Examinando por Materia "Machine learning"

Mostrando 1 - 6 de 6

Digital soil mapping of metals and metalloids in croplands using multiple geospatial data and machine learning, implemented in GEE, for the Peruvian Mantaro Valley
(Elsevier, 2024-03-29) Pizarro Carcausto, Samuel; Vera Vilchez, Jesús; Huamani, Joseph; Cruz, Juancarlos; Lastra, Sphyros; Solórzano Acosta, Richard; Verástegui Martínez, Patricia
Quality and safety of the soil are essential to ensure social and economic development and provides the supply of contaminant free food. With agriculture intensification, expansion of urban zones, construction of roads, and mining, some agricultural soils sites become polluted increasing environmental risks to ecosystems functions and human health. Hence the need know the spatial distribution of elements in soils, we mapped 25 elements, namely Ca, Mg, Sr, Ba, Be, K, Na, As, Sb, Se, Tl, Cd, Zn, Al, Pb, Hg, Cr, Ni, Cu, Mo, Ag, Fe, Co, Mn and V, using various geospatial datasets, such as remote sensing, climate, topography, soil data, and distance, to establish the spatial estimation models of spatial distribution trained trough machine learning model with a supervised dataset of 109 topsoil samples, into Google earth engine platform. Using R2, RMSE and MAE to assess the prediction accuracy. First Random Forest gave satisfactory results in predicting the distribution of analyzed elements in soil, being improved for some elements when adds more trees. Additionally, each element analyzed has a different combination of environmental covariates as predictor, mainly soil, climate, topographic and distance variables especially croplands close to rivers, with less importance for spectral variables. Our results suggest that is possible to identify polluted soils and improved regulations to minimize harm to environmental health and human health, for short-to-medium-term environmental risk control.
Estimation of forage biomass in oat (Avena sativa) using agronomic variables through UAV multispectral imaging
(MDPI, 2024-10-06) Urquizo Barrera, Julio Cesar; Ccopi Trucios, Dennis; Ortega Quispe, Kevin; Castañeda Tinco, Italo; Patricio Rosales, Solanch; Passuni Huayta, Jorge; Figueroa Venegas, Deyanira; Enriquez Pinedo, Lucia; Ore Aquino, Zoila; Pizarro Carcausto, Samuel
Accurate and timely estimation of oat biomass is crucial for the development of sustainable and efficient agricultural practices. This research focused on estimating and predicting forage oat biomass using UAV and agronomic variables. A Matrice 300 equipped with a multispectral camera was used for 14 flights, capturing 21 spectral indices per flight. Concurrently, agronomic data were collected at six stages synchronized with UAV flights. Data analysis involved correlations and Principal Component Analysis (PCA) to identify significant variables. Predictive models for forage biomass were developed using various machine learning techniques: linear regression, Random Forests (RFs), Support Vector Machines (SVMs), and Neural Networks (NNs). The Random Forest model showed the best performance, with a coefficient of determination R2 of 0.52 on the test set, followed by Support Vector Machines with an R2 of 0.50. Differences in root mean square error (RMSE) and mean absolute error (MAE) among the models highlighted variations in prediction accuracy. This study underscores the effectiveness of photogrammetry, UAV, and machine learning in estimating forage biomass, demonstrating that the proposed approach can provide relatively accurate estimations for this purpose.
From rangelands to cropland, land-use change and its impact on soil organic carbon variables in a Peruvian Andean highlands: a machine learning modeling approach
(Springer, 2024-09-09) Carbajal, Mariella; Ramirez, David A.; Turin Canchaya, Cecilia Claudia; Schaeffer, Sean M.; Konkel, Julie; Ninanya, Johan; Rinza, Javier; De Mendiburu, Felipe; Zorogastua, Percy; Villaorduña, Liliana; Quiroz, Roberto
Andean highland soils contain significant quantities of soil organic carbon (SOC); however, more efforts still need to be made to understand the processes behind the accumulation and persistence of SOC and its fractions. This study modeled SOC variables—SOC, refractory SOC (RSOC), and the 13C isotope composition of SOC (d13CSOC)—using machine learning (ML) algorithms in the Central Andean Highlands of Peru, where grasslands and wetlands (‘‘bofedales’’) dominate the landscape surrounded by Junin National Reserve. A total of 198 soil samples (0.3 m depth) were collected to assess SOC variables. Four ML algorithms—random forest (RF), support vector machine (SVM), artificial neural networks (ANNs), and eXtreme gradient boosting (XGB)—were used to model SOC variablesusing remote sensing data, land-use and landcover (LULC, nine categories), climate topography, and sampled physical–chemical soil variables. RF was the best algorithm for SOC and d13CSOC prediction, whereas ANN was the best to model RSOC. ‘‘Bofedales’’ showed 2–3 times greater SOC (11.2 ± 1.60%) and RSOC (1.10 ± 0.23%) and more depleted d13CSOC (- 27.0 ± 0.44 &) than other LULC, which reflects high C persistent, turnover rates, and plant productivity. This highlights the importance of ‘‘bofedales’’ as SOC reservoirs. LULC and vegetation indices close to the near-infrared bands were the most critical environmental predictors to model C variables SOC and d13CSOC. In contrast, climatic indices were more important environmental predictors for RSOC. This study’s outcomes suggest the potential of ML methods, with a particular emphasis on RF, for mapping SOC and its fractions in the Andean highlands.
Implementing cloud computing for the digital mapping of agricultural soil properties from high resolution UAV multispectral imagery
(MDPI, 2023-06-20) Pizarro Carcausto, Samuel Edwin; Pricope, Narcisa G.; Figueroa Venegas, Deyanira Antonella; Carbajal Llosa, Carlos Miguel; Quispe Huincho, Miriam Rocío; Vera Vilchez, Jesús Emilio; Alejandro Méndez, Lidiana Rene; Achallma Mendoza, Lino; González Tovar, Izamar Estrella; Salazar Coronel, Wilian; Loayza, Hildo; Cruz Luis, Juancarlos Alejandro; Arbizu Berrocal, Carlos Irvin
The spatial heterogeneity of soil properties has a significant impact on crop growth, making it difficult to adopt site-specific crop management practices. Traditional laboratory-based analyses are costly, and data extrapolation for mapping soil properties using high-resolution imagery becomes a computationally expensive procedure, taking days or weeks to obtain accurate results using a desktop workstation. To overcome these challenges, cloud-based solutions such as Google Earth Engine (GEE) have been used to analyze complex data with machine learning algorithms. In this study, we explored the feasibility of designing and implementing a digital soil mapping approach in the GEE platform using high-resolution reflectance imagery derived from a thermal infrared and multispectral camera Altum (MicaSense, Seattle, WA, USA). We compared a suite of multispectral-derived soil and vegetation indices with in situ measurements of physical-chemical soil properties in agricultural lands in the Peruvian Mantaro Valley. The prediction ability of several machine learning algorithms (CART, XGBoost, and Random Forest) was evaluated using R2, to select the best predicted maps (R2 > 0.80), for ten soil properties, including Lime, Clay, Sand, N, P, K, OM, Al, EC, and pH, using multispectral imagery and derived products such as spectral indices and a digital surface model (DSM). Our results indicate that the predictions based on spectral indices, most notably, SRI, GNDWI, NDWI, and ExG, in combination with CART and RF algorithms are superior to those based on individual spectral bands. Additionally, the DSM improves the model prediction accuracy, especially for K and Al. We demonstrate that high-resolution multispectral imagery processed in the GEE platform has the potential to develop soil properties prediction models essential in establishing adaptive soil monitoring programs for agricultural regions.
Landsat images and GIS techniques as key tools for historical analysis of landscape change and fragmentation
(Elsevier, 2024-07-28) Gómez Fernández, Darwin; Salas López, Rolando; Zabaleta Santisteban, Jhon A.; Medina Medina, Angel J.; Goñas Goñas, Malluri; Silva López, Jhonsy O.; Oliva Cruz, Manuel; Rojas Briceño, Nilton B.
Monitoring and evaluation of landscape fragmentation is important in numerous research areas, such as natural resource protection and management, sustainable development, and climate change. One of the main challenges in image classification is the intricate selection of parameters, as the optimal combination significantly affects the accuracy and reliability of the final results. This research aimed to analyze landscape change and fragmentation in northwestern Peru. We utilized accurate land cover and land use (LULC) maps derived from Landsat imagery using Google Earth Engine (GEE) and ArcGIS software. For this, we identified the best dataset based on its highest overall accuracy, and kappa index; then we performed an analysis of variance (ANOVA) to assess the differences in accuracies among the datasets, finally, we obtained the LULC and fragmentation maps and analyzed them. We generated 31 datasets resulting from the combination of spectral bands, indices of vegetation, water, soil and clusters. Our analysis revealed that dataset 19, incorporating spectral bands along with water and soil indices, emerged as the optimal choice. Regarding the number of trees utilized in classification, we determined that using between 10 and 400 decision trees in Random Forest classification doesn't significantly affect overall accuracy or the Kappa index, but we observed a slight cumulative increase in accuracy metrics when using 100 decision trees. Additionally, between 1989 and 2023, the categories Artificial surfaces, Agricultural areas, and Scrub/ Herbaceous vegetation exhibit a positive rate of change, while the categories Forest and Open spaces with little or no vegetation display a decreasing trend. Consequently, the areas of patches and perforated have expanded in terms of area units, contributing to a reduction in forested areas (Core 3) due to fragmentation. As a result, forested areas smaller than 500 acres (Core 1 and 2) have increased. Finally, our research provides a methodological framework for image classification and assessment of landscape change and fragmentation, crucial information for decision makers in a current agricultural zone of northwestern Peru.
Mango varietal discrimination using hyperspectral imaging and machine learning
(Springer Nature, 2024-07-29) Castro, Wilson; Tene, Baldemar; Castro, Jorge; Guivin, Alex; Ruesta Campoverde, Nelson Asdrubal; Avila George, Himer
Mango is a highly diverse tropical fruit with numerous varieties that differ in flavor, texture, and chemical composition. Consequently, identifying fraudulent substitutions of mango varieties poses a significant challenge using traditional techniques. Therefore, there is an increasing need for new methods to discriminate between mango varieties. Hyperspectral imaging coupled with machine learning techniques presents a promising approach for varietal discrimination. In this study, mango samples of eleven varieties were collected from a germplasm bank, with four slices obtained from each sample. Hyperspectral images were acquired in the Vis–NIR and NIR ranges for each slice, and spectral profiles were extracted and pretreated. Three discrimination models, linear discriminant analysis, K-nearest neighbor, and artificial neural networks, were implemented and validated using relevant wavelengths selected through a covering array feature selection algorithm. The performance of these models was evaluated using precision, accuracy, and F-score metrics. The average spectral profiles of the studied varieties exhibited a similar behavior with slight differences, which could be used for classification within the evaluated ranges. The optimal number of variables selected to refine the models was 17 for the UV–Vis–NIR range and 21 for the NIR range, with an accuracy ranging between 0.752 and 0.972. This study concludes that hyperspectral imaging combined with machine learning techniques can effectively discriminate between different varieties of mango.