diff --git a/tools/CMakeLists.txt b/tools/CMakeLists.txt
index 109898750..555fcfa0d 100644
--- a/tools/CMakeLists.txt
+++ b/tools/CMakeLists.txt
@@ -34,6 +34,7 @@ list( APPEND grib_tools_bins_EXTRA
              grib_repair
              grib_to_json
              grib_check_gaussian_grid
+             bufr_split_by_subtype
              )
 
 # Install generic tools
diff --git a/tools/bufr_split_by_subtype.c b/tools/bufr_split_by_subtype.c
new file mode 100644
index 000000000..c3fe70a62
--- /dev/null
+++ b/tools/bufr_split_by_subtype.c
@@ -0,0 +1,158 @@
+/*
+ * Copyright 2005-2018 ECMWF.
+ *
+ * This software is licensed under the terms of the Apache Licence Version 2.0
+ * which can be obtained at http://www.apache.org/licenses/LICENSE-2.0.
+ *
+ * In applying this licence, ECMWF does not waive the privileges and immunities granted to it by
+ * virtue of its status as an intergovernmental organisation nor does it submit to any jurisdiction.
+ */
+
+/*
+ * Description:
+ * Split an input BUFR file into output files according to 'rdbSubtype'.
+ * The output files are named split_rdbSubtype.<rdbSubtype>.bufr.
+ * This is much faster than bufr_copy/bufr_filter
+ *
+ */
+
+#include "grib_api_internal.h"
+#include <assert.h>
+
+static int verbose = 0;
+static void usage(const char* prog)
+{
+    printf("usage: %s [-v] infile\n",prog);
+    exit(1);
+}
+static int file_exists(const char* path)
+{
+    struct stat s;
+    int stat_val = stat(path,&s);
+    if (stat_val != 0) return 0; /*error doing stat*/
+    return S_ISREG( s.st_mode );
+}
+
+static int split_file_by_subtype(FILE* in, const char* filename, unsigned long *count)
+{
+    void* mesg=NULL;
+    FILE* out;
+    size_t size=0;
+    off_t offset=0;
+    int err=GRIB_SUCCESS;
+    char ofilename[2048]={0,};
+    grib_context* c=grib_context_get_default();
+
+    if (!in) return 1;
+    sprintf(ofilename, "split_rdbSubtype.bufr"); /*default name*/
+
+    while ( err!=GRIB_END_OF_FILE ) {
+        mesg=wmo_read_any_from_file_malloc(in, 0, &size, &offset, &err);
+        if (mesg!=NULL && err==0) {
+            /*Decode subtype from mesg*/
+            unsigned long edition = 0;
+            long pos_edition = 7*8;
+            const long nbits_edition = 8; /*1 byte*/
+            long section1Flags=0;
+            long nbits_section1Flags=1*8;
+            long pos_section1Flags=17*8;
+            long section1Length = 0;
+            const long nbits_section1Length = 24; /*3 bytes*/
+            long pos_section1Length = 8*8;
+
+            edition = (long)grib_decode_unsigned_long(mesg, &pos_edition, nbits_edition);
+            section1Length=(long)grib_decode_unsigned_long(mesg, &pos_section1Length, nbits_section1Length);
+            if (edition==3) pos_section1Flags=15*8;
+            section1Flags=(long)grib_decode_unsigned_long(mesg, &pos_section1Flags, nbits_section1Flags);
+
+            sprintf(ofilename, "split_rdbSubtype.bufr");
+            if (section1Flags!=0) {
+                long oldSubtype=0;
+                long newSubtype=0;
+                long rdbSubtype=0;
+                long nbits_oldSubtype=1*8;
+                long nbits_newSubtype=2*8;
+                long pos_oldSubtype = (8 + section1Length + 5)*8;
+                long pos_newSubtype = (8 + section1Length + 49)*8;
+
+                oldSubtype=(long)grib_decode_unsigned_long(mesg, &pos_oldSubtype, nbits_oldSubtype);
+                newSubtype=(long)grib_decode_unsigned_long(mesg, &pos_newSubtype, nbits_newSubtype);
+                /*printf("oldSubtype=%d   newSubtype=%d\n", oldSubtype, newSubtype);*/
+                
+                if (oldSubtype<255) rdbSubtype=oldSubtype;
+                else                rdbSubtype=newSubtype;
+                /*printf("=> rdbSubtype=%d\n", rdbSubtype);*/
+                sprintf(ofilename, "split_rdbSubtype.%ld.bufr", rdbSubtype);
+            }
+            if (verbose) {
+                if (!file_exists(ofilename)) {
+                    printf("Writing output to %s\n", ofilename);
+                }
+            }
+            out=fopen(ofilename,"a");
+            if (!out) {
+                fprintf(stderr,"ERROR: Failed to append to file '%s'\n", ofilename);
+                perror(ofilename);
+                return GRIB_IO_PROBLEM;
+            }
+            if (fwrite(mesg,1,size,out)!=size) {
+                perror(ofilename);
+                fclose(out);
+                return GRIB_IO_PROBLEM;
+            }
+            grib_context_free(c,mesg);
+            fclose(out);
+            (*count)++;
+        }
+    }
+
+    if (err==GRIB_END_OF_FILE) err=GRIB_SUCCESS;
+
+    return err;
+}
+
+int main(int argc,char* argv[])
+{
+    FILE* infh = NULL;
+    char* filename;
+    int i, status=0;
+    struct stat s;
+    int err=0;
+    unsigned long count=0;
+
+    if (argc!=2 && argc!=3) usage(argv[0]);
+
+    i=1;
+    if (strcmp(argv[i], "-v")==0) {
+        i++;
+        verbose = 1;
+        if (argc !=3) usage(argv[0]);
+    }
+
+    filename=argv[i];
+    if (stat(filename, &s)==0) {
+        if (S_ISDIR(s.st_mode)) {
+            fprintf(stderr, "ERROR: %s: Is a directory\n", filename);
+            return 1;
+        }
+    }
+    infh=fopen(filename,"r");
+    if (!infh) {
+        perror(filename);
+        return 1;
+    }
+
+    count=0;
+    err=split_file_by_subtype(infh, filename, &count);
+    if (err) {
+        fprintf(stderr,"ERROR: Failed to split file %s", filename);
+        fprintf(stderr,"\n");
+        status = 1;
+    } else {
+        if (verbose) printf ("%7lu %s\n", count, filename);
+    }
+
+    fclose(infh);
+
+    return status;
+}