Many soil samples contain coarse fragments, which are pieces of rock or other material that are larger than 2 mm in diameter. The proportion of coarse fragments in a soil sample is used in the calculation of the stock of soil organic carbon in the sample. Data on the proportion of coarse fragments in many soil samples is missing. This issue is to develop a method to estimate the proportion of coarse fragments in a soil sample.
We will first develop a random forest model to classify soil samples as having or not having coarse fragments. Next we will develop a regression model to estimate the proportion of coarse fragments in soil samples that are classified as having coarse fragments. The second model will be trained on soil samples that are known as having coarse fragments. We expect that this strategy will avoid estimating small amounts of coarse fragments for samples that do not have any.
Many soil samples contain coarse fragments, which are pieces of rock or other material that are larger than 2 mm in diameter. The proportion of coarse fragments in a soil sample is used in the calculation of the stock of soil organic carbon in the sample. Data on the proportion of coarse fragments in many soil samples is missing. This issue is to develop a method to estimate the proportion of coarse fragments in a soil sample.
We will first develop a random forest model to classify soil samples as having or not having coarse fragments. Next we will develop a regression model to estimate the proportion of coarse fragments in soil samples that are classified as having coarse fragments. The second model will be trained on soil samples that are known as having coarse fragments. We expect that this strategy will avoid estimating small amounts of coarse fragments for samples that do not have any.