"""X11: 宅地開発 × 農地転用 2016-2020 — 土地変換圧力の市町別メカニズム 広島県全20市町の「都市化圧力」を4データセット統合で定量化: 新築動向 (L20) × 宅地開発 (L21) × 農地転用 (L24) × 農林施策 (L25) L20/L21 の研究は「建物・開発の量的分布」を問うたが、 本記事は「農地→宅地の転換圧力はどこで起き、農林漁業政策と矛盾するか」 という政策矛盾の問いを追加する横断記事。 使用データセット: #1332–1516 新築動向 20件 (L20 の第2記事) #1351–1370 宅地開発 20件 (L21 の第2記事) #1391–1407 農地転用 17件 (L24 の第2記事) #1408–1424 農林漁業施策 17件 (L25 の第2記事) ─── 合計 74 dataset_id ─── 形式: X系横断記事 (Format B, RQ × 3 + 仮説 H1-H9) 実行時間目安: 60〜90 秒 """ from __future__ import annotations import sys, time, json, zipfile, glob from pathlib import Path sys.path.insert(0, str(Path(__file__).parent)) from _common import (ROOT, ASSETS, LESSONS, render_lesson, code, figure) import numpy as np import pandas as pd import matplotlib matplotlib.use("Agg") import matplotlib.pyplot as plt import matplotlib.patches as mpatches import geopandas as gpd import shapely plt.rcParams["font.family"] = "Yu Gothic" plt.rcParams["axes.unicode_minus"] = False t0 = time.time() print("=== X11 宅地開発 × 農地転用 土地変換分析 ===") TARGET_CRS = "EPSG:6671" CITY_NAME = { 101:"広島市中区",102:"広島市東区",103:"広島市南区",104:"広島市西区", 105:"広島市安佐南区",106:"広島市安佐北区",107:"広島市安芸区",108:"広島市佐伯区", 202:"呉市",203:"竹原市",204:"三原市",205:"尾道市",207:"福山市", 208:"府中市",209:"三次市",210:"庄原市",211:"大竹市",212:"東広島市", 213:"廿日市市",214:"安芸高田市",215:"江田島市", 302:"府中町",304:"海田町",307:"熊野町",309:"坂町", 369:"安芸太田町",462:"世羅町", } # 広島市 8区 → 「広島市」に集約 def aggr(city_cd): if 101 <= city_cd <= 108: return 100 # 広島市 return city_cd CITY_NAME[100] = "広島市" # ============================================================================= # 1. L20 新築動向 (46,613件) # ============================================================================= print("\n[1] L20 新築動向 ...") t1 = time.time() dfs20 = [] for zp in sorted(glob.glob(str(ROOT/"data"/"extras"/"L20_new_construction"/"shinchiku_*.zip"))): with zipfile.ZipFile(zp) as zf: gj = [f for f in zf.namelist() if f.endswith(".geojson")][0] with zf.open(gj) as f: gdf = gpd.read_file(f) dfs20.append(gdf[["CITY_CD","YOTO_SUB","TATE_H","NEW_YCD"]]) df20 = pd.concat(dfs20, ignore_index=True) df20["CITY_CD2"] = df20["CITY_CD"].apply(aggr) print(f" L20: {len(df20):,} 件, {time.time()-t1:.1f}s") # NEW_YCD: 1=2016, 2=2017, 3=2018, 4=2019, 5=2020 YEAR_MAP = {1:2016, 2:2017, 3:2018, 4:2019, 5:2020} df20["year"] = df20["NEW_YCD"].map(YEAR_MAP) # 用途グループ YOTO_GROUP = { "一戸建ての住宅":"住宅", "長屋":"住宅", "共同住宅":"住宅", } df20["yoto_grp"] = df20["YOTO_SUB"].map(YOTO_GROUP).fillna("非住宅") # 市町別集計 n20_city = df20.groupby("CITY_CD2").size().rename("新築件数").sort_values(ascending=False) n20_city_name = n20_city.rename(index=CITY_NAME) print(f" 市町別 Top3: {n20_city_name.head(3).to_dict()}") # ============================================================================= # 2. L21 宅地開発 (面積 m²) # ============================================================================= print("\n[2] L21 宅地開発 ...") t1 = time.time() dfs21 = [] for zp in sorted(glob.glob(str(ROOT/"data"/"extras"/"L21_housing_development"/"takuchi_*.zip"))): with zipfile.ZipFile(zp) as zf: gj = [f for f in zf.namelist() if f.endswith(".geojson")][0] with zf.open(gj) as f: gdf = gpd.read_file(f, crs="EPSG:4326") if "KAIHATU_A" in gdf.columns: area_m2 = pd.to_numeric(gdf["KAIHATU_A"], errors="coerce").fillna(0) else: # ALT スキーマ (江田島市・三原市・竹原市): ジオメトリ面積を使用 gdf_proj = gdf.to_crs(TARGET_CRS) area_m2 = gdf_proj.geometry.area dfs21.append(pd.DataFrame({"CITY_CD": gdf["CITY_CD"], "area_m2": area_m2})) df21 = pd.concat(dfs21, ignore_index=True) df21["CITY_CD2"] = df21["CITY_CD"].apply(aggr) area21_city = df21.groupby("CITY_CD2")["area_m2"].sum() / 10_000 # → ha area21_city = area21_city.rename("宅地開発_ha").sort_values(ascending=False) print(f" L21: {len(df21):,} 件, 総面積: {area21_city.sum():.0f} ha, {time.time()-t1:.1f}s") # ============================================================================= # 3. L24 農地転用 (面積 ha) # ============================================================================= print("\n[3] L24 農地転用 ...") t1 = time.time() dfs24 = [] for zp in sorted(glob.glob(str(ROOT/"data"/"extras"/"L24_farmland_conversion"/"farm_*.zip"))): with zipfile.ZipFile(zp) as zf: gj = [f for f in zf.namelist() if f.endswith(".geojson")][0] with zf.open(gj) as f: gdf = gpd.read_file(f) dfs24.append(gdf[["CITY_CD","AREA"]]) df24 = pd.concat(dfs24, ignore_index=True) df24["CITY_CD2"] = df24["CITY_CD"].apply(aggr) area24_city = df24.groupby("CITY_CD2")["AREA"].sum().rename("農地転用_ha").sort_values(ascending=False) print(f" L24: {len(df24):,} ポリゴン, 総面積: {area24_city.sum():.0f} ha, {time.time()-t1:.1f}s") # ============================================================================= # 4. L25 農林漁業施策 (ポリゴン面積) # ============================================================================= print("\n[4] L25 農林漁業施策 ...") t1 = time.time() dfs25 = [] for zp in sorted(glob.glob(str(ROOT/"data"/"extras"/"L25_agroforestry_policy"/"agro_*.zip"))): with zipfile.ZipFile(zp) as zf: gj = [f for f in zf.namelist() if f.endswith(".geojson")][0] with zf.open(gj) as f: gdf = gpd.read_file(f) gdf_proj = gdf.to_crs(TARGET_CRS) gdf_proj["area_m2"] = gdf_proj.geometry.area dfs25.append(gdf_proj[["CITY_CD","area_m2","geometry"]]) gdf25_all = gpd.GeoDataFrame(pd.concat(dfs25, ignore_index=True), geometry="geometry", crs=TARGET_CRS) gdf25_all["CITY_CD2"] = gdf25_all["CITY_CD"].apply(aggr) area25_city = (gdf25_all.groupby("CITY_CD2")["area_m2"].sum() / 10_000).rename("農林施策_ha") print(f" L25: {len(gdf25_all):,} ポリゴン, 総面積: {area25_city.sum():.0f} ha, {time.time()-t1:.1f}s") # ============================================================================= # 5. 行政区域 (面積用) # ============================================================================= print("\n[5] 行政区域 ...") t1 = time.time() ADMIN_GPKG = ROOT/"data"/"extras"/"L44_storm_surge"/"_cache"/"admin_diss.gpkg" admin = gpd.read_file(ADMIN_GPKG).to_crs(TARGET_CRS) admin["CITY_CD2"] = admin["CITY_CD"].apply(aggr) admin_area = admin.groupby("CITY_CD2")["admin_km2"].sum().rename("行政面積_km2") print(f" admin: {len(admin)} polys, {time.time()-t1:.1f}s") # ============================================================================= # 6. 統合テーブル # ============================================================================= print("\n[6] 統合分析 ...") df_all = pd.DataFrame({ "新築件数": n20_city, "宅地開発_ha": area21_city, "農地転用_ha": area24_city, "農林施策_ha": area25_city, "行政面積_km2": admin_area, }).fillna(0) df_all.index = pd.Index(df_all.index) df_all["市町名"] = df_all.index.map(lambda c: CITY_NAME.get(c, str(c))) df_all["新築密度_per_km2"] = df_all["新築件数"] / df_all["行政面積_km2"].replace(0, np.nan) df_all["開発効率_件per_ha"] = df_all["新築件数"] / df_all["宅地開発_ha"].replace(0, np.nan) df_all["転換率"] = df_all["農地転用_ha"] / df_all["宅地開発_ha"].replace(0, np.nan) # Lorenz / Gini（新築件数） def lorenz_gini(vals): v = np.sort(vals[vals > 0]) n = len(v) if n == 0: return np.array([0,1]), np.array([0,1]), 0.0 cum = np.cumsum(v) / v.sum() lorenz_x = np.linspace(0, 1, n+1) lorenz_y = np.concatenate([[0], cum]) trapz = getattr(np, "trapezoid", None) or np.trapz gini = 1 - 2 * trapz(lorenz_y, lorenz_x) return lorenz_x, lorenz_y, gini lx, ly, gini20 = lorenz_gini(df_all["新築件数"].values) lx21, ly21, gini21 = lorenz_gini(df_all["宅地開発_ha"].values) lx24, ly24, gini24 = lorenz_gini(df_all["農地転用_ha"].values) print(f" 新築件数 Gini: {gini20:.3f}, 宅地開発 Gini: {gini21:.3f}, 農地転用 Gini: {gini24:.3f}") # 上位3市のシェア top3_cities = df_all["新築件数"].sort_values(ascending=False).head(3).index top3_share = df_all.loc[top3_cities, "新築件数"].sum() / df_all["新築件数"].sum() * 100 print(f" 上位3市: {[CITY_NAME.get(c,str(c)) for c in top3_cities]}, シェア: {top3_share:.1f}%") # 年別新築件数 year_cnt = df20.groupby("year").size() print(f" 年別: {year_cnt.to_dict()}") # ============================================================================= # 7. 政策矛盾分析 (L25 農林施策 × L24 農地転用 GIS overlay) # ============================================================================= print("\n[7] 政策矛盾: 農林施策 × 農地転用 ...") t1 = time.time() dfs24_geo = [] for zp in sorted(glob.glob(str(ROOT/"data"/"extras"/"L24_farmland_conversion"/"farm_*.zip"))): with zipfile.ZipFile(zp) as zf: gj = [f for f in zf.namelist() if f.endswith(".geojson")][0] with zf.open(gj) as f: gdf = gpd.read_file(f) gdf = gdf.to_crs(TARGET_CRS) gdf["CITY_CD2"] = gdf["CITY_CD"].apply(aggr) dfs24_geo.append(gdf[["CITY_CD","CITY_CD2","AREA","geometry"]]) gdf24_all = gpd.GeoDataFrame(pd.concat(dfs24_geo, ignore_index=True), geometry="geometry", crs=TARGET_CRS) gdf24_all["geometry"] = gdf24_all.geometry.buffer(0) # sjoin: 農地転用ポリゴン が 農林施策ポリゴン と交差するか try: conflict_join = gpd.sjoin( gdf24_all[["CITY_CD","CITY_CD2","AREA","geometry"]].reset_index(drop=True), gdf25_all[["CITY_CD2","geometry"]].reset_index(drop=True), how="inner", predicate="intersects", lsuffix="farm", rsuffix="agro" ) n_conflict = conflict_join["CITY_CD_farm"].nunique() if "CITY_CD_farm" in conflict_join.columns else conflict_join.index.nunique() conflict_area_ha = conflict_join["AREA"].sum() conflict_cities = len(conflict_join.groupby("CITY_CD2_farm")) if "CITY_CD2_farm" in conflict_join.columns else len(conflict_join.groupby("CITY_CD2_left")) if "CITY_CD2_left" in conflict_join.columns else 0 print(f" 政策矛盾: 農地転用 × 農林施策 重複 {n_conflict} ポリゴン, 面積 {conflict_area_ha:.1f} ha, {time.time()-t1:.1f}s") # city-level conflict cc = "CITY_CD2_farm" if "CITY_CD2_farm" in conflict_join.columns else "CITY_CD2_left" if cc not in conflict_join.columns: cc = conflict_join.columns[conflict_join.columns.str.startswith("CITY_CD2")][0] conflict_by_city = (conflict_join.groupby(cc)["AREA"].sum() .rename("矛盾農転_ha").sort_values(ascending=False)) conflict_city_count = (conflict_by_city > 0).sum() print(f" 政策矛盾 市町数: {conflict_city_count}") except Exception as e: print(f" 政策矛盾 sjoin error: {e}") conflict_area_ha = 0.0 conflict_city_count = 0 conflict_by_city = pd.Series(dtype=float) # ============================================================================= # 8. 図の生成 # ============================================================================= print("\n=== 図生成 ===") # 図1: Lorenz 曲線 3本 t1 = time.time() fig, axes = plt.subplots(1, 2, figsize=(13, 5)) ax = axes[0] for (lx_, ly_, g_), label, color in [ ((lx, ly, gini20), f"新築件数 (Gini={gini20:.3f})", "#2563eb"), ((lx21, ly21, gini21), f"宅地開発面積 (Gini={gini21:.3f})", "#16a34a"), ((lx24, ly24, gini24), f"農地転用面積 (Gini={gini24:.3f})", "#dc2626"), ]: ax.plot(lx_, ly_, label=label, linewidth=1.8) ax.plot([0,1],[0,1],"k--",linewidth=1,alpha=0.5,label="均等線") ax.fill_between(lx, ly, [0,*lx[:-1]], alpha=0.08, color="#2563eb") ax.set_xlabel("累積市町割合", fontsize=11); ax.set_ylabel("累積シェア", fontsize=11) ax.set_title("広島県 土地変換 Lorenz 曲線\n（新築件数・宅地開発面積・農地転用面積）", fontsize=12) ax.legend(fontsize=9); ax.set_aspect("equal"); ax.grid(alpha=0.3) # 図1右: 年別新築件数推移 ax2 = axes[1] year_cnt.plot(ax=ax2, kind="bar", color="#3b82f6", edgecolor="#1d4ed8", linewidth=0.5) ax2.set_xlabel("年度", fontsize=11); ax2.set_ylabel("新築件数", fontsize=11) ax2.set_title(f"広島県 年別新築件数 (2016-2020)\n計 {len(df20):,} 件", fontsize=12) for p, v in zip(ax2.patches, year_cnt.values): ax2.text(p.get_x()+p.get_width()/2, p.get_height()+50, f"{v:,}", ha="center", fontsize=9) ax2.tick_params(axis="x", rotation=0) plt.tight_layout() p1 = ASSETS/"X11_lorenz_trend.png" plt.savefig(p1, dpi=120); plt.close() print(f" 図1: {p1.name}, {time.time()-t1:.1f}s") # 図2: 市町別新築件数 Top20 (広島市集約) t1 = time.time() fig, ax = plt.subplots(figsize=(12, 6)) top20 = df_all.sort_values("新築件数", ascending=False).head(20) bars = ax.bar(range(len(top20)), top20["新築件数"], color=["#1d4ed8" if n in ["広島市","福山市"] else "#3b82f6" for n in top20["市町名"]], edgecolor="#1e40af", linewidth=0.5) ax.set_xticks(range(len(top20))) ax.set_xticklabels(top20["市町名"], rotation=45, ha="right", fontsize=9) ax.set_ylabel("新築件数", fontsize=11) ax.set_title(f"市町別 新築建物件数 Top20 (2016-2020)\n上位3市シェア: {top3_share:.1f}%", fontsize=12) for bar, v in zip(bars, top20["新築件数"]): ax.text(bar.get_x()+bar.get_width()/2, bar.get_height()+50, f"{v:,}", ha="center", fontsize=8) plt.tight_layout() p2 = ASSETS/"X11_newbuild_by_city.png" plt.savefig(p2, dpi=120); plt.close() print(f" 図2: {p2.name}, {time.time()-t1:.1f}s") # 図3: 用途構成 (pie + bar) t1 = time.time() fig, axes = plt.subplots(1, 2, figsize=(13, 5)) ax = axes[0] yoto_top = df20["YOTO_SUB"].value_counts().head(8) other_n = len(df20) - yoto_top.sum() if other_n > 0: yoto_top["その他"] = other_n colors_pie = ["#1d4ed8","#3b82f6","#60a5fa","#93c5fd","#bfdbfe","#dbeafe","#eff6ff","#e2e8f0","#cbd5e1"] wedges, texts, autotexts = ax.pie( yoto_top, labels=yoto_top.index, autopct="%1.1f%%", colors=colors_pie[:len(yoto_top)], startangle=90, textprops={"fontsize":8} ) ax.set_title(f"新築建物 用途構成\n(n={len(df20):,}件)", fontsize=12) ax2 = axes[1] # 市町別住宅率 Top/Bottom yoto_city = df20.groupby("CITY_CD2")["yoto_grp"].value_counts(normalize=True).unstack(fill_value=0) if "住宅" in yoto_city.columns: yoto_city = yoto_city["住宅"].rename(index=CITY_NAME).sort_values() ax2.barh(range(len(yoto_city)), yoto_city.values*100, color=["#dc2626" if v < 0.7 else "#3b82f6" for v in yoto_city.values], edgecolor="#555", linewidth=0.4) ax2.set_yticks(range(len(yoto_city))) ax2.set_yticklabels(yoto_city.index, fontsize=8) ax2.set_xlabel("住宅系比率 (%)", fontsize=11) ax2.set_title("市町別 住宅系新築比率\n（赤 = 70%未満）", fontsize=12) ax2.axvline(x=70, color="red", linestyle="--", linewidth=1.2, label="70%閾値") ax2.legend(fontsize=9) plt.tight_layout() p3 = ASSETS/"X11_yoto_composition.png" plt.savefig(p3, dpi=120); plt.close() print(f" 図3: {p3.name}, {time.time()-t1:.1f}s") # 図4: 宅地開発面積 vs 新築件数 散布 t1 = time.time() fig, ax = plt.subplots(figsize=(9, 7)) df_plot = df_all[(df_all["宅地開発_ha"]>0) & (df_all["新築件数"]>0)].copy() ax.scatter(df_plot["宅地開発_ha"], df_plot["新築件数"], s=60, color="#3b82f6", edgecolor="#1d4ed8", alpha=0.8, linewidth=0.5) for _, row in df_plot.iterrows(): if row["新築件数"] > 1000 or row["宅地開発_ha"] > 10: ax.annotate(row["市町名"], (row["宅地開発_ha"], row["新築件数"]), fontsize=8, xytext=(3,3), textcoords="offset points") # 回帰線 from numpy.polynomial.polynomial import polyfit if len(df_plot) > 3: x_, y_ = df_plot["宅地開発_ha"].values, df_plot["新築件数"].values c = polyfit(x_, y_, 1) xr = np.linspace(0, x_.max(), 100) ax.plot(xr, c[0]+c[1]*xr, "r--", linewidth=1.2, alpha=0.7, label="回帰線") r = np.corrcoef(x_, y_)[0,1] ax.legend(fontsize=9, title=f"Pearson r={r:.3f}") ax.set_xlabel("宅地開発面積 (ha)", fontsize=11) ax.set_ylabel("新築件数", fontsize=11) ax.set_title("宅地開発面積 vs 新築件数 (市町別)\n2016-2020", fontsize=12) plt.tight_layout() p4 = ASSETS/"X11_dev_vs_newbuild.png" plt.savefig(p4, dpi=120); plt.close() print(f" 図4: {p4.name}, {time.time()-t1:.1f}s") # 図5: 農地転用 vs 宅地開発 散布（転換効率） t1 = time.time() fig, ax = plt.subplots(figsize=(9, 7)) df_farm = df_all[(df_all["農地転用_ha"]>0) & (df_all["宅地開発_ha"]>0)].copy() ax.scatter(df_farm["農地転用_ha"], df_farm["宅地開発_ha"], s=60, color="#16a34a", edgecolor="#15803d", alpha=0.8, linewidth=0.5) for _, row in df_farm.iterrows(): if row["農地転用_ha"] > 5 or row["宅地開発_ha"] > 10: ax.annotate(row["市町名"], (row["農地転用_ha"], row["宅地開発_ha"]), fontsize=8, xytext=(3,3), textcoords="offset points") # 等値線 (農転 = 宅地) max_v = max(df_farm["農地転用_ha"].max(), df_farm["宅地開発_ha"].max()) ax.plot([0,max_v],[0,max_v],"k--",linewidth=1,alpha=0.5, label="農転＝宅地開発（1:1線）") if len(df_farm) > 3: r_farm = np.corrcoef(df_farm["農地転用_ha"].values, df_farm["宅地開発_ha"].values)[0,1] ax.legend(fontsize=9, title=f"Pearson r={r_farm:.3f}") ax.set_xlabel("農地転用面積 (ha)", fontsize=11) ax.set_ylabel("宅地開発面積 (ha)", fontsize=11) ax.set_title("農地転用面積 vs 宅地開発面積 (市町別)\n1:1線より上 = 宅地開発が農転を超過", fontsize=12) plt.tight_layout() p5 = ASSETS/"X11_farmland_vs_dev.png" plt.savefig(p5, dpi=120); plt.close() print(f" 図5: {p5.name}, {time.time()-t1:.1f}s") # 図6: 農地転用 市町別面積 棒グラフ t1 = time.time() fig, axes = plt.subplots(1, 2, figsize=(13, 5)) ax = axes[0] farm_plot = area24_city.rename(index=CITY_NAME).sort_values(ascending=False).head(15).iloc[::-1] ax.barh(range(len(farm_plot)), farm_plot.values, color="#16a34a", edgecolor="#15803d", linewidth=0.5) ax.set_yticks(range(len(farm_plot))); ax.set_yticklabels(farm_plot.index, fontsize=9) ax.set_xlabel("農地転用面積 (ha)", fontsize=11) ax.set_title(f"市町別 農地転用面積 Top15\n(2016-2020, 全17市町計 {area24_city.sum():.0f} ha)", fontsize=12) for i,v in enumerate(farm_plot.values): ax.text(v+0.1, i, f"{v:.1f}", va="center", fontsize=8) ax2 = axes[1] # 農林施策面積 市町別 agro_plot = area25_city.rename(index=CITY_NAME).sort_values(ascending=False).head(15).iloc[::-1] ax2.barh(range(len(agro_plot)), agro_plot.values, color="#15803d", edgecolor="#166534", linewidth=0.5) ax2.set_yticks(range(len(agro_plot))); ax2.set_yticklabels(agro_plot.index, fontsize=9) ax2.set_xlabel("農林漁業施策面積 (ha)", fontsize=11) ax2.set_title(f"市町別 農林漁業施策面積 Top15\n(2022, 全17市町計 {area25_city.sum():.0f} ha)", fontsize=12) plt.tight_layout() p6 = ASSETS/"X11_farmland_agro.png" plt.savefig(p6, dpi=120); plt.close() print(f" 図6: {p6.name}, {time.time()-t1:.1f}s") # 図7: 建物高さ分布 ヒストグラム t1 = time.time() fig, axes = plt.subplots(1, 2, figsize=(13, 5)) ax = axes[0] h_vals = df20["TATE_H"].dropna() h_vals = h_vals[(h_vals > 0) & (h_vals < 100)] ax.hist(h_vals, bins=40, color="#3b82f6", edgecolor="#1d4ed8", linewidth=0.3, alpha=0.85) ax.axvline(x=h_vals.median(), color="red", linestyle="--", linewidth=1.5, label=f"中央値 {h_vals.median():.1f}m") ax.axvline(x=20, color="orange", linestyle="--", linewidth=1.2, label="20m (高層閾値)") ax.set_xlabel("建物高さ (m)", fontsize=11); ax.set_ylabel("棟数", fontsize=11) ax.set_title(f"新築建物 高さ分布\n(n={len(h_vals):,}棟)", fontsize=12) ax.legend(fontsize=9) ax2 = axes[1] # 高層建築 (>20m) 市町別件数 high_rise = df20[df20["TATE_H"] > 20].groupby("CITY_CD2").size().rename(index=CITY_NAME).sort_values(ascending=False).head(10) if len(high_rise) > 0: ax2.bar(range(len(high_rise)), high_rise.values, color="#1d4ed8", edgecolor="#1e3a8a", linewidth=0.5) ax2.set_xticks(range(len(high_rise))) ax2.set_xticklabels(high_rise.index, rotation=45, ha="right", fontsize=9) ax2.set_ylabel("高層建築件数 (>20m)", fontsize=11) ax2.set_title(f"市町別 高層建築 (>20m) 件数 Top10\n(全体 {(df20['TATE_H']>20).sum():,}件)", fontsize=12) plt.tight_layout() p7 = ASSETS/"X11_height_distribution.png" plt.savefig(p7, dpi=120); plt.close() print(f" 図7: {p7.name}, {time.time()-t1:.1f}s") # 図8: 転換率 (農地転用/宅地開発) 市町別バブル t1 = time.time() fig, ax = plt.subplots(figsize=(10, 6)) df_bubble = df_all[(df_all["農地転用_ha"]>0) & (df_all["宅地開発_ha"]>0)].copy() df_bubble["転換率"] = df_bubble["農地転用_ha"] / df_bubble["宅地開発_ha"] df_bubble = df_bubble.sort_values("転換率", ascending=False) colors_bubble = ["#dc2626" if r > 1 else "#16a34a" for r in df_bubble["転換率"]] bars = ax.bar(range(len(df_bubble)), df_bubble["転換率"], color=colors_bubble, edgecolor="#555", linewidth=0.4) ax.axhline(y=1.0, color="black", linestyle="--", linewidth=1.2, alpha=0.7, label="農転＝宅地開発（1:1）") ax.set_xticks(range(len(df_bubble))) ax.set_xticklabels(df_bubble["市町名"], rotation=45, ha="right", fontsize=8) ax.set_ylabel("転換率（農地転用面積 / 宅地開発面積）", fontsize=10) ax.set_title("市町別 農地転用 / 宅地開発 比率 (2016-2020)\n赤 = 農転が宅地開発を超過（農地の余剰転用？）", fontsize=12) ax.legend(fontsize=9) handles = [mpatches.Patch(facecolor="#dc2626",label=">1.0: 農転超過"), mpatches.Patch(facecolor="#16a34a",label="≤1.0: 宅地開発超過")] ax.legend(handles=handles, fontsize=9) plt.tight_layout() p8 = ASSETS/"X11_conversion_ratio.png" plt.savefig(p8, dpi=120); plt.close() print(f" 図8: {p8.name}, {time.time()-t1:.1f}s") # ============================================================================= # 9. CSV 出力 # ============================================================================= print("\n=== CSV 出力 ===") c1 = ASSETS/"X11_city_summary.csv" df_all[["市町名","新築件数","宅地開発_ha","農地転用_ha","農林施策_ha","行政面積_km2", "新築密度_per_km2","開発効率_件per_ha","転換率"]].to_csv(c1, encoding="utf-8-sig") c2 = ASSETS/"X11_yoto_sub.csv" df20["YOTO_SUB"].value_counts().reset_index().rename(columns={"YOTO_SUB":"件数","index":"用途"}).to_csv(c2, index=False, encoding="utf-8-sig") c3 = ASSETS/"X11_year_trend.csv" year_cnt.reset_index().rename(columns={"year":"年","count":"件数"}).to_csv(c3, index=False, encoding="utf-8-sig") c4 = ASSETS/"X11_lorenz.csv" pd.DataFrame({ "項目":["新築件数","宅地開発_ha","農地転用_ha"], "Gini係数":[round(gini20,4), round(gini21,4), round(gini24,4)] }).to_csv(c4, index=False, encoding="utf-8-sig") c5 = ASSETS/"X11_conflict.csv" if len(conflict_by_city) > 0: conflict_by_city.rename(index=CITY_NAME).reset_index().to_csv(c5, index=False, encoding="utf-8-sig") else: pd.DataFrame(columns=["市町名","矛盾農転_ha"]).to_csv(c5, index=False, encoding="utf-8-sig") print(" CSV 5件出力") # ============================================================================= # 10. 仮説検証 # ============================================================================= def jp(c): return "支持" if c else "反証" h1 = jp(gini20 >= 0.4) h2_top3 = [CITY_NAME.get(c,str(c)) for c in top3_cities] h2 = jp(top3_share >= 70) housing_rate_all = (df20["yoto_grp"]=="住宅").sum() / len(df20) * 100 h3 = jp(housing_rate_all >= 80) # H4: 高層 > 20m high_pct = (df20["TATE_H"] > 20).sum() / len(df20) * 100 h4 = jp(high_pct < 5) # 高層が少数 → 戸建て中心 # H5: 農転 vs 宅地開発 相関 if len(df_farm) > 3: r_f = np.corrcoef(df_farm["農地転用_ha"].values, df_farm["宅地開発_ha"].values)[0,1] else: r_f = 0 h5 = jp(r_f >= 0.5) # H6: Gini 農地転用 h6 = jp(gini24 >= 0.5) # H7: 政策矛盾 (農林施策 × 農転 重複) 市町数 >= 3 h7 = jp(conflict_city_count >= 3) # H8: 転換率 > 1 の市町 >= 3 over1 = (df_all["転換率"] > 1).sum() h8 = jp(over1 >= 3) # H9: 宅地開発 Gini と 農地転用 Gini の差 <= 0.1 (同じく偏在) h9 = jp(abs(gini21 - gini24) <= 0.15) # ============================================================================= # 11. HTML 出力 # ============================================================================= print("\n=== HTML 組立 ===") elapsed = time.time() - t0 sections = [ ("研究の問い (RQ) と仮説", f"""

L20 × L21 × L24 × L25 統合横断研究

L20（新築動向）と L21（宅地開発）は「建物の量的分布」を分析したが、 農地はどこから転換されたのか、農林施策との政策矛盾はないかを問うていない。 本記事は農地転用（L24）と農林漁業施策（L25）を加え、 「農地→宅地の転換圧力と政策矛盾」を 4 データセット統合で定量化する。

主 RQ

広島県 (2016-2020) の農地→宅地転換はどの市町で集中し、 農林漁業施策ポリゴン内での農地転用（政策矛盾）の面積と市町数はいくらか。

立てた仮説 (H1〜H9)

RQ1 — 新築建物の集中構造

RQ2 — 農地→宅地転換メカニズム

RQ3 — 農林施策との政策矛盾

"""), ("RQ1: 新築建物の集中構造", f"""

Lorenz 曲線と年別推移

{figure("X11_lorenz_trend.png", "図1: 土地変換 3 指標の Lorenz 曲線（左）と年別新築件数推移（右）")}

市町別新築件数

{figure("X11_newbuild_by_city.png", "図2: 市町別 新築建物件数 Top20 (2016-2020)")}

用途構成と住宅比率

{figure("X11_yoto_composition.png", "図3: 用途構成パイチャート（左）と市町別住宅系比率（右）")}

建物高さ分布

{figure("X11_height_distribution.png", "図7: 建物高さ分布ヒストグラム（左）と高層棟数 Top10（右）")} """), ("RQ2: 農地→宅地 転換メカニズム", f"""

農地転用と宅地開発の相関

{figure("X11_farmland_vs_dev.png", "図5: 農地転用面積 vs 宅地開発面積 (市町別)")} {figure("X11_dev_vs_newbuild.png", "図4: 宅地開発面積 vs 新築件数 (市町別)")}

宅地開発と新築件数は概ね正の相関 — 開発面積が大きいほど多くの建物が建てられる。 ただし外れ値（宅地開発は大きいが新築が少ない市町）は「開発後の宅地売れ残り」を示唆。

農地転用面積 市町別分布

{figure("X11_farmland_agro.png", "図6: 農地転用面積 Top15（左）と農林漁業施策面積 Top15（右）")}

転換率（農転/宅地開発）

{figure("X11_conversion_ratio.png", "図8: 市町別 転換率（農地転用面積 / 宅地開発面積）")} """), ("RQ3: 農林施策との政策矛盾", f"""

農林漁業施策ポリゴン × 農地転用 重複（政策矛盾ゾーン）

農林漁業施策（L25）は農地・森林・漁業インフラを保全する政策指定ゾーン。 このゾーン内で農地転用（L24）が行われていれば、保全政策と土地利用変換が矛盾する。

考察 — 農地保全と都市開発の綱引き

"""), ("仮説検証と考察", f"""

仮説 vs 結果

仮説判定根拠
H1: 新築件数 Gini ≥ 0.4{h1}Gini = {gini20:.3f}
H2: 上位3市シェア ≥ 70%{h2}{top3_share:.1f}% ({", ".join(h2_top3)})
H3: 住宅系 ≥ 80%{h3}{housing_rate_all:.1f}%
H4: 高層 <5%{h4}{high_pct:.1f}%
H5: 農転 vs 宅地 r ≥ 0.5{h5}r = {r_f:.3f}
H6: 農転 Gini ≥ 0.5{h6}Gini = {gini24:.3f}
H7: 政策矛盾市町 ≥ 3{h7}{conflict_city_count} 市町で矛盾発生
H8: 転換率 >1 の市町 ≥ 3{h8}{over1} 市町
H9: 宅地vs農転 Gini差 ≤ 0.15{h9}差 = {abs(gini21-gini24):.3f}

総合考察

"""), ("発展課題 (結果から導かれる新たな問い)", f"""
  1. 転換率 > 1.0 市町の詳細調査:
  2. 建替え vs 新規開発の比率推計:
"""), ("使用データセットと再現方法", f"""

使用 DoBoX データセット

ID名称件数役割
#1332–1516 都市計画区域情報_新築動向_2016-2020（20市町）20新築建物 4.6万件
#1351–1370 都市計画区域情報_宅地開発状況_2016-2020（20市町）20宅地開発面積
#1391–1407 都市計画区域情報_農地転用状況_2016-2020（17市町）17農地転用面積
#1408–1424 都市計画区域情報_農林漁業関係施策_2022（17市町）17政策ゾーン（政策矛盾検証）

合計: 74 dataset_id

再現方法

"""), ] html_out = render_lesson( num=11, title="X11: 宅地開発 × 農地転用 2016-2020 — 土地変換圧力の市町別メカニズム", tags=["X系","横断研究","宅地開発","農地転用","農林施策","政策矛盾","Gini","Lorenz"], time="90分", level="リテラシ基礎+α", data_label=( '新築動向 #1332-1516 (20件) × ' '宅地開発 #1351-1370 (20件) × ' '農地転用 #1391-1407 (17件) × ' '農林施策 #1408-1424 (17件)' ), sections=sections, script_filename="lessons/X11_land_conversion.py", ) html_out = html_out.replace("", '', 1) out_path = LESSONS/"X11_land_conversion.html" out_path.write_text(html_out, encoding="utf-8") print(f"\n[HTML] {out_path.name} ({out_path.stat().st_size:,} bytes)") print(f"=== DONE in {time.time()-t0:.1f}s ===")