Removing time_nextloop from Kernel loop

docs/examples_v3/tutorial_kernelloop.ipynb

@@ -26,7 +26,7 @@
     "\n",
     "When you run a Parcels simulation (i.e. a call to `pset.execute()`), the Kernel loop is the main part of the code that is executed. This part of the code loops through all particles and executes the Kernels that are defined for each particle.\n",
     "\n",
-    "In order to make sure that the displacements of a particle in the different Kernels can be summed, all Kernels add to a _change_ in position (`particles.dlon`, `particles.dlat`, and `particles.ddepth`). This is important, because there are situations where movement kernels would otherwise not commute. Take the example of advecting particles by currents _and_ winds. If the particle would first be moved by the currents and then by the winds, the result could be different from first moving by the winds and then by the currents. Instead, by adding the changes in position, the ordering of the Kernels has no consequence on the particle displacement."
+    "In order to make sure that the displacements of a particle in the different Kernels can be summed, all Kernels add to a _change_ in position (`particles.dlon`, `particles.dlat`, and `particles.dz`). This is important, because there are situations where movement kernels would otherwise not commute. Take the example of advecting particles by currents _and_ winds. If the particle would first be moved by the currents and then by the winds, the result could be different from first moving by the winds and then by the currents. Instead, by adding the changes in position, the ordering of the Kernels has no consequence on the particle displacement."
    ]
   },
   {
@@ -40,29 +40,27 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "Below is a structured overview of the Kernel loop is implemented. Note that this is for longitude only, but the same process is applied for latitude and depth.\n",
+    "Below is a structured overview of the Kernel loop is implemented. Note that this is for `lon` only, but the same process is applied for `lat` and `z`.\n",
     "\n",
-    "1. Initialise an extra Variable `particles.lon=0` and `particles.time_nextloop = particles.time`\n",
+    "1. Initialise an extra Variable `particles.dlon=0`\n",
     "\n",
     "2. Within the Kernel loop, for each particle:<br>\n",
     "\n",
     "    1. Update `particles.lon += particles.dlon`<br>\n",
     "\n",
-    "    2. Set variable `particles.dlon = 0`<br>\n",
+    "    2. Update `particles.time += particles.dt` (except for on the first iteration of the Kernel loop)<br>\n",
     "\n",
-    "    3. Update `particles.time = particles.time_nextloop`\n",
+    "    3. Set variable `particles.dlon = 0`<br>\n",
     "\n",
     "    4. For each Kernel in the list of Kernels:\n",
     "        \n",
     "        1. Execute the Kernel\n",
     "        \n",
     "        2. Update `particles.dlon` by adding the change in longitude, if needed<br>\n",
     "\n",
-    "    5. Update `particles.time_nextloop += particles.dt`<br>\n",
+    "    5. If `outputdt` is a multiple of `particle.time`, write `particle.lon` and `particle.time` to zarr output file<br>\n",
     "\n",
-    "    6. If `outputdt` is a multiple of `particle.time`, write `particle.lon` and `particle.time` to zarr output file<br>\n",
-    "\n",
-    "Besides having commutable Kernels, the main advantage of this implementation is that, when using Field Sampling with e.g. `particle.temp = fieldset.Temp[particle.time, particle.depth, particle.lat, particle.lon]`, the particle location stays the same throughout the entire Kernel loop. Additionally, this implementation ensures that the particle location is the same as the location of the sampled field in the output file."
+    "Besides having commutable Kernels, the main advantage of this implementation is that, when using Field Sampling with e.g. `particle.temp = fieldset.Temp[particle.time, particle.z, particle.lat, particle.lon]`, the particle location stays the same throughout the entire Kernel loop. Additionally, this implementation ensures that the particle location is the same as the location of the sampled field in the output file."
    ]
   },
   {
@@ -155,10 +153,10 @@
    "source": [
     "def wind_kernel(particle, fieldset, time):\n",
     "    particle_dlon += (\n",
-    "        fieldset.UWind[time, particle.depth, particle.lat, particle.lon] * particle.dt\n",
+    "        fieldset.UWind[time, particle.z, particle.lat, particle.lon] * particle.dt\n",
     "    )\n",
     "    particle_dlat += (\n",
-    "        fieldset.VWind[time, particle.depth, particle.lat, particle.lon] * particle.dt\n",
+    "        fieldset.VWind[time, particle.z, particle.lat, particle.lon] * particle.dt\n",
     "    )"
    ]
   },
@@ -242,30 +240,16 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "## Caveats"
+    "## Caveat: Avoid updating particle locations directly in Kernels"
    ]
   },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "There are a few important considerations to take into account when writing Kernels\n",
-    "\n",
-    "### 1. Avoid updating particle locations directly in Kernels\n",
     "It is better not to update `particle.lon` directly in a Kernel, as it can interfere with the loop above. Assigning a value to `particle.lon` in a Kernel will throw a warning. \n",
     "\n",
-    "Instead, update the local variable `particle.dlon`.\n",
-    "\n",
-    "### 2. Be careful with updating particle variables that do not depend on Fields.\n",
-    "While assigning the interpolated value of a `Field` to a Particle goes well in the loop above, this is not necessarily so for assigning other attributes. For example, a line like `particle.age += particle.dt` is executed directly so may result in the age being `dt` at `time = 0` in the output file. \n",
-    "\n",
-    "A workaround is to either initialise the age to `-dt`, or to increase the `age` only when `particle.time > 0` (using an `np.where` statement).\n",
-    "\n",
-    "\n",
-    "### 3. The last time is not written to file\n",
-    "Because the location at the start of the loop is written at the end of the Kernel loop, the last `particle.time` of the particle is not written to file. This is similar behaviour to e.g. `np.arange(start, stop)`, which also doesn't include the `stop` value itself. \n",
-    "\n",
-    "If you do want to write the last time to file, you can increase the `runtime` or `endtime` by `dt` (although this may cause an OutsideTimeInterval if your run was to the end of the available hydrodynamic data), or you can call `pfile.write_latest_locations(pset, time=pset[0].time_nextloop)`. Note that in the latter case, the particle locations (longitude, latitude and depth) will be updated, but other variables will _not_ be updated as the Kernels are not run again."
+    "Instead, update the local variable `particle.dlon`."
    ]
   },
   {
@@ -355,7 +339,7 @@
    "source": [
     "def KeepInOcean(particle, fieldset, time):\n",
     "    if particle.state == StatusCode.ErrorThroughSurface:\n",
-    "        particle_ddepth = 0.0\n",
+    "        particle_dz = 0.0\n",
     "        particle.state = StatusCode.Success"
    ]
   },

src/parcels/_core/kernel.py

@@ -75,7 +75,7 @@ def __init__(
         self._fieldset = fieldset
         self._ptype = ptype
 
-        self._positionupdate_kernels_added = False
+        self._positionupdate_kernel_added = False
 
         for f in pyfuncs:
             self.check_fieldsets_in_kernels(f)
@@ -111,23 +111,19 @@ def remove_deleted(self, pset):
         if len(indices) > 0:
             pset.remove_indices(indices)
 
-    def add_positionupdate_kernels(self):
+    def add_positionupdate_kernel(self):
         # Adding kernels that set and update the coordinate changes
-        def Setcoords(particles, fieldset):  # pragma: no cover
+        def PositionUpdate(particles, fieldset):  # pragma: no cover
             particles.lon += particles.dlon
             particles.lat += particles.dlat
             particles.z += particles.dz
+            particles.time += particles.dt
 
             particles.dlon = 0
             particles.dlat = 0
             particles.dz = 0
 
-            particles.time = particles.time_nextloop
-
-        def UpdateTime(particles, fieldset):  # pragma: no cover
-            particles.time_nextloop = particles.time + particles.dt
-
-        self._pyfuncs = (Setcoords + self + UpdateTime)._pyfuncs
+        self._pyfuncs = (PositionUpdate + self)._pyfuncs
 
     def check_fieldsets_in_kernels(self, pyfunc):  # TODO v4: this can go into another method? assert_is_compatible()?
         """
@@ -234,14 +230,13 @@ def execute(self, pset, endtime, dt):
 
         pset._data["state"][:] = StatusCode.Evaluate
 
-        if not self._positionupdate_kernels_added:
-            self.add_positionupdate_kernels()
-            self._positionupdate_kernels_added = True
-
         while (len(pset) > 0) and np.any(np.isin(pset.state, [StatusCode.Evaluate, StatusCode.Repeat])):
-            time_to_endtime = compute_time_direction * (endtime - pset.time_nextloop)
+            time_to_endtime = compute_time_direction * (endtime - pset.time)
 
-            if all(time_to_endtime <= 0):
+            evaluate_particles = (np.isin(pset.state, [StatusCode.Success, StatusCode.Evaluate])) & (
+                time_to_endtime >= 0
+            )
+            if not np.any(evaluate_particles):
                 return StatusCode.Success
 
             # adapt dt to end exactly on endtime
@@ -251,7 +246,6 @@ def execute(self, pset, endtime, dt):
                 pset.dt = np.minimum(np.maximum(pset.dt, -time_to_endtime), 0)
 
             # run kernels for all particles that need to be evaluated
-            evaluate_particles = (pset.state == StatusCode.Evaluate) & (pset.dt != 0)
             for f in self._pyfuncs:
                 f(pset[evaluate_particles], self._fieldset)
 
@@ -265,9 +259,9 @@ def execute(self, pset, endtime, dt):
             if not hasattr(self.fieldset, "RK45_tol"):
                 pset._data["dt"][:] = dt
 
-            # Reset particle state for particles that signalled success and have not reached endtime yet
-            particles_to_evaluate = (pset.state == StatusCode.Success) & (time_to_endtime > 0)
-            pset[particles_to_evaluate].state = StatusCode.Evaluate
+            # Set particle state for particles that reached endtime
+            particles_endofloop = (pset.state == StatusCode.Evaluate) & (pset.time == endtime)
+            pset[particles_endofloop].state = StatusCode.EndofLoop
 
             # delete particles that signalled deletion
             self.remove_deleted(pset)
@@ -284,4 +278,9 @@ def execute(self, pset, endtime, dt):
                     else:
                         error_func(pset[inds].z, pset[inds].lat, pset[inds].lon)
 
+            # Only add PositionUpdate kernel at the end of the first execute call to avoid adding dt to time too early
+            if not self._positionupdate_kernel_added:
+                self.add_positionupdate_kernel()
+                self._positionupdate_kernel_added = True
+
         return pset

src/parcels/_core/particle.py

@@ -180,7 +180,6 @@ def get_default_particle(spatial_dtype: np.float32 | np.float64) -> ParticleClas
                 dtype=_SAME_AS_FIELDSET_TIME_INTERVAL.VALUE,
                 attrs={"standard_name": "time", "units": "seconds", "axis": "T"},
             ),
-            Variable("time_nextloop", dtype=_SAME_AS_FIELDSET_TIME_INTERVAL.VALUE, to_write=False),
             Variable(
                 "trajectory",
                 dtype=np.int64,

src/parcels/_core/particlefile.py

@@ -71,6 +71,9 @@ def __init__(self, store, outputdt, chunks=None, create_new_zarrfile=True):
         if not isinstance(outputdt, np.timedelta64):
             raise ValueError(f"Expected outputdt to be a np.timedelta64 or datetime.timedelta, got {type(outputdt)}")
 
+        if outputdt <= np.timedelta64(0, "s"):
+            raise ValueError(f"outputdt must be a positive non-zero timedelta. Got {outputdt=!r}")
+
         self._outputdt = outputdt
 
         _assert_valid_chunks_tuple(chunks)
@@ -242,23 +245,6 @@ def _write_particle_data(self, *, particle_data, pclass, time_interval, time, in
 
         particle_data["obs_written"][indices_to_write] = obs + 1
 
-    def write_latest_locations(self, pset, time):
-        """Write the current (latest) particle locations to zarr file.
-        This can be useful at the end of a pset.execute(), when the last locations are not written yet.
-        Note that this only updates the locations, not any of the other Variables. Therefore, use with care.
-
-        Parameters
-        ----------
-        pset :
-            ParticleSet object to write
-        time :
-            Time at which to write ParticleSet. Note that typically this would be pset.time_nextloop
-        """
-        for var in ["lon", "lat", "z"]:
-            pset._data[f"{var}"] += pset._data[f"d{var}"]
-        pset._data["time"] = pset._data["time_nextloop"]
-        self.write(pset, time)
-
 
 def _get_store_from_pathlike(path: Path | str) -> DirectoryStore:
     path = str(Path(path))  # Ensure valid path, and convert to string
@@ -301,21 +287,21 @@ def _to_write_particles(particle_data, time):
         (
             np.less_equal(
                 time - np.abs(particle_data["dt"] / 2),
-                particle_data["time_nextloop"],
-                where=np.isfinite(particle_data["time_nextloop"]),
+                particle_data["time"],
+                where=np.isfinite(particle_data["time"]),
             )
             & np.greater_equal(
                 time + np.abs(particle_data["dt"] / 2),
-                particle_data["time_nextloop"],
-                where=np.isfinite(particle_data["time_nextloop"]),
+                particle_data["time"],
+                where=np.isfinite(particle_data["time"]),
             )  # check time - dt/2 <= particle_data["time"] <= time + dt/2
             | (
                 (np.isnan(particle_data["dt"]))
-                & np.equal(time, particle_data["time_nextloop"], where=np.isfinite(particle_data["time_nextloop"]))
+                & np.equal(time, particle_data["time"], where=np.isfinite(particle_data["time"]))
             )  # or dt is NaN and time matches particle_data["time"]
         )
         & (np.isfinite(particle_data["trajectory"]))
-        & (np.isfinite(particle_data["time_nextloop"]))
+        & (np.isfinite(particle_data["time"]))
     )[0]
 
 
@@ -324,9 +310,6 @@ def _convert_particle_data_time_to_float_seconds(particle_data, time_interval):
     particle_data = particle_data.copy()
 
     particle_data["time"] = ((particle_data["time"] - time_interval.left) / np.timedelta64(1, "s")).astype(np.float64)
-    particle_data["time_nextloop"] = (
-        (particle_data["time_nextloop"] - time_interval.left) / np.timedelta64(1, "s")
-    ).astype(np.float64)
     particle_data["dt"] = (particle_data["dt"] / np.timedelta64(1, "s")).astype(np.float64)
     return particle_data
 

src/parcels/_core/particleset.py

@@ -124,7 +124,6 @@ def __init__(
                 lat=lat,
                 z=z,
                 time=time,
-                time_nextloop=time,
                 trajectory=trajectory_ids,
             ),
         )
@@ -421,7 +420,7 @@ def update_dt_dtype(self, dt_dtype: np.dtype):
         dt_dtype : np.dtype
             New dtype for dt.
         """
-        if dt_dtype not in [np.timedelta64, "timedelta64[ns]", "timedelta64[ms]", "timedelta64[s]"]:
+        if dt_dtype not in [np.timedelta64, "timedelta64[ns]", "timedelta64[μs]", "timedelta64[ms]", "timedelta64[s]"]:
             raise ValueError(f"dt_dtype must be a numpy timedelta64 dtype. Got {dt_dtype=!r}")
 
         self._data["dt"] = self._data["dt"].astype(dt_dtype)
@@ -516,15 +515,16 @@ def execute(
                 raise ValueError(
                     f"The runtime must be a datetime.timedelta or np.timedelta64 object. Got {type(runtime)}"
                 ) from e
+            if runtime < np.timedelta64(0, "s"):
+                raise ValueError(f"The runtime must be a non-negative timedelta. Got {runtime=!r}")
 
         start_time, end_time = _get_simulation_start_and_end_times(
-            self.fieldset.time_interval, self._data["time_nextloop"], runtime, endtime, sign_dt
+            self.fieldset.time_interval, self._data["time"], runtime, endtime, sign_dt
         )
 
         # Set the time of the particles if it hadn't been set on initialisation
         if np.isnat(self._data["time"]).any():
             self._data["time"][:] = start_time
-            self._data["time_nextloop"][:] = start_time
 
         outputdt = output_file.outputdt if output_file else None
 
@@ -536,7 +536,7 @@ def execute(
             pbar = tqdm(total=(end_time - start_time) / np.timedelta64(1, "s"), file=sys.stdout)
             pbar.set_description("Integration time: " + str(start_time))
 
-        next_output = start_time + sign_dt * outputdt if output_file else None
+        next_output = start_time if output_file else None
 
         time = start_time
         while sign_dt * (time - end_time) < 0:
@@ -553,7 +553,7 @@ def execute(
                     if output_file:
                         output_file.write(self, next_output)
                     if np.isfinite(outputdt):
-                        next_output += outputdt
+                        next_output += outputdt * sign_dt
 
             if verbose_progress:
                 pbar.set_description("Integration time: " + str(time))

src/parcels/_core/statuscodes.py

@@ -20,6 +20,7 @@ class StatusCode:
     """Class defining the status codes for particles.state."""
 
     Success = 0
+    EndofLoop = 1
     Evaluate = 10
     Repeat = 20
     Delete = 30