{
  "morris": {
    "method": "morris",
    "model": "GR4J_MANAGED",
    "variant": "LuMod 0.1.3.0 GR4J; BasinForge conceptual managed-water layer",
    "objective": "nse",
    "score_period": "training data after warmup only",
    "training_steps": 1935,
    "seed": 42,
    "levels": 4,
    "trajectories": 8,
    "evaluations": 40,
    "parameters": [
      {
        "parameter": "wu_baseflow_fraction",
        "bounds": [
          0.1,
          0.3
        ],
        "mu": 0.011157023030545533,
        "mu_star": 0.01283875734774994,
        "sigma": 0.00993963584052799,
        "effects": [
          0.02152377801423072,
          0.015248841290568083,
          0.016362623844982216,
          -0.0035537219681051613,
          0.01039953495830237,
          0.010454590671914121,
          -0.003173215300712462,
          0.02199375273318438
        ]
      },
      {
        "parameter": "wu_groundwater_recession",
        "bounds": [
          0.01,
          0.1
        ],
        "mu": 0.005948009243014293,
        "mu_star": 0.005948009243014293,
        "sigma": 0.0034824174421994704,
        "effects": [
          0.013386030434299812,
          0.00611726247300054,
          0.005772513731213513,
          0.004739616880898634,
          0.006276462413321748,
          0.0008089485255636242,
          0.004726351257387795,
          0.00575688822842868
        ]
      },
      {
        "parameter": "wu_demand_scale",
        "bounds": [
          0.5,
          1.5
        ],
        "mu": 0.0025100437783668675,
        "mu_star": 0.0025100437783668675,
        "sigma": 0.0012391059347691327,
        "effects": [
          0.0019221999446099036,
          0.0018269497492143794,
          0.0041530415015813005,
          0.004724399371171806,
          0.0014791540964013694,
          0.0014791540964013694,
          0.0024342813579704137,
          0.002061170109584398
        ]
      },
      {
        "parameter": "wu_consumption_scale",
        "bounds": [
          0.5,
          1.0
        ],
        "mu": 0.002329101029698806,
        "mu_star": 0.002329101029698806,
        "sigma": 0.0010218011683904914,
        "effects": [
          0.0006405483312066362,
          0.0031416941400351783,
          0.002386386790714068,
          0.003114166283229469,
          0.0030472389956968104,
          0.0031022947093085618,
          0.0023078367036604153,
          0.0008926422837393089
        ]
      }
    ],
    "interpretation": "Normalized Morris effects on the minimized training objective. mu_star ranks influence within the reported bounds; sigma indicates effect variability, including possible interaction and nonlinearity."
  },
  "sobol": {
    "method": "sobol_jansen",
    "model": "GR4J_MANAGED",
    "variant": "LuMod 0.1.3.0 GR4J; BasinForge conceptual managed-water layer",
    "objective": "nse",
    "score_period": "training data after warmup only",
    "seed": 42,
    "base_samples": 64,
    "bootstrap_replicates": 50,
    "evaluations": 384,
    "parameters": [
      {
        "parameter": "wu_baseflow_fraction",
        "bounds": [
          0.1,
          0.3
        ],
        "first_order": 0.5187551069128638,
        "first_order_ci95": [
          0.31211512727820645,
          0.6444509574395172
        ],
        "total_order": 0.47371777941592685,
        "total_order_ci95": [
          0.3128939391789899,
          0.6100166110520077
        ],
        "first_order_estimator": "1 - mean((f(B) - f(A_Bi))^2) / (2 Var(Y))",
        "total_order_estimator": "mean((f(A) - f(A_Bi))^2) / (2 Var(Y))"
      },
      {
        "parameter": "wu_groundwater_recession",
        "bounds": [
          0.01,
          0.1
        ],
        "first_order": 0.4580569463519244,
        "first_order_ci95": [
          0.32865801616062074,
          0.6121995326159431
        ],
        "total_order": 0.4389723237935868,
        "total_order_ci95": [
          0.3172184158963748,
          0.6394772090176398
        ],
        "first_order_estimator": "1 - mean((f(B) - f(A_Bi))^2) / (2 Var(Y))",
        "total_order_estimator": "mean((f(A) - f(A_Bi))^2) / (2 Var(Y))"
      },
      {
        "parameter": "wu_demand_scale",
        "bounds": [
          0.5,
          1.5
        ],
        "first_order": 0.07710795961417982,
        "first_order_ci95": [
          -0.24365905424274495,
          0.32327014024075174
        ],
        "total_order": 0.04672029504625307,
        "total_order_ci95": [
          0.02916339850581352,
          0.07071098839840644
        ],
        "first_order_estimator": "1 - mean((f(B) - f(A_Bi))^2) / (2 Var(Y))",
        "total_order_estimator": "mean((f(A) - f(A_Bi))^2) / (2 Var(Y))"
      },
      {
        "parameter": "wu_consumption_scale",
        "bounds": [
          0.5,
          1.0
        ],
        "first_order": 0.04107857725765085,
        "first_order_ci95": [
          -0.2926939095126685,
          0.3276025686513298
        ],
        "total_order": 0.01894064890070636,
        "total_order_ci95": [
          0.011814187729041001,
          0.02733129369522831
        ],
        "first_order_estimator": "1 - mean((f(B) - f(A_Bi))^2) / (2 Var(Y))",
        "total_order_estimator": "mean((f(A) - f(A_Bi))^2) / (2 Var(Y))"
      }
    ],
    "interpretation": "Variance-based global effects over the reported independent parameter bounds. Finite-sample first-order estimates may be negative or sum above one. Bootstrap intervals describe index sampling variability, not predictive uncertainty."
  }
}