Efficient sampling for predictive nutrient mapping in farm-scale crop management

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00027049%3A_____%2F24%3AN0000099" target="_blank" >RIV/00027049:_____/24:N0000099 - isvavai.cz</a>

Výsledek na webu

—

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Efficient sampling for predictive nutrient mapping in farm-scale crop management

Popis výsledku v původním jazyce

Plant nutrition and balanced fertilization require very accurate information from the field at the lowest possible cost and therefore limited number of samples. It is therefore essential to optimise sampling schemes. The aim of this study is to compare widely used sampling schemes combined with variable sample size for the prediction of common soil macronutrients. Conditional Latin hypercube sampling (cLHS), feature space coverage sampling with k-means (FSCS) and simple random sampling (SRS) were compared. The effect of sampling scheme and sample size on the accuracy of predicted nutrient maps was investigated in a real case study field (35 ha) with heterogeneous soil properties. A total of 200 training points were placed in 6 grids: cLHS and FSCS with 10, 30 and 60 samples, corresponding to 1 sample per 3, 1 and 0.5 ha, respectively. For the numerical experiment with different sampling frequencies, all 200 training samples were interpolated into a set of nutrient maps, which were considered as an error-free dataset for both calibration and validation samples included in the predictive modelling. Sampling grids with variable sample sizes from 2 to 60 were created using SRS, cLHS and FSCS in combination with a pragmatic set of environmental covariates. Each network of each method was automatically generated 100 times using the same algorithm settings. These were used to make predictions using the covariates. The performance of the models was monitored. The results show the advantage of using FSCS, which shows both less variation in prediction accuracy compared to SRS and cLHS and better results with sparse sampling.

Název v anglickém jazyce

Efficient sampling for predictive nutrient mapping in farm-scale crop management

Popis výsledku anglicky

Plant nutrition and balanced fertilization require very accurate information from the field at the lowest possible cost and therefore limited number of samples. It is therefore essential to optimise sampling schemes. The aim of this study is to compare widely used sampling schemes combined with variable sample size for the prediction of common soil macronutrients. Conditional Latin hypercube sampling (cLHS), feature space coverage sampling with k-means (FSCS) and simple random sampling (SRS) were compared. The effect of sampling scheme and sample size on the accuracy of predicted nutrient maps was investigated in a real case study field (35 ha) with heterogeneous soil properties. A total of 200 training points were placed in 6 grids: cLHS and FSCS with 10, 30 and 60 samples, corresponding to 1 sample per 3, 1 and 0.5 ha, respectively. For the numerical experiment with different sampling frequencies, all 200 training samples were interpolated into a set of nutrient maps, which were considered as an error-free dataset for both calibration and validation samples included in the predictive modelling. Sampling grids with variable sample sizes from 2 to 60 were created using SRS, cLHS and FSCS in combination with a pragmatic set of environmental covariates. Each network of each method was automatically generated 100 times using the same algorithm settings. These were used to make predictions using the covariates. The performance of the models was monitored. The results show the advantage of using FSCS, which shows both less variation in prediction accuracy compared to SRS and cLHS and better results with sparse sampling.

Druh

O - Ostatní výsledky

CEP obor

—

OECD FORD obor

40104 - Soil science

Projekt

<a href="/cs/project/QK21010247" target="_blank" >QK21010247: Optimalizace hospodaření na nevyrovnaných pozemcích využitím efektivního mapování půdních podmínek a zohlednění změn vláhových poměrů s cílem stabilizace dosahovaných výnosových úrovní</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)<br>I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Rok uplatnění

2024

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů