diff --git a/notebooks/Sampling_comparison.ipynb b/notebooks/Sampling_comparison.ipynb
index c5d10b1..00aa09c 100644
--- a/notebooks/Sampling_comparison.ipynb
+++ b/notebooks/Sampling_comparison.ipynb
@@ -11,7 +11,7 @@
     "\n",
     "<table style=\"border-collapse: collapse; border: none;\">\n",
     "    <tr>\n",
-    "        <td style=\"border: none;\"><img src=\"../figures/figure3_histogram/Figure3a.png\" width=\"500\"></td>\n",
+    "        <td style=\"border: none;\"><img src=\"https://raw.githubusercontent.com/PolarWandering/PaleoSampling/57e98e6cc79638b28949893508981cbfcc3a1643/figures/figure3_histogram/Figure3a.png\" width=\"500\"></td>\n",
     "        <td style=\"border: none; vertical-align: bottom; width: 300px;\">\n",
     "            <b>Figure 3a:</b> Comparison between two different sampling strategies to determine a mean paleomagnetic pole position in the presence of outliers for a fixed number of total samples (n = 100). The red histograms and curve are strategy 1 where we have one sample per site (n0 = 1), one hundred sites (N = 100) and we use the Vandamme filter. The blue histograms and curve are strategy 2 where n0 = 5, (N = 20) and we filter all the outliers (perfect detection algorithm) for (a) p_outlier = 0.10.\n",
     "        </td>\n",