diff --git a/include/heyoka/model/sgp4.hpp b/include/heyoka/model/sgp4.hpp
index cfbf0b0b1..955476074 100644
--- a/include/heyoka/model/sgp4.hpp
+++ b/include/heyoka/model/sgp4.hpp
@@ -69,6 +69,10 @@ class HEYOKA_DLL_PUBLIC_INLINE_CLASS sgp4_propagator
     template <typename Input, typename... KwArgs>
     static auto parse_ctor_args(const Input &in, const KwArgs &...kw_args)
     {
+        if (in.data_handle() == nullptr) [[unlikely]] {
+            throw std::invalid_argument("Cannot initialise an sgp4_propagator with a null list of satellites");
+        }
+
         if (in.extent(1) == 0u) [[unlikely]] {
             throw std::invalid_argument("Cannot initialise an sgp4_propagator with an empty list of satellites");
         }
diff --git a/test/model_sgp4.cpp b/test/model_sgp4.cpp
index 442c0f69b..316205a6e 100644
--- a/test/model_sgp4.cpp
+++ b/test/model_sgp4.cpp
@@ -7,6 +7,8 @@
 // with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
 
 #include <cstddef>
+#include <ranges>
+#include <stdexcept>
 #include <vector>
 
 #include <boost/math/constants/constants.hpp>
@@ -150,7 +152,7 @@ TEST_CASE("propagator")
     const prop_t::in_1d<double> tm_in{tm.data(), 2};
 
     for (auto cm : {false, true}) {
-        prop_t prop{md_input_t{ins.data(), 9, 2}, kw::compact_mode = cm};
+        prop_t prop{md_input_t{ins.data(), 2}, kw::compact_mode = cm};
 
         std::vector<double> outs(12u);
         prop_t::out_2d out{outs.data(), 6, 2};
@@ -222,7 +224,7 @@ TEST_CASE("propagator batch")
     const prop_t::in_2d<double> tm_in{tm.data(), 2, 2};
 
     for (auto cm : {false, true}) {
-        prop_t prop{md_input_t{ins.data(), 9, 2}, kw::compact_mode = cm};
+        prop_t prop{md_input_t{ins.data(), 2}, kw::compact_mode = cm};
 
         std::vector<double> outs(24u);
         prop_t::out_3d out{outs.data(), 2, 6, 2};
@@ -288,3 +290,84 @@ TEST_CASE("propagator batch")
         REQUIRE(out(1, 5, 1) == approximately(1.785277174529374, 10000.));
     }
 }
+
+TEST_CASE("error handling")
+{
+    detail::edb_disabler ed;
+
+    using Catch::Matchers::Message;
+    using md_input_t = mdspan<const double, extents<std::size_t, 9, std::dynamic_extent>>;
+    using prop_t = model::sgp4_propagator<double>;
+
+    // Propagator with null list or zero satellites.
+    REQUIRE_THROWS_MATCHES((prop_t{md_input_t{nullptr, 0}}), std::invalid_argument,
+                           Message("Cannot initialise an sgp4_propagator with a null list of satellites"));
+
+    std::vector<double> input(9u);
+
+    REQUIRE_THROWS_MATCHES((prop_t{md_input_t{input.data(), 0}}), std::invalid_argument,
+                           Message("Cannot initialise an sgp4_propagator with an empty list of satellites"));
+}
+
+TEST_CASE("derivatives")
+{
+    detail::edb_disabler ed;
+
+    using md_input_t = mdspan<const double, extents<std::size_t, 9, std::dynamic_extent>>;
+    using prop_t = model::sgp4_propagator<double>;
+
+    // First compute the order-2 derivatives of the whole model.
+    const auto sgp4_func = model::sgp4();
+    const auto inputs = make_vars("n0", "e0", "i0", "node0", "omega0", "m0", "bstar", "tsince");
+    const auto dt = diff_tensors(sgp4_func, std::vector(inputs.begin(), inputs.begin() + 6), kw::diff_order = 2);
+
+    // Make a compiled function with the derivatives.
+    auto diff_cf = cfunc<double>(dt | std::views::transform([](const auto &p) { return p.second; }), inputs);
+
+    // Create a propagator with derivatives.
+    const std::vector<double> ins = {revday2radmin(13.75091047972192),
+                                     revday2radmin(15.50103472202482),
+                                     0.0024963,
+                                     0.0007417,
+                                     deg2rad(90.2039),
+                                     deg2rad(51.6439),
+                                     deg2rad(55.5633),
+                                     deg2rad(211.2001),
+                                     deg2rad(320.5956),
+                                     deg2rad(17.6667),
+                                     deg2rad(91.4738),
+                                     deg2rad(85.6398),
+                                     0.75863e-3,
+                                     .38792e-4,
+                                     2460486.5,
+                                     2458826.5,
+                                     0.6478633000000116,
+                                     0.6933954099999937};
+
+    const auto tm = std::array{1440., 0.};
+    const prop_t::in_1d<double> tm_in{tm.data(), 2};
+
+    prop_t prop{md_input_t{ins.data(), 2}, kw::diff_order = 2};
+
+    // Prepare the input buffer for the cfunc.
+    std::vector<double> cf_in(ins.begin(), ins.begin() + 14);
+    cf_in.insert(cf_in.end(), tm.begin(), tm.end());
+    cfunc<double>::in_2d cf_in_span(cf_in.data(), 8, 2);
+
+    // Prepare the output buffers.
+    std::vector<double> cf_out(dt.size() * 2u), prop_out(cf_out);
+    cfunc<double>::out_2d cf_out_span(cf_out.data(), dt.size(), 2);
+    cfunc<double>::out_2d prop_out_span(prop_out.data(), dt.size(), 2);
+
+    // Evaluate the cfunc.
+    diff_cf(cf_out_span, cf_in_span);
+
+    // Evaluate the propagation.
+    prop(prop_out_span, tm_in);
+
+    for (std::size_t i = 0; i < prop_out_span.extent(0); ++i) {
+        for (std::size_t j = 0; j < prop_out_span.extent(1); ++j) {
+            REQUIRE(prop_out_span(i, j) == approximately(cf_out_span(i, j)));
+        }
+    }
+}