00:00:00.000000000', '2021-12-01T00:00:00.000000000'))\n",
    "era = era.mean(dim='time')\n",
    "era = era.to_dataframe().reset_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create dataframe from centroids of each ERA5 cell (hence point geometries)\n",
    "era = gpd.GeoDataFrame(era, geometry=gpd.points_from_xy(era.x, era.y)).set_crs(\"EPSG:3857\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "# to create an ERA5 square grid, calculate the distance between two neighboring ERA5 centroids and create a square buffer\n",
    "a1 = era.geometry[1]\n",
    "a2 = era.geometry[2]\n",
    "buf = a1.distance(a2) / 2\n",
    "era.geometry = era.geometry.buffer(buf, cap_style = 3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "The `op` parameter is deprecated and will be removed in a future release. Please use the `predicate` parameter instead.\n"
     ]
    }
   ],
   "source": [
    "# Left spatial join final with era, matching a final polygon with an ERA5 cell if the former is covered by the latter\n",
    "# covered by in the following sense: https://shapely.readthedocs.io/en/stable/manual.html\n",
    "final_era = gpd.sjoin(final, era, how=\"left\", op=\"covered_by\") # quite fast"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['code', 'lithuania_', 'geometry', 'name', 'group_i', 'id', 'Programa',\n",
       "       'RGR', 'I group', 'II group', 'III group', 'IV group', 'V group',\n",
       "       'VI group', 'VII group', 'VIII group', 'is_eco', 'distance', 'code',\n",
       "       'name', 'group_i', 'id', 'Programa', 'RGR', 'I group', 'II group',\n",
       "       'III group', 'IV group', 'V group', 'VI group', 'VII group',\n",
       "       'VIII group', 'is_eco', 'index_right', 'x', 'y', 'spatial_ref', 'u10',\n",
       "       'v10', 'd2m', 't2m', 'evabs', 'evaow', 'evatc', 'evavt', 'fal',\n",
       "       'lai_hv', 'lai_lv', 'pev', 'ro', 'src', 'skt', 'asn', 'snowc', 'rsn',\n",
       "       'sde', 'sd', 'es', 'sf', 'smlt', 'stl1', 'stl2', 'stl3', 'stl4', 'ssro',\n",
       "       'slhf', 'ssr', 'str', 'sp', 'sro', 'sshf', 'ssrd', 'strd', 'tsn', 'e',\n",
       "       'tp', 'swvl1', 'swvl2', 'swvl3', 'swvl4'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# the final, matched 2020-21 dataset augmented with ERA5\n",
    "final_era.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "# smart deduplication of column names that would cause downstream problems\n",
    "final_era.columns = pd.io.parsers.base_parser.ParserBase({'names':final_era.columns, 'usecols':None})._maybe_dedup_names(final_era.columns)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "# define double_eco treatment, binary indicator for eco-friendly parcels in both years\n",
    "final_era['double_eco'] = np.where((final_era['is_eco'] == 1)  & (final_era['is_eco.1'] == 1), 1, 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "# drop obsolete info and na values\n",
    "final_era = final_era.drop(columns=['RGR', 'RGR.1']).dropna()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 5. Causal analysis with EconML"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "# standard scaler for all ERA5 data\n",
    "scaler = StandardScaler()\n",
    "final_era.iloc[:, 36:78] = scaler.fit_transform(final_era.iloc[:, 36:78])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create \"crop sequence\" variable, concatenating the 2020 and 2021 crop code for each parcel\n",
    "final_era['crop_seq'] = final_era['code'] + '-' + final_era['code.1']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DGP-DGP    40818\n",
       "GPŽ-GPŽ    19561\n",
       "KVŽ-KVŽ    11242\n",
       "KVŽ-RAŽ     7987\n",
       "KTŽ-KTŽ     7381\n",
       "Name: crop_seq, dtype: int64"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# the most prominent crop sequences found in the dataset\n",
    "final_era.crop_seq.value_counts()[:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "# the plan is to account for heterogeneity caused by different crops\n",
    "# create dummies for the most prominent crop sequences only (those seen > 10000 times)\n",
    "cats = final_era.crop_seq.value_counts()[lambda x: x > 10000].index\n",
    "temp = pd.get_dummies(pd.Categorical(final_era['crop_seq'], categories=cats))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "# horizontal concatenation of the dummies to the final dataset\n",
    "final_era = pd.concat([final_era, temp], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "final_era = final_era.dropna()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "# in these preliminary results, we restrict the dataset to the three most prominent crop combinations, and learn CATEs for them\n",
    "final_era_f = final_era.loc[~((final_era['DGP-DGP'] == 0) & (final_era['GPŽ-GPŽ'] == 0) & (final_era['KVŽ-KVŽ'] == 0)), :]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Y = outcome, T = treatment, W = controls, X = heterogeneity controls\n",
    "Y = final_era_f.lithuania_\n",
    "T = final_era_f.double_eco.astype(int) # only 7k double eco\n",
    "W = final_era_f.iloc[:, 36:78] # ERA5, if need crop groups on W can use e.g. np.r_[3:46, 51:59]\n",
    "X = final_era_f.iloc[:, -3:] # the type of rotation, with binary variables for the most prominent rotations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "# further restrict ERA5 controls to the most significant climate variables\n",
    "W = final_era_f.loc[:, ['u10', 'v10', 't2m', 'evavt', 'lai_lv', 'ro', 'stl1', 'tp', 'swvl1']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<econml.dml.dml.LinearDML at 0x7f3468df2950>"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# train a LinearDML, do not fit cate intercept here for simpler interpretation of point estimates\n",
    "est0 = LinearDML(model_y=RandomForestRegressor(random_state=42), model_t=RandomForestClassifier(random_state=42), random_state=123, fit_cate_intercept = False)\n",
    "est0.fit(Y, T, X=X, W=W)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CATE Intercept Results:  No intercept was fitted!\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>Coefficient Results</caption>\n",
       "<tr>\n",
       "   <td></td>  <th>point_estimate</th> <th>stderr</th>  <th>zstat</th> <th>pvalue</th> <th>ci_lower</th> <th>ci_upper</th>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>X0</th>      <td>0.058</td>      <td>0.141</td>  <td>0.407</td>  <td>0.684</td>   <td>-0.22</td>    <td>0.335</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>X1</th>     <td>-0.077</td>      <td>0.174</td> <td>-0.443</td>  <td>0.658</td>  <td>-0.418</td>    <td>0.264</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>X2</th>     <td>-0.088</td>      <td>0.275</td> <td>-0.319</td>  <td>0.749</td>  <td>-0.628</td>    <td>0.452</td> \n",
       "</tr>\n",
       "</table><br/><br/><sub>A linear parametric conditional average treatment effect (CATE) model was fitted:<br/>$Y = \\Theta(X)\\cdot T + g(X, W) + \\epsilon$<br/>where for every outcome $i$ and treatment $j$ the CATE $\\Theta_{ij}(X)$ has the form:<br/>$\\Theta_{ij}(X) = \\phi(X)' coef_{ij} + cate\\_intercept_{ij}$<br/>where $\\phi(X)$ is the output of the `featurizer` or $X$ if `featurizer`=None. Coefficient Results table portrays the $coef_{ij}$ parameter vector for each outcome $i$ and treatment $j$. Intercept Results table portrays the $cate\\_intercept_{ij}$ parameter.</sub>"
      ],
      "text/plain": [
       "<class 'econml.utilities.Summary'>\n",
       "\"\"\"\n",
       "                  Coefficient Results                   \n",
       "========================================================\n",
       "   point_estimate stderr zstat  pvalue ci_lower ci_upper\n",
       "--------------------------------------------------------\n",
       "X0          0.058  0.141  0.407  0.684    -0.22    0.335\n",
       "X1         -0.077  0.174 -0.443  0.658   -0.418    0.264\n",
       "X2         -0.088  0.275 -0.319  0.749   -0.628    0.452\n",
       "--------------------------------------------------------\n",
       "\n",
       "<sub>A linear parametric conditional average treatment effect (CATE) model was fitted:\n",
       "$Y = \\Theta(X)\\cdot T + g(X, W) + \\epsilon$\n",
       "where for every outcome $i$ and treatment $j$ the CATE $\\Theta_{ij}(X)$ has the form:\n",
       "$\\Theta_{ij}(X) = \\phi(X)' coef_{ij} + cate\\_intercept_{ij}$\n",
       "where $\\phi(X)$ is the output of the `featurizer` or $X$ if `featurizer`=None. Coefficient Results table portrays the $coef_{ij}$ parameter vector for each outcome $i$ and treatment $j$. Intercept Results table portrays the $cate\\_intercept_{ij}$ parameter.</sub>\n",
       "\"\"\""
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# inference on learned causal parameters\n",
    "est0.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/latex": [
       "$\\displaystyle -0.00253638347343484$"
      ],
      "text/plain": [
       "-0.002536383473434845"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# average CATEs over X to get a simple ATE estimate (inference only over the training set restricted to most prominent crop combination)\n",
    "est0.effect(X).mean()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 6. Visualizing the results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "# restrict the big dataset to random bbox (includes crop combinations are than the prominent ones)\n",
    "eco_to_viz0 = final_era.cx[2840415.203:2853466.007, 7252025.724:7261150.492].geometry.reset_index().drop(columns=['index'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "# assign the corresponding estimated CATEs to the new df\n",
    "eco_to_viz0 = eco_to_viz0.assign(C=est0.effect(final_era.cx[2840415.203:2853466.007, 7252025.724:7261150.492].iloc[:, -3:]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 576x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# visualize CATE map with geopandas (all rows)\n",
    "eco_to_viz0.plot(column='C', legend=True, figsize = (8,4));"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "# export shape file for QGIS beautification\n",
    "eco_to_viz0.to_file(\"results_all.shp\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 576x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# do the same for the restricted dataset (see parcels it also saw during training)\n",
    "eco_to_viz1 = final_era_f.cx[2840415.203:2853466.007, 7252025.724:7261150.492].geometry.reset_index().drop(columns=['index'])\n",
    "eco_to_viz1 = eco_to_viz1.assign(C=est0.effect(final_era_f.cx[2840415.203:2853466.007, 7252025.724:7261150.492].iloc[:, -3:]))\n",
    "eco_to_viz1.plot(column='C', legend=True, figsize = (8,4));"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "# export shape file for QGIS beautification\n",
    "eco_to_viz1.to_file(\"results_X_only.shp\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "eiffel",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.11"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "7258697ed51efff87c53f50f40c40e8fe823fc1f0641f496f75556fc2dc2d153"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}