diff --git a/src/care/DefaultMacros.h b/src/care/DefaultMacros.h
index 0f58821f..62c48072 100644
--- a/src/care/DefaultMacros.h
+++ b/src/care/DefaultMacros.h
@@ -267,6 +267,11 @@ OMP_FOR_BEGIN for (auto INDEX = _care_openmp_for_loop_begin_ndx; INDEX < _care_o
 
 #define CARE_CHECKED_PARALLEL_LOOP_END(CHECK) CARE_CHECKED_OPENMP_FOR_LOOP_END(CHECK)
 
+#define CARE_CHECKED_REDUCE_LOOP_START(INDEX, START_INDEX, END_INDEX, CHECK) \
+   CARE_CHECKED_PARALLEL_LOOP_START(INDEX, START_INDEX, END_INDEX, CHECK)
+
+#define CARE_CHECKED_REDUCE_LOOP_END(CHECK) CARE_CHECKED_PARALLEL_LOOP_END(CHECK)
+
 ////////////////////////////////////////////////////////////////////////////////
 ///
 /// @brief Macros that start and end a GPU RAJA loop of length one. If GPU is
@@ -545,17 +550,28 @@ OMP_FOR_BEGIN for (auto INDEX = _care_openmp_for_loop_begin_ndx; INDEX < _care_o
 #define CARE_SET_THREAD_ID(INDEX)
 #endif
 
-#define CARE_CHECKED_PARALLEL_LOOP_START(INDEX, START_INDEX, END_INDEX, CHECK) { \
+#define CARE_CHECKED_POLICY_LOOP_START(POLICY, INDEX, START_INDEX, END_INDEX, CHECK) { \
    auto _care_checked_loop_end = END_INDEX; \
    decltype(_care_checked_loop_end) _care_checked_loop_begin = START_INDEX; \
    if (_care_checked_loop_end > _care_checked_loop_begin) { \
       CARE_NEST_BEGIN(CHECK) \
-      care::forall(care::parallel{}, __FILE__, __LINE__, _care_checked_loop_begin, _care_checked_loop_end, [=] CARE_DEVICE (decltype(_care_checked_loop_end) INDEX) { \
+      care::forall(POLICY{}, __FILE__, __LINE__, _care_checked_loop_begin, _care_checked_loop_end, [=] CARE_DEVICE (decltype(_care_checked_loop_end) INDEX) { \
          CARE_SET_THREAD_ID(INDEX)
 
-#define CARE_CHECKED_PARALLEL_LOOP_END(CHECK) }); \
+#define CARE_CHECKED_POLICY_LOOP_END(CHECK) }); \
    CARE_NEST_END(CHECK) }}
 
+#define CARE_CHECKED_PARALLEL_LOOP_START(INDEX, START_INDEX, END_INDEX, CHECK) \
+   CARE_CHECKED_POLICY_LOOP_START(care::parallel,INDEX, START_INDEX, END_INDEX, CHECK)
+
+#define CARE_CHECKED_PARALLEL_LOOP_END(CHECK) CARE_CHECKED_POLICY_LOOP_END(CHECK)
+
+#define CARE_CHECKED_REDUCE_LOOP_START(INDEX, START_INDEX, END_INDEX, CHECK) \
+   CARE_CHECKED_POLICY_LOOP_START(care::parallel_reduce,INDEX, START_INDEX, END_INDEX, CHECK)
+
+#define CARE_CHECKED_REDUCE_LOOP_END(CHECK) CARE_CHECKED_POLICY_LOOP_END(CHECK)
+
+
 ////////////////////////////////////////////////////////////////////////////////
 ///
 /// @brief Macros that start and end a GPU RAJA loop of length one. If GPU is
@@ -830,9 +846,9 @@ OMP_FOR_BEGIN for (auto INDEX = _care_openmp_for_loop_begin_ndx; INDEX < _care_o
 /// @arg[in] END_INDEX The ending index (exclusive)
 ///
 ////////////////////////////////////////////////////////////////////////////////
-#define CARE_REDUCE_LOOP(INDEX, START_INDEX, END_INDEX) CARE_CHECKED_PARALLEL_LOOP_START(INDEX, START_INDEX, END_INDEX, care_reduce_loop_check)
+#define CARE_REDUCE_LOOP(INDEX, START_INDEX, END_INDEX) CARE_CHECKED_REDUCE_LOOP_START(INDEX, START_INDEX, END_INDEX, care_reduce_loop_check)
 
-#define CARE_REDUCE_LOOP_END CARE_CHECKED_PARALLEL_LOOP_END(care_reduce_loop_check)
+#define CARE_REDUCE_LOOP_END CARE_CHECKED_REDUCE_LOOP_END(care_reduce_loop_check)
 
 ////////////////////////////////////////////////////////////////////////////////
 ///
diff --git a/src/care/forall.h b/src/care/forall.h
index ea2ad41b..22813b1e 100644
--- a/src/care/forall.h
+++ b/src/care/forall.h
@@ -46,6 +46,10 @@ namespace care {
    struct ExecutionPolicyToSpace<RAJA::hip_exec<CARE_CUDA_BLOCK_SIZE, CARE_CUDA_ASYNC>> {
       static constexpr const chai::ExecutionSpace value = chai::GPU;
    };
+   template <>
+   struct ExecutionPolicyToSpace<RAJAReductionExec> {
+      static constexpr const chai::ExecutionSpace value = chai::GPU;
+   };
 #endif
 
 #if CARE_ENABLE_GPU_SIMULATION_MODE
@@ -228,6 +232,37 @@ namespace care {
 #endif
       PluginData::setParallelContext(false);
 
+#if CARE_ENABLE_PARALLEL_LOOP_BACKWARDS
+      s_reverseLoopOrder = false;
+#endif
+   }
+
+   ////////////////////////////////////////////////////////////////////////////////
+   ///
+   /// @author Peter Robinson
+   ///
+   /// @brief Execute using the care::RAJAReductionExec policy
+   ///
+   /// @arg[in] parallel_reducew Used to choose this overload of forall
+   /// @arg[in] fileName The name of the file where this function is called
+   /// @arg[in] lineNumber The line number in the file where this function is called
+   /// @arg[in] start The starting index (inclusive)
+   /// @arg[in] end The ending index (exclusive)
+   /// @arg[in] body The loop body to execute at each index
+   ///
+   ////////////////////////////////////////////////////////////////////////////////
+   template <typename LB>
+   void forall(parallel_reduce, const char * fileName, const int lineNumber,
+               const int start, const int end, LB&& body) {
+#if CARE_ENABLE_PARALLEL_LOOP_BACKWARDS
+      s_reverseLoopOrder = true;
+#endif
+      PluginData::setParallelContext(true);
+
+      forall(RAJAReductionExec{}, fileName, lineNumber, start, end, std::forward<LB>(body));
+
+      PluginData::setParallelContext(false);
+
 #if CARE_ENABLE_PARALLEL_LOOP_BACKWARDS
       s_reverseLoopOrder = false;
 #endif
diff --git a/src/care/policies.h b/src/care/policies.h
index 23eae95d..3252ff3c 100644
--- a/src/care/policies.h
+++ b/src/care/policies.h
@@ -15,6 +15,7 @@ namespace care {
    struct openmp {};
    struct gpu {};
    struct parallel {};
+   struct parallel_reduce {};
    struct raja_fusible {};
    struct raja_fusible_seq {};
    struct managed_ptr_read {};
@@ -27,6 +28,7 @@ namespace care {
       openmp,
       gpu,
       parallel,
+      parallel_reduce,
       managed_ptr_read,
       managed_ptr_write
    };
@@ -65,6 +67,12 @@ using RAJADeviceExec = RAJA::seq_exec;
 
 #endif // CARE_GPUCC
 
+//  reduction kernel policy
+#if defined(__HIPCC__)
+using RAJAReductionExec = RAJA::hip_exec_occ_calc<CARE_CUDA_BLOCK_SIZE, CARE_CUDA_ASYNC>;
+#else
+using RAJAReductionExec = RAJADeviceExec;
+#endif