Backup automatico script del 2026-02-22 07:00

2026-02-22 07:00:03 +01:00
parent 11b6768fa3
commit c25c309a15
3 changed files with 499 additions and 250 deletions
@@ -16,6 +16,7 @@ BASE_DIR = os.path.dirname(os.path.abspath(__file__))
 DEFAULT_PATTERNS = ["*.log", "*_log.txt"]
 EXCLUDED_FILES = {
    "circondario_log.txt",
+    "irrigation_cron.log",
    "road_weather.log",
    "snow_radar.log",
 }
@@ -55,11 +55,18 @@ MODEL_NAMES = {
    "italia_meteo_arpae_icon_2i": "ICON Italia (ARPAE 2i)"
 }

+# Per Casa/Italia: forecast_days per modello lungo termine (come Agent Irrigazione / OPEN_METEO_MODELS.md)
+LONG_TERM_FORECAST_DAYS = {
+    "italia_meteo_arpae_icon_2i": 10,
+    "ecmwf_ifs": 10,
+    "meteofrance_seamless": 4,
+}
+
 def choose_models_by_country(cc, is_home=False):
    """
    Seleziona modelli meteo ottimali.
-    - Per Casa e Italia: solo ICON Italia (ARPAE 2i); AROME HD non copre San Marino.
-    - Per altre località: usa best match di Open-Meteo (senza specificare models).
+    - Per Casa e Italia: 0-2d mediana ICON Italia + AROME HD; 3-10d mediana ICON Italia + ECMWF IFS + ARPEGE.
+    - Per altre località: best match Open-Meteo.
    Ritorna (short_term_models, long_term_models)
    """
    cc = cc.upper() if cc else "UNKNOWN"
@@ -67,8 +74,11 @@ def choose_models_by_country(cc, is_home=False):
    long_term_default = ["gfs_global", "ecmwf_ifs04"]
    
    if is_home or cc == "IT":
-        # ICON Italia (0–72h) + ECMWF IFS per i giorni successivi (dove Icon Italia non arriva)
-        return ["italia_meteo_arpae_icon_2i"], ["ecmwf_ifs"]
+        # 0-2d: due modelli ad alta risoluzione (mediana). 3-10d: tre modelli (mediana, come Irrigazione).
+        return (
+            ["italia_meteo_arpae_icon_2i", "meteofrance_arome_france_hd"],
+            ["italia_meteo_arpae_icon_2i", "ecmwf_ifs", "meteofrance_seamless"],
+        )
    else:
        return None, long_term_default

@@ -153,22 +163,28 @@ def get_weather_multi_model(lat, lon, short_term_models, long_term_models, forec
        except:
            results["best_match"] = None
    else:
-        # Modelli specifici (per Casa: AROME + ICON, per Italia: ICON ARPAE)
+        # Modelli specifici: 0-2d ICON Italia + AROME HD (mediana); AROME HD solo 2 giorni
        for model in short_term_models:
            url = "https://api.open-meteo.com/v1/forecast"
-            # ICON Italia (ARPAE 2i): parametri come da API, senza precipitation_probability
            if model == "italia_meteo_arpae_icon_2i":
                hourly_params = "rain,showers,snowfall,snow_depth,precipitation,temperature_2m,weathercode,windspeed_10m,winddirection_10m"
                daily_params = "temperature_2m_max,temperature_2m_min,showers_sum,rain_sum,snowfall_sum,precipitation_sum,precipitation_hours,weathercode,winddirection_10m_dominant,windspeed_10m_max,windgusts_10m_max"
+                fd_short = min(forecast_days, 7)
+            elif model == "meteofrance_arome_france_hd":
+                # AROME HD: 2 giorni, set variabili ridotto (no snow_depth/showers in output)
+                hourly_params = "temperature_2m,precipitation,snowfall,rain,weathercode,windspeed_10m,winddirection_10m"
+                daily_params = "temperature_2m_max,temperature_2m_min,precipitation_sum,precipitation_hours,snowfall_sum,rain_sum,weathercode,winddirection_10m_dominant,windspeed_10m_max,windgusts_10m_max"
+                fd_short = 2
            else:
                hourly_params = "temperature_2m,precipitation,snowfall,rain,snow_depth,weathercode,windspeed_10m,windgusts_10m,winddirection_10m,dewpoint_2m,relative_humidity_2m,cloud_cover,soil_temperature_0cm"
                daily_params = "temperature_2m_max,temperature_2m_min,precipitation_sum,precipitation_hours,snowfall_sum,showers_sum,rain_sum,weathercode,winddirection_10m_dominant,windspeed_10m_max,windgusts_10m_max"
+                fd_short = min(forecast_days, 3)
            params = {
                "latitude": lat, "longitude": lon,
                "hourly": hourly_params,
                "daily": daily_params,
                "timezone": timezone if timezone else TZ_STR, "models": model,
-                "forecast_days": min(forecast_days, 7) if model == "italia_meteo_arpae_icon_2i" else min(forecast_days, 3)
+                "forecast_days": fd_short
            }
            try:
                resp = open_meteo_get(url, params=params, timeout=(5, 20))
@@ -218,21 +234,25 @@ def get_weather_multi_model(lat, lon, short_term_models, long_term_models, forec
            except:
                results[model] = None
    
-    # Recupera modelli a lungo termine (dopo 72h, dove Icon Italia non arriva)
+    # Recupera modelli a lungo termine (3-10d): tre modelli per mediana (come Agent Irrigazione)
    for model in (long_term_models or []):
        url = "https://api.open-meteo.com/v1/forecast"
-        # ECMWF IFS: parametri come da API (rain, showers, snowfall) + campi necessari per il report
+        fd_long = LONG_TERM_FORECAST_DAYS.get(model, forecast_days)
        if model == "ecmwf_ifs":
            hourly_params = "rain,showers,snowfall,precipitation,temperature_2m,weathercode,windspeed_10m,winddirection_10m"
            daily_params = "temperature_2m_max,temperature_2m_min,showers_sum,rain_sum,snowfall_sum,precipitation_sum,precipitation_hours,weathercode,winddirection_10m_dominant,windspeed_10m_max,windgusts_10m_max"
+        elif model == "meteofrance_seamless":
+            hourly_params = "temperature_2m,precipitation,snowfall,rain,weathercode,windspeed_10m,windgusts_10m,winddirection_10m"
+            daily_params = "temperature_2m_max,temperature_2m_min,precipitation_sum,precipitation_hours,snowfall_sum,rain_sum,weathercode,winddirection_10m_dominant,windspeed_10m_max,windgusts_10m_max"
        else:
-            hourly_params = "temperature_2m,precipitation,snowfall,rain,snow_depth,weathercode,windspeed_10m,windgusts_10m,winddirection_10m,dewpoint_2m,relative_humidity_2m,cloud_cover,soil_temperature_0cm"
-            daily_params = "temperature_2m_max,temperature_2m_min,precipitation_sum,precipitation_hours,snowfall_sum,showers_sum,rain_sum,weathercode,winddirection_10m_dominant,windspeed_10m_max,windgusts_10m_max"
+            # ICON Italia e altri
+            hourly_params = "rain,showers,snowfall,snow_depth,precipitation,temperature_2m,weathercode,windspeed_10m,winddirection_10m"
+            daily_params = "temperature_2m_max,temperature_2m_min,showers_sum,rain_sum,snowfall_sum,precipitation_sum,precipitation_hours,weathercode,winddirection_10m_dominant,windspeed_10m_max,windgusts_10m_max"
        params = {
            "latitude": lat, "longitude": lon,
            "hourly": hourly_params,
            "daily": daily_params,
-            "timezone": timezone if timezone else TZ_STR, "models": model, "forecast_days": forecast_days
+            "timezone": timezone if timezone else TZ_STR, "models": model, "forecast_days": fd_long
        }
        try:
            resp = open_meteo_get(url, params=params, timeout=(5, 25))
@@ -265,6 +285,144 @@ def get_weather_multi_model(lat, lon, short_term_models, long_term_models, forec
    
    return results

+
+def _normalize_time_key(t):
+    """Normalizza timestamp per confronto (YYYY-MM-DDTHH:MM)."""
+    if not t or not isinstance(t, str):
+        return str(t) if t else ""
+    return t.strip()[:16]
+
+
+def _median_or_single(values):
+    """Mediana dei valori numerici; ignora None. Con 2 valori restituisce la media dei due."""
+    nums = [float(v) for v in values if v is not None]
+    if not nums:
+        return None
+    if len(nums) == 1:
+        return nums[0]
+    return median(nums)
+
+
+# Chiavi che esistono solo su ICON Italia (no merge, si tiene il valore da quel modello)
+HOURLY_KEYS_ICON_ONLY = ["snow_depth", "showers"]
+DAILY_KEYS_ICON_ONLY = ["showers_sum"]
+
+
+def _merge_hourly_median(hourly_by_model, single_source_keys=None, single_source_model=None):
+    """
+    Unisce hourly da più modelli: mediana per ogni timestamp.
+    single_source_keys: per queste chiavi si prende il valore solo da single_source_model (es. ICON Italia per snow_depth, showers).
+    """
+    single_source_keys = single_source_keys or []
+    time_idx = {}
+    all_times = []
+    for _model, h in hourly_by_model:
+        times = h.get("time", []) or []
+        for t in times:
+            k = _normalize_time_key(str(t)) if t else ""
+            if k and k not in time_idx:
+                time_idx[k] = len(all_times)
+                all_times.append(t if isinstance(t, str) else k)
+    if not all_times:
+        return {"time": [], "temperature_2m": [], "precipitation": [], "snowfall": [], "rain": [], "weathercode": [], "windspeed_10m": [], "winddirection_10m": [], "snow_depth": [], "dewpoint_2m": [], "cloud_cover": [], "soil_temperature_0cm": []}
+    # Raccogli tutte le chiavi numeriche dal primo modello che le ha
+    all_keys = []
+    for _model, h in hourly_by_model:
+        for key in (h.keys() - {"time"}):
+            if key not in all_keys:
+                all_keys.append(key)
+    out = {"time": all_times}
+    for key in all_keys:
+        out[key] = []
+        for ref_t in all_times:
+            ref_k = _normalize_time_key(str(ref_t))
+            if key in single_source_keys and single_source_model:
+                val = None
+                for m, h in hourly_by_model:
+                    if m != single_source_model:
+                        continue
+                    times = h.get("time", []) or []
+                    arr = h.get(key, []) or []
+                    for i, t in enumerate(times):
+                        if _normalize_time_key(str(t)) == ref_k and i < len(arr) and arr[i] is not None:
+                            try:
+                                val = float(arr[i]) if key != "weathercode" else (int(arr[i]) if arr[i] is not None else None)
+                            except (TypeError, ValueError):
+                                pass
+                            break
+                    break
+                out[key].append(val)
+            else:
+                vals = []
+                for _m, h in hourly_by_model:
+                    times = h.get("time", []) or []
+                    arr = h.get(key, []) or []
+                    for i, t in enumerate(times):
+                        if _normalize_time_key(str(t)) == ref_k and i < len(arr) and arr[i] is not None:
+                            try:
+                                vals.append(float(arr[i]))
+                            except (TypeError, ValueError):
+                                pass
+                            break
+                out[key].append(_median_or_single(vals) if vals else None)
+    return out
+
+
+def _merge_daily_median(daily_by_model, single_source_keys=None, single_source_model=None):
+    """Unisce daily da più modelli: mediana per data. single_source_keys: valore solo da single_source_model."""
+    single_source_keys = single_source_keys or []
+    time_idx = {}
+    all_times = []
+    for _model, d in daily_by_model:
+        times = d.get("time", []) or []
+        for t in times:
+            key = str(t)[:10] if t else ""
+            if key and key not in time_idx:
+                time_idx[key] = len(all_times)
+                all_times.append(key)
+    if not all_times:
+        return {"time": [], "temperature_2m_max": [], "temperature_2m_min": [], "precipitation_sum": [], "precipitation_hours": [], "snowfall_sum": [], "showers_sum": [], "rain_sum": [], "weathercode": [], "winddirection_10m_dominant": [], "windspeed_10m_max": [], "windgusts_10m_max": []}
+    all_keys = []
+    for _model, d in daily_by_model:
+        for key in (d.keys() - {"time"}):
+            if key not in all_keys:
+                all_keys.append(key)
+    out = {"time": all_times}
+    for key in all_keys:
+        out[key] = []
+        for date_str in all_times:
+            if key in single_source_keys and single_source_model:
+                val = None
+                for m, d in daily_by_model:
+                    if m != single_source_model:
+                        continue
+                    times = d.get("time", []) or []
+                    arr = d.get(key, []) or []
+                    for i, t in enumerate(times):
+                        if str(t)[:10] == date_str and i < len(arr) and arr[i] is not None:
+                            try:
+                                val = float(arr[i]) if key != "weathercode" else (int(arr[i]) if arr[i] is not None else None)
+                            except (TypeError, ValueError):
+                                pass
+                            break
+                    break
+                out[key].append(val)
+            else:
+                vals = []
+                for _m, d in daily_by_model:
+                    times = d.get("time", []) or []
+                    arr = d.get(key, []) or []
+                    for i, t in enumerate(times):
+                        if str(t)[:10] == date_str and i < len(arr) and arr[i] is not None:
+                            try:
+                                vals.append(float(arr[i]))
+                            except (TypeError, ValueError):
+                                pass
+                            break
+                out[key].append(_median_or_single(vals) if vals else None)
+    return out
+
+
 def merge_multi_model_forecast(models_data, forecast_days=10):
    """Combina dati da modelli a breve e lungo termine in un forecast unificato"""
    merged = {
@@ -299,183 +457,141 @@ def merge_multi_model_forecast(models_data, forecast_days=10):
        "models_used": []
    }
    
-    # Trova modello a breve termine disponibile (cerca tutti i modelli con type "short_term")
-    # Priorità: ICON Italia per snow_depth, altrimenti primo disponibile
-    short_term_data = None
-    short_term_model = None
-    icon_italia_data = None
-    icon_italia_model = None
+    cutoff_day = 2  # 0-2d alta risoluzione, 3-10d mediana tre modelli
+    short_term_list = [(m, models_data[m]) for m in models_data if models_data.get(m) and models_data[m].get("model_type") == "short_term"]
+    long_term_list = [(m, models_data[m]) for m in models_data if models_data.get(m) and models_data[m].get("model_type") == "long_term"]
    
-    # Prima cerca ICON Italia (ha snow_depth quando disponibile)
-    # Cerca anche altri modelli che potrebbero avere snow_depth (icon_d2, etc.)
-    for model in models_data.keys():
-        if models_data[model] and models_data[model].get("model_type") == "short_term":
-            # Priorità a ICON Italia, ma cerca anche altri modelli con snow_depth
-            if model == "italia_meteo_arpae_icon_2i":
-                icon_italia_data = models_data[model]
-                icon_italia_model = model
-            # ICON-D2 può avere anche snow_depth
-            elif model == "icon_d2" and icon_italia_data is None:
-                # Usa ICON-D2 come fallback se ICON Italia non disponibile
-                hourly_data = models_data[model].get("hourly", {})
-                snow_depth_values = hourly_data.get("snow_depth", []) if hourly_data else []
-                # Verifica se ha dati di snow_depth validi
-                has_valid_snow_depth = False
-                if snow_depth_values:
-                    for sd in snow_depth_values[:24]:
-                        if sd is not None:
-                            try:
-                                if float(sd) > 0:
-                                    has_valid_snow_depth = True
-                                    break
-                            except (ValueError, TypeError):
-                                continue
-                if has_valid_snow_depth:
-                    icon_italia_data = models_data[model]
-                    icon_italia_model = model
-    
-    # Poi cerca primo modello disponibile (per altri parametri)
-    for model in models_data.keys():
-        if models_data[model] and models_data[model].get("model_type") == "short_term":
-            short_term_data = models_data[model]
-            short_term_model = model
-            break
-    
-    # Trova modello a lungo termine disponibile (cerca tutti i modelli con type "long_term")
-    long_term_data = None
-    long_term_model = None
-    for model in models_data.keys():
-        if models_data[model] and models_data[model].get("model_type") == "long_term":
-            long_term_data = models_data[model]
-            long_term_model = model
-            break
-    
-    if not short_term_data and not long_term_data:
+    if not short_term_list and not long_term_list:
        return None
    
-    # Usa dati a breve termine per primi 2-3 giorni, poi passa a lungo termine
-    cutoff_day = 2  # Usa modelli ad alta risoluzione per primi 2 giorni
+    daily_keys = list(merged["daily"].keys())
+    hourly_keys = list(merged["hourly"].keys())
    
-    if short_term_data:
-        # Gestisci best_match o modelli specifici
-        if short_term_model == "best_match":
-            model_display = "Best Match"
-        else:
-            model_display = MODEL_NAMES.get(short_term_model, short_term_model)
-        short_daily = short_term_data.get("daily", {})
-        short_hourly = short_term_data.get("hourly", {})
-        # Prendi dati daily: tutti i giorni se è l'unico modello, altrimenti primi cutoff_day+1
-        short_daily_times_all = short_daily.get("time", [])
-        short_daily_times = short_daily_times_all[:cutoff_day+1] if long_term_data else short_daily_times_all
+    def ensure_merged_keys(merged, daily_times, hourly_times):
+        for k in daily_keys:
+            if k == "time":
+                continue
+            while len(merged["daily"][k]) < len(merged["daily"]["time"]):
+                merged["daily"][k].append(None)
+        for k in hourly_keys:
+            if k == "time":
+                continue
+            while len(merged["hourly"][k]) < len(merged["hourly"]["time"]):
+                merged["hourly"][k].append(None)
    
-        # Verifica se ICON Italia ha dati di snow_depth (controllo diretto, non solo il flag)
-        has_icon_snow_depth = False
-        if icon_italia_data:
-            icon_hourly = icon_italia_data.get("hourly", {})
-            icon_snow_depth = icon_hourly.get("snow_depth", []) if icon_hourly else []
-            if icon_snow_depth:
-                for sd in icon_snow_depth[:72]:  # Controlla prime 72h
-                    if sd is not None:
-                        try:
-                            if float(sd) > 0:
-                                has_icon_snow_depth = True
-                                break
-                        except (ValueError, TypeError):
-                            continue
-        
-        num_days = len(short_daily_times)
-        if has_icon_snow_depth or (icon_italia_data and icon_italia_data.get("has_snow_depth_data")):
-            icon_display = MODEL_NAMES.get("italia_meteo_arpae_icon_2i", "ICON Italia")
-            merged["models_used"].append(f"{model_display} + {icon_display} (snow_depth) (0-{num_days}d)")
-        else:
-            merged["models_used"].append(f"{model_display} (0-{num_days}d)")
-        
-        for i, day_time in enumerate(short_daily_times):
-            merged["daily"]["time"].append(day_time)
-            for key in ["temperature_2m_max", "temperature_2m_min", "precipitation_sum", "precipitation_hours", "snowfall_sum", "showers_sum", "rain_sum", "weathercode", "winddirection_10m_dominant", "windspeed_10m_max", "windgusts_10m_max"]:
-                val = short_daily.get(key, [])[i] if i < len(short_daily.get(key, [])) else None
-                merged["daily"][key].append(val)
-        
-        # Prendi dati hourly dal modello a breve termine
-        # Priorità: usa snow_depth da ICON Italia se disponibile, altrimenti dal modello principale
-        short_hourly_times = short_hourly.get("time", [])
-        icon_italia_hourly = icon_italia_data.get("hourly", {}) if icon_italia_data else {}
-        icon_italia_hourly_times = icon_italia_hourly.get("time", []) if icon_italia_hourly else []
-        icon_italia_snow_depth = icon_italia_hourly.get("snow_depth", []) if icon_italia_hourly else []
-        # Crea mappa timestamp -> snow_depth per ICON Italia (per corrispondenza esatta o approssimata)
-        icon_snow_depth_map = {}
-        if icon_italia_hourly_times and icon_italia_snow_depth:
-            for idx, ts in enumerate(icon_italia_hourly_times):
-                if idx < len(icon_italia_snow_depth) and icon_italia_snow_depth[idx] is not None:
-                    try:
-                        val_cm = float(icon_italia_snow_depth[idx])
-                        if val_cm >= 0:  # Solo valori validi (già in cm)
-                            icon_snow_depth_map[ts] = val_cm
-                    except (ValueError, TypeError):
+    # ---- 0-2 giorni: uno o due modelli short-term ----
+    if short_term_list:
+        if len(short_term_list) >= 2:
+            # Mediana ICON Italia + AROME HD; snow_depth e showers solo da ICON Italia
+            short_daily_by_model = [(m, d.get("daily", {}) or {}) for m, d in short_term_list]
+            short_hourly_by_model = [(m, d.get("hourly", {}) or {}) for m, d in short_term_list]
+            merged_short_daily = _merge_daily_median(short_daily_by_model, single_source_keys=DAILY_KEYS_ICON_ONLY, single_source_model="italia_meteo_arpae_icon_2i")
+            merged_short_hourly = _merge_hourly_median(short_hourly_by_model, single_source_keys=HOURLY_KEYS_ICON_ONLY, single_source_model="italia_meteo_arpae_icon_2i")
+            short_daily_times = (merged_short_daily.get("time") or [])[:cutoff_day + 1] if long_term_list else (merged_short_daily.get("time") or [])
+            short_hourly_times = merged_short_hourly.get("time") or []
+            cutoff_h = (cutoff_day + 1) * 24 if long_term_list else len(short_hourly_times)
+            short_hourly_times = short_hourly_times[:cutoff_h]
+            names_short = " + ".join(MODEL_NAMES.get(m, m) for m, _ in short_term_list[:2])
+            merged["models_used"].append(f"{names_short} (mediana) (0-{len(short_daily_times)}d)")
+            for i, day_time in enumerate(short_daily_times):
+                merged["daily"]["time"].append(day_time)
+                for key in daily_keys:
+                    if key == "time":
                        continue
+                    arr = merged_short_daily.get(key, [])
+                    merged["daily"][key].append(arr[i] if i < len(arr) else None)
+            for i, hour_time in enumerate(short_hourly_times):
+                merged["hourly"]["time"].append(hour_time)
+                for key in hourly_keys:
+                    if key == "time":
+                        continue
+                    arr = merged_short_hourly.get(key, [])
+                    merged["hourly"][key].append(arr[i] if i < len(arr) else None)
+        else:
+            # Un solo modello short-term (es. best_match o fallback)
+            short_term_model, short_term_data = short_term_list[0]
+            short_daily = short_term_data.get("daily", {}) or {}
+            short_hourly = short_term_data.get("hourly", {}) or {}
+            short_daily_times_all = short_daily.get("time", []) or []
+            short_daily_times = short_daily_times_all[:cutoff_day + 1] if long_term_list else short_daily_times_all
+            model_display = "Best Match" if short_term_model == "best_match" else MODEL_NAMES.get(short_term_model, short_term_model)
+            merged["models_used"].append(f"{model_display} (0-{len(short_daily_times)}d)")
+            for i, day_time in enumerate(short_daily_times):
+                merged["daily"]["time"].append(day_time)
+                for key in daily_keys:
+                    if key == "time":
+                        continue
+                    arr = short_daily.get(key, [])
+                    merged["daily"][key].append(arr[i] if i < len(arr) else None)
+            short_hourly_times = short_hourly.get("time", []) or []
+            cutoff_h = (cutoff_day + 1) * 24 if long_term_list else len(short_hourly_times)
+            for i, hour_time in enumerate(short_hourly_times[:cutoff_h]):
+                merged["hourly"]["time"].append(hour_time)
+                for key in hourly_keys:
+                    if key == "time":
+                        continue
+                    arr = short_hourly.get(key, [])
+                    merged["hourly"][key].append(arr[i] if i < len(arr) else None)
+        ensure_merged_keys(merged, merged["daily"]["time"], merged["hourly"]["time"])
    
-        cutoff_hour = (cutoff_day + 1) * 24 if long_term_data else len(short_hourly_times)
-        for i, hour_time in enumerate(short_hourly_times[:cutoff_hour]):
-            merged["hourly"]["time"].append(hour_time)
-            for key in ["temperature_2m", "precipitation", "snowfall", "rain", "weathercode", "windspeed_10m", "winddirection_10m", "dewpoint_2m", "cloud_cover", "soil_temperature_0cm"]:
-                val = short_hourly.get(key, [])[i] if i < len(short_hourly.get(key, [])) else None
-                merged["hourly"][key].append(val)
-            # Per snow_depth: usa ICON Italia se disponibile (corrispondenza per timestamp), altrimenti modello principale
-            # NOTA: I valori sono già convertiti in cm durante il recupero dall'API
-            val_snow_depth = None
-            # Cerca corrispondenza esatta per timestamp
-            if hour_time in icon_snow_depth_map:
-                # Usa snow_depth da ICON Italia per questo timestamp (già in cm)
-                val_snow_depth = icon_snow_depth_map[hour_time]
-            else:
-                # Fallback 1: cerca corrispondenza per ora approssimata (se i timestamp non corrispondono esattamente)
-                # Estrai solo la parte ora (YYYY-MM-DDTHH) per corrispondenza approssimata
-                hour_time_base = hour_time[:13] if len(hour_time) >= 13 else hour_time  # "2025-01-09T12"
-                for icon_ts, icon_val in icon_snow_depth_map.items():
-                    if icon_ts.startswith(hour_time_base):
-                        val_snow_depth = icon_val
-                        break
-                # Fallback 2: se non trovato, cerca il valore più vicino nello stesso giorno
-                if val_snow_depth is None and hour_time_base:
-                    day_date_str = hour_time[:10] if len(hour_time) >= 10 else None  # "2025-01-09"
-                    if day_date_str:
-                        # Cerca tutti i valori di ICON Italia per lo stesso giorno
-                        same_day_values = [v for ts, v in icon_snow_depth_map.items() if ts.startswith(day_date_str)]
-                        if same_day_values:
-                            # Usa il primo valore disponibile per quel giorno (approssimazione)
-                            val_snow_depth = same_day_values[0]
-                # Fallback 3: usa snow_depth dal modello principale se ICON Italia non disponibile
-                if val_snow_depth is None and i < len(short_hourly.get("snow_depth", [])):
-                    val_snow_depth = short_hourly.get("snow_depth", [])[i]
-            merged["hourly"]["snow_depth"].append(val_snow_depth)
-    
-    if long_term_data:
-        merged["models_used"].append(f"{MODEL_NAMES.get(long_term_model, long_term_model)} ({cutoff_day+1}-{forecast_days}d)")
-        long_daily = long_term_data.get("daily", {})
-        long_hourly = long_term_data.get("hourly", {})
-        
-        # Prendi dati daily dal modello a lungo termine per i giorni successivi
-        long_daily_times = long_daily.get("time", [])
-        start_idx = cutoff_day + 1
-        
-        for i in range(start_idx, min(len(long_daily_times), forecast_days)):
-            merged["daily"]["time"].append(long_daily_times[i])
-            for key in ["temperature_2m_max", "temperature_2m_min", "precipitation_sum", "precipitation_hours", "snowfall_sum", "showers_sum", "rain_sum", "weathercode", "winddirection_10m_dominant", "windspeed_10m_max", "windgusts_10m_max"]:
-                val = long_daily.get(key, [])[i] if i < len(long_daily.get(key, [])) else None
-                merged["daily"][key].append(val)
-        
-        # Per i dati hourly, completa con dati a lungo termine se necessario
-        long_hourly_times = long_hourly.get("time", [])
-        current_hourly_count = len(merged["hourly"]["time"])
-        needed_hours = forecast_days * 24
-        
-        if current_hourly_count < needed_hours:
-            start_hour_idx = current_hourly_count
+    # ---- 3-10 giorni: uno o più modelli long-term ----
+    if long_term_list:
+        if len(long_term_list) >= 2:
+            long_daily_by_model = [(m, d.get("daily", {}) or {}) for m, d in long_term_list]
+            long_hourly_by_model = [(m, d.get("hourly", {}) or {}) for m, d in long_term_list]
+            merged_long_daily = _merge_daily_median(long_daily_by_model, single_source_keys=DAILY_KEYS_ICON_ONLY, single_source_model="italia_meteo_arpae_icon_2i")
+            merged_long_hourly = _merge_hourly_median(long_hourly_by_model, single_source_keys=HOURLY_KEYS_ICON_ONLY, single_source_model="italia_meteo_arpae_icon_2i")
+            long_daily_times = merged_long_daily.get("time") or []
+            long_hourly_times = merged_long_hourly.get("time") or []
+            names_long = " + ".join(MODEL_NAMES.get(m, m) for m, _ in long_term_list[:3])
+            merged["models_used"].append(f"{names_long} (mediana) ({cutoff_day+1}-{forecast_days}d)")
+            start_idx = cutoff_day + 1
+            for i, day_time in enumerate(long_daily_times):
+                day_num = i
+                if day_num < start_idx:
+                    continue
+                if day_num >= forecast_days:
+                    break
+                merged["daily"]["time"].append(day_time)
+                for key in daily_keys:
+                    if key == "time":
+                        continue
+                    arr = merged_long_daily.get(key, [])
+                    merged["daily"][key].append(arr[i] if i < len(arr) else None)
+            start_hour_idx = (cutoff_day + 1) * 24
+            needed_hours = forecast_days * 24
            for i in range(start_hour_idx, min(len(long_hourly_times), needed_hours)):
                merged["hourly"]["time"].append(long_hourly_times[i])
-                for key in ["temperature_2m", "precipitation", "snowfall", "snow_depth", "rain", "weathercode", "windspeed_10m", "winddirection_10m", "dewpoint_2m", "cloud_cover", "soil_temperature_0cm"]:
-                    val = long_hourly.get(key, [])[i] if i < len(long_hourly.get(key, [])) else None
-                    merged["hourly"][key].append(val)
+                for key in hourly_keys:
+                    if key == "time":
+                        continue
+                    arr = merged_long_hourly.get(key, [])
+                    merged["hourly"][key].append(arr[i] if i < len(arr) else None)
+        else:
+            long_term_model, long_term_data = long_term_list[0]
+            long_daily = long_term_data.get("daily", {}) or {}
+            long_hourly = long_term_data.get("hourly", {}) or {}
+            merged["models_used"].append(f"{MODEL_NAMES.get(long_term_model, long_term_model)} ({cutoff_day+1}-{forecast_days}d)")
+            long_daily_times = long_daily.get("time", []) or []
+            start_idx = cutoff_day + 1
+            for i in range(start_idx, min(len(long_daily_times), forecast_days)):
+                merged["daily"]["time"].append(long_daily_times[i])
+                for key in daily_keys:
+                    if key == "time":
+                        continue
+                    arr = long_daily.get(key, [])
+                    merged["daily"][key].append(arr[i] if i < len(arr) else None)
+            long_hourly_times = long_hourly.get("time", []) or []
+            current_hourly_count = len(merged["hourly"]["time"])
+            needed_hours = forecast_days * 24
+            for i in range(current_hourly_count, min(len(long_hourly_times), needed_hours)):
+                merged["hourly"]["time"].append(long_hourly_times[i])
+                for key in hourly_keys:
+                    if key == "time":
+                        continue
+                    arr = long_hourly.get(key, [])
+                    merged["hourly"][key].append(arr[i] if i < len(arr) else None)
+        ensure_merged_keys(merged, merged["daily"]["time"], merged["hourly"]["time"])
    
    return merged

@@ -18,6 +18,7 @@ import logging
 import os
 import sys
 from logging.handlers import RotatingFileHandler
+from statistics import median as _median
 from typing import Dict, List, Optional, Tuple
 from zoneinfo import ZoneInfo

@@ -42,11 +43,22 @@ TZINFO = ZoneInfo(TZ)

 # Open-Meteo: due fonti
 # - Suolo (tutti i layer): ICON Seamless (DWD) - copertura Europa centrale
-# - Meteo (ET₀, precipitazioni, T°): ICON Italia - risoluzione spaziale migliore per Italia/San Marino
+# - Meteo (ET₀, precipitazioni, T°): analisi a tre modelli con mediana (vedi OPEN_METEO_MODELS.md)
 OPEN_METEO_URL = "https://api.open-meteo.com/v1/forecast"
 MODEL_SOIL = "icon_seamless"   # Dati suolo (0-1, 1-3, 3-9, 9-27, 27-81 cm) e T° suolo; forecast_days=8
-MODEL_WEATHER = "italia_meteo_arpae_icon_2i"  # ET₀, precipitazioni, temperatura, radiazione
+MODEL_WEATHER = "italia_meteo_arpae_icon_2i"  # Retrocompatibilità / primo modello
 MODEL_ICON = MODEL_WEATHER  # Retrocompatibilità
+# Tre modelli per mediana (Europa/Italia: ICON Italia + ECMWF IFS + ARPEGE/Météo-France; ARPEGE preferito a GFS)
+WEATHER_MODELS_THREE = [
+    "italia_meteo_arpae_icon_2i",  # ~3 d utili, 2 km Italia/SM
+    "ecmwf_ifs",                   # 15 d, ~9 km
+    "meteofrance_seamless",        # ARPEGE+AROME, 4 d, 0.1° Europa
+]
+WEATHER_MODELS_FORECAST_DAYS = {
+    "italia_meteo_arpae_icon_2i": 10,
+    "ecmwf_ifs": 10,
+    "meteofrance_seamless": 4,
+}
 HTTP_HEADERS = {"User-Agent": "Smart-Irrigation-Advisor/2.0"}

 # Files
@@ -372,6 +384,176 @@ def fetch_weather_icon_italia(lat: float, lon: float, timezone: str = TZ) -> Opt
        return None


+def _weather_params_common(lat: float, lon: float, timezone: str) -> Dict:
+    """Parametri comuni hourly/daily per fetch meteo (ET₀, precipitazioni, ecc.)."""
+    return {
+        "latitude": lat,
+        "longitude": lon,
+        "timezone": timezone,
+        "hourly": ",".join([
+            "precipitation",
+            "snowfall",
+            "temperature_2m",
+            "relative_humidity_2m",
+            "et0_fao_evapotranspiration",
+            "vapour_pressure_deficit",
+            "direct_radiation",
+            "diffuse_radiation",
+            "shortwave_radiation",
+            "sunshine_duration",
+        ]),
+        "daily": ",".join([
+            "precipitation_sum",
+            "snowfall_sum",
+            "et0_fao_evapotranspiration_sum",
+            "sunshine_duration",
+        ]),
+    }
+
+
+def fetch_weather_single_model(
+    lat: float, lon: float, timezone: str, model: str, forecast_days: int = 10
+) -> Optional[Dict]:
+    """
+    Recupera dati meteo per un singolo modello Open-Meteo (stessa struttura di fetch_weather_icon_italia).
+    """
+    params = _weather_params_common(lat, lon, timezone)
+    params["forecast_days"] = forecast_days
+    params["models"] = model
+    try:
+        r = open_meteo_get(OPEN_METEO_URL, params=params, headers=HTTP_HEADERS, timeout=(5, 30))
+        r.raise_for_status()
+        return r.json()
+    except Exception as e:
+        LOGGER.debug("Open-Meteo single model %s error: %s", model, e)
+        return None
+
+
+def _median_or_single(values: List[Optional[float]]) -> Optional[float]:
+    """Mediana dei valori numerici; ignora None. Se nessun valore valido, ritorna None."""
+    nums = [float(v) for v in values if v is not None]
+    if not nums:
+        return None
+    if len(nums) == 1:
+        return nums[0]
+    return _median(nums)
+
+
+def _merge_daily_three_models_median(daily_list: List[Dict]) -> Dict:
+    """
+    Unisce i daily di più risposte meteo: per ogni data (unione di tutte) calcola la mediana
+    di et0_fao_evapotranspiration_sum, precipitation_sum, snowfall_sum, sunshine_duration.
+    """
+    time_idx: Dict[str, int] = {}
+    all_times: List[str] = []
+    for d in daily_list:
+        times = d.get("time", []) or []
+        for t in times:
+            key = str(t)[:10] if t else ""
+            if key and key not in time_idx:
+                time_idx[key] = len(all_times)
+                all_times.append(key)
+    if not all_times:
+        return {"time": [], "et0_fao_evapotranspiration_sum": [], "precipitation_sum": [], "snowfall_sum": [], "sunshine_duration": []}
+    # Per ogni data, indice in ogni daily
+    daily_keys = ["et0_fao_evapotranspiration_sum", "precipitation_sum", "snowfall_sum", "sunshine_duration"]
+    out: Dict[str, List] = {k: [] for k in daily_keys}
+    out["time"] = all_times
+    for date_str in all_times:
+        for key in daily_keys:
+            vals = []
+            for d in daily_list:
+                times = d.get("time", []) or []
+                arr = d.get(key, []) or []
+                for i, t in enumerate(times):
+                    if str(t)[:10] == date_str and i < len(arr) and arr[i] is not None:
+                        try:
+                            vals.append(float(arr[i]))
+                        except (TypeError, ValueError):
+                            pass
+                        break
+            out[key].append(_median_or_single(vals) if vals else None)
+    return out
+
+
+def _merge_hourly_three_models_median(hourly_list: List[Dict]) -> Dict:
+    """
+    Unisce gli hourly di più risposte: per ogni timestamp (unione) calcola la mediana
+    per ogni variabile numerica. Usa il primo dizionario per la lista dei nomi chiave.
+    """
+    time_idx: Dict[str, int] = {}
+    all_times: List[str] = []
+    for h in hourly_list:
+        times = h.get("time", []) or []
+        for t in times:
+            k = _normalize_time_key(str(t)) if t else ""
+            if k and k not in time_idx:
+                time_idx[k] = len(all_times)
+                all_times.append(t if isinstance(t, str) else k)
+    if not all_times:
+        keys = [k for k in (list(hourly_list[0].keys()) if hourly_list else []) if k != "time"]
+        return {"time": [], **{k: [] for k in keys}}
+    keys = [k for k in (list(hourly_list[0].keys()) if hourly_list else []) if k != "time"]
+    out: Dict[str, List] = {"time": all_times}
+    for key in keys:
+        out[key] = []
+        for ref_t in all_times:
+            ref_k = _normalize_time_key(str(ref_t))
+            vals = []
+            for h in hourly_list:
+                times = h.get("time", []) or []
+                arr = h.get(key, []) or []
+                for i, t in enumerate(times):
+                    if _normalize_time_key(str(t)) == ref_k and i < len(arr) and arr[i] is not None:
+                        try:
+                            vals.append(float(arr[i]))
+                        except (TypeError, ValueError):
+                            pass
+                        break
+            out[key].append(_median_or_single(vals) if vals else None)
+    return out
+
+
+def fetch_weather_three_models(lat: float, lon: float, timezone: str = TZ) -> Optional[Dict]:
+    """
+    Recupera meteo da tre modelli (ICON Italia, ECMWF IFS, ARPEGE/Météo-France) e restituisce
+    un unico payload con daily e hourly ottenuti dalla mediana per ogni giorno/ora.
+    Vedi OPEN_METEO_MODELS.md per la motivazione (ARPEGE preferito a GFS per Europa/Italia).
+    """
+    daily_list: List[Dict] = []
+    hourly_list: List[Dict] = []
+    meta = None
+    for model in WEATHER_MODELS_THREE:
+        fd = WEATHER_MODELS_FORECAST_DAYS.get(model, 10)
+        data = fetch_weather_single_model(lat, lon, timezone, model, forecast_days=fd)
+        if not data:
+            continue
+        if meta is None:
+            meta = {
+                "latitude": data.get("latitude"),
+                "longitude": data.get("longitude"),
+                "timezone": data.get("timezone"),
+            }
+        d = data.get("daily", {}) or {}
+        h = data.get("hourly", {}) or {}
+        if d.get("time"):
+            daily_list.append(d)
+        if h.get("time"):
+            hourly_list.append(h)
+    if not daily_list or meta is None:
+        LOGGER.warning("Three-model weather: no valid responses; fallback to single ICON Italia.")
+        return fetch_weather_icon_italia(lat, lon, timezone)
+    merged_daily = _merge_daily_three_models_median(daily_list)
+    merged_hourly = _merge_hourly_three_models_median(hourly_list) if hourly_list else {}
+    return {
+        "latitude": meta["latitude"],
+        "longitude": meta["longitude"],
+        "timezone": meta["timezone"],
+        "hourly": merged_hourly,
+        "daily": merged_daily,
+    }
+
+
 def _normalize_time_key(t: str) -> str:
    """Normalizza timestamp per confronto (es. '2026-02-05T16:00' e '2026-02-05T16:00:00' → stesso key)."""
    if not t or not isinstance(t, str):
@@ -437,11 +619,12 @@ def _merge_hourly_by_time(soil_hourly: Dict, weather_hourly: Dict, weather_daily

 def fetch_soil_and_weather(lat: float, lon: float, timezone: str = TZ) -> Optional[Dict]:
    """
-    Recupera dati combinati: suolo da ICON Seamless (tutti i layer), meteo da ICON Italia.
+    Recupera dati combinati: suolo da ICON Seamless (tutti i layer), meteo da analisi
+    a tre modelli (ICON Italia + ECMWF IFS + ARPEGE) con mediana di ET₀ e precipitazioni.
    In caso di fallimento suolo, prova fallback con singola fonte (solo ICON Italia).
    """
    soil_data = fetch_soil_icon_seamless(lat, lon, timezone)
-    weather_data = fetch_weather_icon_italia(lat, lon, timezone)
+    weather_data = fetch_weather_three_models(lat, lon, timezone)
    if not weather_data:
        return None
    hourly_w = weather_data.get("hourly", {}) or {}
@@ -506,8 +689,8 @@ def fetch_soil_and_weather_fallback(lat: float, lon: float, timezone: str = TZ)


 def fetch_weather_only(lat: float, lon: float, timezone: str = TZ) -> Optional[Dict]:
-    """Recupera solo dati meteo (senza suolo)."""
-    return fetch_weather_icon_italia(lat, lon, timezone)
+    """Recupera solo dati meteo (senza suolo): analisi a tre modelli con mediana."""
+    return fetch_weather_three_models(lat, lon, timezone)


 # =============================================================================
@@ -1353,6 +1536,15 @@ def build_irrigation_chart_bytes(
    ax1.plot(x, m27, "^-", color="C3", label="Umidità 27-81cm", markersize=4)
    ax1.axhline(y=SOIL_MOISTURE_DEEP_STRESS * 100, color="gray", linestyle="--", alpha=0.7, label=f"Trigger {SOIL_MOISTURE_DEEP_STRESS*100:.0f}%")
    ax1.axhline(y=SOIL_MOISTURE_WILTING_POINT * 100, color="brown", linestyle="--", alpha=0.7, label=f"Appassimento {SOIL_MOISTURE_WILTING_POINT*100:.0f}%")
+    # Linea verticale "oggi"
+    today_iso = now.date().isoformat()
+    now_idx = None
+    for i, d in enumerate(dates):
+        if d and str(d).startswith(today_iso[:10]):
+            now_idx = i
+            break
+    if now_idx is not None:
+        ax1.axvline(x=now_idx, color="red", linewidth=1, linestyle="-", alpha=0.9)
    ax1.legend(loc="upper right", fontsize=7)
    ax1.grid(True, alpha=0.3)
    ax1.set_ylim(bottom=0)
@@ -1362,6 +1554,8 @@ def build_irrigation_chart_bytes(
    precip_vals = [float(p) if p is not None else 0.0 for p in precip_list]
    ax2.bar([i - 0.2 for i in x], et0_vals, 0.35, label="ET₀", color="C0", alpha=0.8)
    ax2.bar([i + 0.2 for i in x], precip_vals, 0.35, label="Precip", color="C1", alpha=0.8)
+    if now_idx is not None:
+        ax2.axvline(x=now_idx, color="red", linewidth=1, linestyle="-", alpha=0.9)
    ax2.legend(loc="upper right", fontsize=7)
    ax2.grid(True, alpha=0.3)
    ax2.set_ylim(bottom=0)
@@ -1828,13 +2022,11 @@ def analyze_irrigation(
    if rain_veto:
        veto_lines.append(f"🌧️ **VETO PIOGGIA**: Ultime 24h ≥ {PRECIP_VETO_MM_24H:.0f} mm — non avviare irrigazione.")
    
-    # Colpo d'occhio
+    # Colpo d'occhio (umidità e prossimi 8 gg sono nel grafico)
    glance = [
        status.strip(),
        f"📍 {location_name} · {now.strftime('%d/%m/%Y %H:%M')}",
        "",
-        moisture_summary_line.strip(),
-        "",
    ]
    if veto_lines:
        glance.append("**Veti**")
@@ -1854,66 +2046,6 @@ def analyze_irrigation(
    ]
    if timing_advice:
        report_parts.append("**Orario** " + " · ".join(timing_advice))
-        report_parts.append("")
-    if planning_8d_line:
-        report_parts.append(planning_8d_line.strip())
-        report_parts.append("")
-    report_parts.append("─"*24)
-    
-    # Dettagli tecnici (compatti, at a glance)
-    details = []
-    soil_temp_0cm = _at(current_idx, soil_temp_0cm_list)
-    soil_temp_54cm = _at(current_idx, soil_temp_54cm_list)
-    temp_parts = []
-    for label, val in [("0cm", soil_temp_0cm), ("6cm", soil_temp_6cm), ("18cm", soil_temp_18cm), ("54cm", soil_temp_54cm)]:
-        if val is not None:
-            temp_parts.append(f"{label} {val:.1f}°C")
-    if temp_parts:
-        trend_str = f" · trend 7gg: {temp_trend}" if temp_trend else ""
-        details.append("🌡️ T° suolo: " + " · ".join(temp_parts) + trend_str)
-    elif soil_temp_0cm_list and current_idx < len(soil_temp_0cm_list) and soil_temp_0cm_list[current_idx] is not None:
-        details.append(f"🌡️ T° suolo 0cm: {float(soil_temp_0cm_list[current_idx]):.1f}°C")
-
-    moist_parts = []
-    any_at_fc = False
-    for label, val in [("0-1", soil_moisture_0_1cm), ("3-9", soil_moisture_3_9cm), ("9-27", soil_moisture_9_27cm), ("27-81", soil_moisture_27_81cm)]:
-        if val is not None:
-            moist_parts.append(f"{label} {val*100:.0f}%")
-            if val >= SOIL_MOISTURE_FIELD_CAPACITY:
-                any_at_fc = True
-        else:
-            moist_parts.append(f"{label} —")
-    if moist_parts:
-        line = "💧 Umidità: " + " · ".join(moist_parts)
-        if any_at_fc:
-            line += " — terreno pieno"
-        details.append(line)
-    if not details:
-        details.append("ℹ️ Dati suolo non disponibili")
-
-    # Una riga: ET₀, VPD, sole, umidità aria
-    meteo_parts = []
-    if et0_avg is not None:
-        meteo_parts.append(f"ET₀ {et0_avg:.1f} mm/d")
-    if vpd_avg is not None:
-        meteo_parts.append(f"VPD {vpd_avg:.2f} kPa")
-    if sunshine_hours is not None:
-        meteo_parts.append(f"Sole {sunshine_hours:.1f}h")
-    if humidity_avg is not None:
-        meteo_parts.append(f"UR {humidity_avg:.0f}%")
-    if meteo_parts:
-        details.append("☀️ " + " · ".join(meteo_parts))
-
-    # Precipitazioni: una riga
-    if future_rain_total > 0:
-        days_short = ", ".join(rainy_days[:3]) if rainy_days else ""
-        details.append(f"🌧️ Precip 5gg: {future_rain_total:.1f} mm — {days_short}")
-    else:
-        details.append("🌧️ Precip 5gg: 0 mm")
-    
-    if details:
-        report_parts.append("**Dettagli**")
-        report_parts.append("\n".join(details))
    
    # Salva stato
    save_state(state)