{"id":581255,"date":"2026-01-16T16:10:00","date_gmt":"2026-01-16T16:10:00","guid":{"rendered":"https:\/\/www.esri.com\/en-us\/industries\/blog\/?post_type=blog&#038;p=581255"},"modified":"2026-01-20T16:07:04","modified_gmt":"2026-01-20T16:07:04","slug":"deep-learning-assessment-in-the-santa-ana-river-watershed","status":"publish","type":"blog","link":"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed","title":{"rendered":"Deep Learning Arundo donax (Arundo) Assessment in the Santa Ana River Watershed"},"content":{"rendered":"<p><em>Authored by: Peter Vitt, Senior GIS Project Manager, Santa Ana Watershed Project Authority<\/em><\/p>\n\n<p>The Santa Ana Watershed Project Authority (SAWPA) was formed in 1975 as a Joint Powers Authority under California law. SAWPA\u2019s mission is to develop and maintain regional plans, programs, and projects that protect the Santa Ana River (SAR) Watershed while maximizing the beneficial uses of its water resources in an economically and environmentally responsible manner.<\/p>\n\n<p>A longtime threat to the SAR Watershed\u2019s water resources is the invasive plant species <em>Arundo<\/em>. <em>Arundo<\/em> is an invasive grass species pervasive in the stream channels of the SAR Watershed. It was likely introduced to the watershed for erosion control along irrigation channels. <em>Arundo<\/em> is problematic because it uses a lot of water, replaces native habitats, and presents a fire hazard due to its very dense growth pattern.<\/p>\n\n<p>A study by the California Invasive Plant Council (Cal-IPC) in 2011 estimated approximately 1,400 acres of <em>Arundo<\/em> within the watershed. Past and present <em>Arundo<\/em> removal efforts conducted by Resource Conservation Districts and other agencies have removed significant amounts of the plant from the watershed. But because of the dense root structure and fast-growing nature of the plant, it is difficult to eradicate. Since it had been nearly 15 years since the last <em>Arundo<\/em> estimate was conducted, SAWPA wanted to do a watershed-wide assessment to evaluate the current magnitude of the problem and get an idea of how effective ongoing removal efforts have been in reducing its presence.<\/p>\n\n<h2 class=\"wp-block-heading\" id=\"h-challenge\">Challenge<\/h2>\n\n<p>The watershed is very large and encompasses approximately 2,800 square miles across four counties in Southern California, including western Riverside and San Bernardino Counties, northern Orange County, and a slice of southeastern Los Angeles County. Because of the watershed\u2019s large size, SAWPA did not have the time or resources to do field investigations, and digitizing <em>Arundo<\/em> plants from aerial imagery across such a large area would not have been accurate or timely.<\/p>\n\n<h2 class=\"wp-block-heading\" id=\"h-solution\">Solution<\/h2>\n\n<p>The solution was to employ Esri\u2019s deep learning tools in ArcGIS Pro to identify <em>Arundo<\/em> on imagery and limit the assessment area to places where <em>Arundo<\/em> was likely present.<\/p>\n\n<p><strong>Identify Areas for Analysis<\/strong><\/p>\n\n<p>Areas for analysis included those where <em>Arundo<\/em> was most likely to be found and where it has historically been found in the watershed. These included:<\/p>\n\n<ul class=\"wp-block-list\">\n<li>Stream channels.<\/li>\n\n<li>Areas where <em>Arundo<\/em> was previously identified and\/or removed by Resource Conservation Districts or other agencies.<\/li>\n\n<li>Areas where significant <em>Arundo<\/em> was identified in the Cal-IPC 2011 Study.<\/li>\n\n<li>Groundwater recharge basins.<\/li>\n<\/ul>\n\n<p>We used the Santa Ana River Regional Water Quality Control Board\u2019s Basin Plan to identify major stream channels and identified smaller stream channels from aerial imagery and the USGS National Hydrography Dataset flowline GIS layer. The Santa Ana Watershed Association (SAWA) provided GIS layers of <em>Arundo<\/em> removal sites, and SAWPA member agencies provided GIS layers for groundwater recharge basins.<\/p>\n\n<p>The project was divided into five phases based upon geography and area type. In addition to deep learning analysis, we conducted a visual inspection of the imagery for <em>Arundo<\/em> along concrete-lined stream channels and dry streambeds with little or no vegetation.<\/p>\n\n<div class=\"wp-block-image\">\n<figure class=\"aligncenter size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"408\" src=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-content\/uploads\/2026\/01\/Deep-Learning-Analysis-1024x408.jpg\" alt=\"\" class=\"wp-image-581256\" srcset=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/Deep-Learning-Analysis-1024x408.jpg 1024w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/Deep-Learning-Analysis-300x120.jpg 300w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/Deep-Learning-Analysis-768x306.jpg 768w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/Deep-Learning-Analysis-1536x612.jpg 1536w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/Deep-Learning-Analysis.jpg 1876w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption class=\"wp-element-caption\">Left: Map showing where <em>Arundo<\/em> analysis was conducted within the watershed. Right: Map showing project phases.<\/figcaption><\/figure>\n<\/div>\n\n<p><strong>Imagery, Hardware, and Software<\/strong><\/p>\n\n<p>We had three-inch resolution four-band aerial imagery for the watershed from a previous landscape analysis project we did with the US Bureau of Reclamation (USBR). This imagery consisted of more than 8,500 tiled GeoTIFFs collected in 2020 and 2021.<\/p>\n\n<p>Initially, hardware consisted of a Dell workstation with an Intel<sup>\u00d2<\/sup> Xeon<sup>\u00d2<\/sup> &nbsp;W-2225 CPU @ 4.10 GHz processor, 64 GB of RAM, and a 5 GB graphics card. We did some deep learning processing tests to get an idea of how well the hardware configuration would perform and found that this configuration was too slow. We increased the RAM to 256 GB, and most importantly, upgraded the graphics card to an NVIDIA GeForce RTX 4090 with 24 GB of memory. After switching to the new hardware configuration, we found that a process that might have taken 10 to 12 hours under the former configuration was reduced to one to two hours using the new hardware configuration.<\/p>\n\n<p>The software used was ArcGIS Pro, the ArcGIS Pro Image Analyst extension, and ArcGIS Pro deep learning framework library installers.<\/p>\n\n<div class=\"wp-block-image\">\n<figure class=\"aligncenter size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"791\" src=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-content\/uploads\/2026\/01\/4_Map_ImageryTiles-1024x791.jpg\" alt=\"\" class=\"wp-image-581257\" srcset=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/4_Map_ImageryTiles-1024x791.jpg 1024w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/4_Map_ImageryTiles-300x232.jpg 300w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/4_Map_ImageryTiles-768x593.jpg 768w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/4_Map_ImageryTiles-1536x1187.jpg 1536w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/4_Map_ImageryTiles-2048x1583.jpg 2048w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption class=\"wp-element-caption\"><em>Map showing the extent of SAWPA imagery and the tiling scheme for GeoTIFFs.<\/em><\/figcaption><\/figure>\n<\/div>\n\n<p><strong>Pre-Arundo Analysis Tasks<\/strong><\/p>\n\n<p>There were a few tasks we needed to complete before beginning the deep learning <em>Arundo<\/em> analysis. The first was a deep learning model type assessment to see which model types would perform best at identifying <em>Arundo<\/em>. The model type is essentially the algorithm the software uses to train your model. We tested four model types: U-net, DeepLab, PSPNet, and MMSegmentation within a four-mile square area and ran QC checks to see which model types worked best. Results of the QC tests indicated that U-net and DeepLab worked the best at identifying <em>Arundo<\/em>. We created a total of seven models for identifying <em>Arundo<\/em> in the watershed\u2014with most of them using the U-net model type. The second pre-analysis task was to break the large analysis area into smaller, more manageable area of interest (AOI) units. AOIs were built based on location, similar ground conditions, and similar imagery collection dates. AOIs were grouped this way because the <em>Arundo<\/em> could look different depending on lighting, ground conditions, atmospheric conditions, and the amount of water in the vegetation. The final pre-analysis task was to merge imagery tiles together for each AOI.<\/p>\n\n<div class=\"wp-block-image\">\n<figure class=\"aligncenter size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"407\" src=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-content\/uploads\/2026\/01\/5_IdentificationProceduresOverview-1024x407.jpg\" alt=\"\" class=\"wp-image-581258\" srcset=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/5_IdentificationProceduresOverview-1024x407.jpg 1024w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/5_IdentificationProceduresOverview-300x119.jpg 300w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/5_IdentificationProceduresOverview-768x305.jpg 768w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/5_IdentificationProceduresOverview.jpg 1340w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption class=\"wp-element-caption\"><em>Figure outlining the basic steps of the deep learning process.<\/em><\/figcaption><\/figure>\n<\/div>\n\n<p><strong>Arundo Analysis<\/strong><\/p>\n\n<p>Our steps for identifying <em>Arundo<\/em> in imagery using deep learning are outlined below:<\/p>\n\n<ol class=\"wp-block-list\">\n<li>Training samples\u2014used the \u201cLabel Objects for Deep Learning\u201d tool to draw training samples on top of the imagery. It was especially important that we collected different examples of <em>Arundo<\/em> and that we were very precise when we drew samples. We created a total of seven models for the watershed, with each model having between 100 to 250 training samples. We used the \u201cExport Training Data for Deep Learning\u201d tool to export image chips for model training. We consulted with Orange County Water District and SAWA environmental staff to help us identify plants that were not obviously <em>Arundo<\/em>, as both agencies have extensive experience with the species.<\/li>\n<\/ol>\n\n<div class=\"wp-block-image\">\n<figure class=\"aligncenter size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"600\" src=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-content\/uploads\/2026\/01\/6_ToolLabelObjects-1024x600.jpg\" alt=\"\" class=\"wp-image-581259\" srcset=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/6_ToolLabelObjects-1024x600.jpg 1024w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/6_ToolLabelObjects-300x176.jpg 300w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/6_ToolLabelObjects-768x450.jpg 768w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/6_ToolLabelObjects.jpg 1176w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption class=\"wp-element-caption\">Screenshot of the \u201cLabel Objects for Deep Learning\u201d tool.<\/figcaption><\/figure>\n<\/div>\n\n<p>2. Train model\u2014used the \u201cTrain Deep Learning Model\u201d tool to train the model. After the model was trained, we examined the model metrics report to see how well it performed. We checked the train and validation curves to make sure they were close, and examined the F1 statistic to evaluate overall model performance. Our target for an acceptable F1 statistic was greater than 0.75. If the F1 statistic was below that score, we would try a different model type or revise the training samples and retrain the model. It often took four or five rounds of trying a different model type and revising the training samples to get the model right and to get as close to the target as possible.<\/p>\n\n<div class=\"wp-block-image\">\n<figure class=\"aligncenter size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"631\" src=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-content\/uploads\/2026\/01\/Unet-Classifier-1024x631.jpg\" alt=\"\" class=\"wp-image-581260\" srcset=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/Unet-Classifier-1024x631.jpg 1024w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/Unet-Classifier-300x185.jpg 300w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/Unet-Classifier-768x473.jpg 768w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/Unet-Classifier.jpg 1294w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption class=\"wp-element-caption\"><em>Screenshot of the model metrics report generated by the \u201cTrain Deep Learning Model\u201d tool.<\/em><\/figcaption><\/figure>\n<\/div>\n\n<p>3. Run the model\u2014used the ArcGIS Pro \u201cClassify Pixels Using Deep Learning\u201d tool to run the model on the AOI imagery to produce a classified raster.<\/p>\n\n<p>4. QC checks\u2014the first step in the QC process was to convert the classified raster into a polygon layer and do a visual inspection to assess the overall accuracy. We did a manual cleanup on the polygon layer to remove any obviously misclassified areas. After the manual cleanup, we used the \u201cCreate Accuracy Assessment Points\u201d tool to generate random locations to check the classification visually against the imagery. We used the \u201cCompute Confusion Matrix\u201d tool to generate accuracy statistics for the classification. Our target accuracy statistics were 80 percent for <em>Arundo<\/em> prediction and 90 percent for the model (average of both classes). Accuracy was calculated as correct predictions divided by total predictions.<\/p>\n\n<div class=\"wp-block-image\">\n<figure class=\"aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"744\" height=\"389\" src=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-content\/uploads\/2026\/01\/9_QC_ConfusionMatrix.jpg\" alt=\"\" class=\"wp-image-581261\" srcset=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/9_QC_ConfusionMatrix.jpg 744w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/9_QC_ConfusionMatrix-300x157.jpg 300w\" sizes=\"auto, (max-width: 744px) 100vw, 744px\" \/><figcaption class=\"wp-element-caption\"><em>Screenshot of confusion matrix output from \u201cCreate Confusion Matrix\u201d tool (imported to Excel).<\/em><\/figcaption><\/figure>\n<\/div>\n\n<p>5. Map&#8211;created maps of detected <em>Arundo <\/em>and generated statistics of the amount detected in the watershed. <\/p>\n\n<div class=\"wp-block-image\">\n<figure class=\"aligncenter size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"791\" src=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-content\/uploads\/2026\/01\/10_Map_DetectedArundo-1024x791.jpg\" alt=\"\" class=\"wp-image-581262\" srcset=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/10_Map_DetectedArundo-1024x791.jpg 1024w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/10_Map_DetectedArundo-300x232.jpg 300w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/10_Map_DetectedArundo-768x593.jpg 768w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/10_Map_DetectedArundo-1536x1187.jpg 1536w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/10_Map_DetectedArundo-2048x1583.jpg 2048w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption class=\"wp-element-caption\"><em>Map showing Arundo detected in the watershed using deep learning analysis.<\/em><\/figcaption><\/figure>\n<\/div>\n\n<div class=\"wp-block-image\">\n<figure class=\"aligncenter size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"546\" src=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-content\/uploads\/2026\/01\/11_ArundoDetectedDeepLearning-1024x546.jpg\" alt=\"\" class=\"wp-image-581263\" srcset=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/11_ArundoDetectedDeepLearning-1024x546.jpg 1024w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/11_ArundoDetectedDeepLearning-300x160.jpg 300w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/11_ArundoDetectedDeepLearning-768x410.jpg 768w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/11_ArundoDetectedDeepLearning.jpg 1532w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption class=\"wp-element-caption\"><em>Screenshot of large Arundo stands detected in the Santa Ana River by deep learning.<\/em><\/figcaption><\/figure>\n<\/div>\n\n<h2 class=\"wp-block-heading\" id=\"h-results\">Results<\/h2>\n\n<p><strong>Project Statistics<\/strong><\/p>\n\n<p>The table below summarizes statistics for each phase of the project. Across the entire watershed, we analyzed more than 56,000 acres and detected just under 286 acres of <em>Arundo<\/em>. Most of the <em>Arundo<\/em> (84 percent) was detected in Phase 1 along the Santa Ana River between Prado Dam and La Cadena Avenue in Colton. We also exceeded our accuracy targets, with <em>Arundo<\/em> prediction at 88 percent and model accuracy (both classes) at just under 94 percent. The project was time-consuming, taking around 500 hours of staff time and more than 400 hours of machine time.<\/p>\n\n<div class=\"wp-block-image\">\n<figure class=\"aligncenter size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"514\" src=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-content\/uploads\/2026\/01\/Table-of-Phases-1024x514.jpg\" alt=\"\" class=\"wp-image-581266\" srcset=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/Table-of-Phases-1024x514.jpg 1024w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/Table-of-Phases-300x151.jpg 300w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/Table-of-Phases-768x386.jpg 768w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/Table-of-Phases-1536x772.jpg 1536w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/Table-of-Phases.jpg 1662w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption class=\"wp-element-caption\">*Weighted to <em>Arundo<\/em> occurrence.<\/figcaption><\/figure>\n<\/div>\n\n<p><strong>Project Takeaways<\/strong><\/p>\n\n<p>Deep learning effectively identified <em>Arundo<\/em> occurrence, providing a solid estimate of its presence in the watershed. The <em>Arundo<\/em> predictive accuracy was 88 percent, and the model accuracy was almost 94 percent over the entire watershed area. The amount of <em>Arundo<\/em> detected was significantly less than the Cal-IPC estimate 15 years ago, indicating that <em>Arundo<\/em> removal efforts have been effective. We might have underestimated <em>Arundo<\/em> somewhat in the drier parts of the upper watershed, as field investigations conducted after the analysis found <em>Arundo<\/em> in some areas where the deep learning analysis didn\u2019t. This could have occurred because different vegetation types were harder to distinguish in drier areas as compared to wetter areas, and some of the imagery in the upper watershed near the mountains was a little blurry.<\/p>\n\n<p>In terms of the amount of <em>Arundo<\/em> detected in the watershed, it is also worth noting that many past <em>Arundo<\/em> removal projects have estimated the amount of <em>Arundo<\/em> removed as the area cleared of <em>Arundo<\/em> rather than the actual amount of <em>Arundo<\/em> plant removed. So sometimes the amount of <em>Arundo<\/em> in an area could be exaggerated.<\/p>\n\n<p>Another big takeaway from the project was the importance of having a powerful graphics card for conducting deep learning analysis. When we upgraded from 5 GB to the 24 GB card, the processing speed increased by six to ten times. Processes that previously took 10 to 12 hours could be accomplished in one to two hours with the upgraded card.<\/p>\n\n<p>While the project was time-consuming\u2014taking more than 500 hours of staff time\u2014there is no way we could have manually digitized <em>Arundo<\/em> within our 56,000-acre study area. The degree of accuracy would certainly have been less than what was achieved from deep learning computer analysis.<\/p>\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"27\" src=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-content\/uploads\/2026\/01\/End-of-Story-Image-1024x27.jpg\" alt=\"\" class=\"wp-image-580920\" srcset=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/End-of-Story-Image-1024x27.jpg 1024w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/End-of-Story-Image-300x8.jpg 300w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/End-of-Story-Image-768x20.jpg 768w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/End-of-Story-Image-1536x41.jpg 1536w, https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/End-of-Story-Image.jpg 1920w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n<h2 class=\"has-text-align-center wp-block-heading\">Stay Connected with Esri&#8217;s Water Team<\/h2>\n\n<p class=\"has-text-align-center\">ArcGIS is an extensive information system that enables modernization of workflows with easy-to-use applications. Strengthen your organization with geospatial solutions that will increase efficiency and provide insight for decision-makers. Visit the&nbsp;<a href=\"https:\/\/www.esri.com\/en-us\/industries\/water\" target=\"_blank\" rel=\"noreferrer noopener\">Esri Water<\/a>&nbsp;web site for more information.<\/p>\n\n<p class=\"has-text-align-center\">Follow #EsriWater on social media:&nbsp;<a href=\"https:\/\/twitter.com\/EsriWater\" target=\"_blank\" rel=\"noreferrer noopener\">X<\/a>&nbsp;|&nbsp;<a href=\"https:\/\/www.linkedin.com\/groups\/6533227\" target=\"_blank\" rel=\"noreferrer noopener\">LinkedIn<\/a><\/p>\n\n<p class=\"has-text-align-center\">Subscribe to the Water Industry newsletter \u201c<a href=\"https:\/\/www.esri.com\/en-us\/industries\/water\/water-news\" target=\"_blank\" rel=\"noreferrer noopener\">Esri News for Water Utilities and Water Resources<\/a>\u201c<\/p>","protected":false},"author":131,"featured_media":0,"parent":0,"menu_order":0,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":"","_links_to":"","_links_to_target":""},"categories":[3872],"tags":[3922,451,6509,841,6511],"class_list":["post-581255","blog","type-blog","status-publish","format-standard","hentry","category-water-resources","tag-analysis","tag-arcgis","tag-deep-learning","tag-environment","tag-watershed","industry-water-resources"],"acf":[],"yoast_head":"<!-- This site is optimized with the Yoast SEO Premium plugin v25.9 (Yoast SEO v25.9) - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>Assessing Arundo in the Santa Ana River Watershed<\/title>\n<meta name=\"description\" content=\"Deep learning identified Arundo in a 56,000-acre study area within the Santa Ana River Watershed, saving hundreds of staff hours.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Assessing Arundo in the Santa Ana River Watershed\" \/>\n<meta property=\"og:description\" content=\"Deep learning identified Arundo in a 56,000-acre study area within the Santa Ana River Watershed, saving hundreds of staff hours.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed\" \/>\n<meta property=\"og:site_name\" content=\"Industry Blogs\" \/>\n<meta property=\"article:modified_time\" content=\"2026-01-20T16:07:04+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-content\/uploads\/2026\/01\/3_Map_ProjectPhases-scaled.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"2560\" \/>\n\t<meta property=\"og:image:height\" content=\"1978\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:title\" content=\"Assessing Arundo in the Santa Ana River Watershed\" \/>\n<meta name=\"twitter:description\" content=\"Deep learning identified Arundo in a 56,000-acre study area within the Santa Ana River Watershed, saving hundreds of staff hours.\" \/>\n<meta name=\"twitter:label1\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data1\" content=\"11 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed\",\"url\":\"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed\",\"name\":\"Assessing Arundo in the Santa Ana River Watershed\",\"isPartOf\":{\"@id\":\"https:\/\/www.esri.com\/en-us\/industries\/blog\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed#primaryimage\"},\"image\":{\"@id\":\"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed#primaryimage\"},\"thumbnailUrl\":\"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-content\/uploads\/2026\/01\/Deep-Learning-Analysis-1024x408.jpg\",\"datePublished\":\"2026-01-16T16:10:00+00:00\",\"dateModified\":\"2026-01-20T16:07:04+00:00\",\"description\":\"Deep learning identified Arundo in a 56,000-acre study area within the Santa Ana River Watershed, saving hundreds of staff hours.\",\"breadcrumb\":{\"@id\":\"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed#primaryimage\",\"url\":\"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/Deep-Learning-Analysis.jpg\",\"contentUrl\":\"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/Deep-Learning-Analysis.jpg\",\"width\":1876,\"height\":748},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\/\/www.esri.com\/en-us\/industries\/blog\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Deep Learning Arundo donax (Arundo) Assessment in the Santa Ana River Watershed\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/www.esri.com\/en-us\/industries\/blog\/#website\",\"url\":\"https:\/\/www.esri.com\/en-us\/industries\/blog\/\",\"name\":\"Industry Blogs\",\"description\":\"\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/www.esri.com\/en-us\/industries\/blog\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/www.esri.com\/en-us\/industries\/blog\/#\/schema\/person\/63b299b98844293e8c561496264c0a03\",\"name\":\"Christa Campbell\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\/\/www.esri.com\/en-us\/industries\/blog\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2021\/01\/ChristaCampbell_2020-150x150.jpg\",\"contentUrl\":\"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2021\/01\/ChristaCampbell_2020-150x150.jpg\",\"caption\":\"Christa Campbell\"},\"description\":\"Christa is an experienced water industry professional with 20 years of success using and promoting technology to solve problems in the water industry. Recognized for Industry thought leadership, strategic thinking, and building success with organizations across the globe. She is a passionate GIS advocate, lifelong learner and collaborator. Builds success jointly with teammates, peers, and customers. Christa is a certified GIS Professional and holds a graduate degree in Geography.\",\"sameAs\":[\"https:\/\/www.esri.com\/en-us\/industries\/water\/overview\",\"https:\/\/www.linkedin.com\/in\/crstacampbell\/\",\"https:\/\/x.com\/CrstaCampbell\"],\"gender\":\"Female\",\"jobTitle\":\"Director Industry Solutions, Water\",\"worksFor\":\"Esri\",\"url\":\"\"}]}<\/script>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"Assessing Arundo in the Santa Ana River Watershed","description":"Deep learning identified Arundo in a 56,000-acre study area within the Santa Ana River Watershed, saving hundreds of staff hours.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed","og_locale":"en_US","og_type":"article","og_title":"Assessing Arundo in the Santa Ana River Watershed","og_description":"Deep learning identified Arundo in a 56,000-acre study area within the Santa Ana River Watershed, saving hundreds of staff hours.","og_url":"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed","og_site_name":"Industry Blogs","article_modified_time":"2026-01-20T16:07:04+00:00","og_image":[{"width":2560,"height":1978,"url":"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-content\/uploads\/2026\/01\/3_Map_ProjectPhases-scaled.jpg","type":"image\/jpeg"}],"twitter_card":"summary_large_image","twitter_title":"Assessing Arundo in the Santa Ana River Watershed","twitter_description":"Deep learning identified Arundo in a 56,000-acre study area within the Santa Ana River Watershed, saving hundreds of staff hours.","twitter_misc":{"Est. reading time":"11 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed","url":"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed","name":"Assessing Arundo in the Santa Ana River Watershed","isPartOf":{"@id":"https:\/\/www.esri.com\/en-us\/industries\/blog\/#website"},"primaryImageOfPage":{"@id":"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed#primaryimage"},"image":{"@id":"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed#primaryimage"},"thumbnailUrl":"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-content\/uploads\/2026\/01\/Deep-Learning-Analysis-1024x408.jpg","datePublished":"2026-01-16T16:10:00+00:00","dateModified":"2026-01-20T16:07:04+00:00","description":"Deep learning identified Arundo in a 56,000-acre study area within the Santa Ana River Watershed, saving hundreds of staff hours.","breadcrumb":{"@id":"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed#primaryimage","url":"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/Deep-Learning-Analysis.jpg","contentUrl":"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/Deep-Learning-Analysis.jpg","width":1876,"height":748},{"@type":"BreadcrumbList","@id":"https:\/\/www.esri.com\/en-us\/industries\/blog\/articles\/deep-learning-assessment-in-the-santa-ana-river-watershed#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/www.esri.com\/en-us\/industries\/blog"},{"@type":"ListItem","position":2,"name":"Deep Learning Arundo donax (Arundo) Assessment in the Santa Ana River Watershed"}]},{"@type":"WebSite","@id":"https:\/\/www.esri.com\/en-us\/industries\/blog\/#website","url":"https:\/\/www.esri.com\/en-us\/industries\/blog\/","name":"Industry Blogs","description":"","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.esri.com\/en-us\/industries\/blog\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Person","@id":"https:\/\/www.esri.com\/en-us\/industries\/blog\/#\/schema\/person\/63b299b98844293e8c561496264c0a03","name":"Christa Campbell","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.esri.com\/en-us\/industries\/blog\/#\/schema\/person\/image\/","url":"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2021\/01\/ChristaCampbell_2020-150x150.jpg","contentUrl":"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2021\/01\/ChristaCampbell_2020-150x150.jpg","caption":"Christa Campbell"},"description":"Christa is an experienced water industry professional with 20 years of success using and promoting technology to solve problems in the water industry. Recognized for Industry thought leadership, strategic thinking, and building success with organizations across the globe. She is a passionate GIS advocate, lifelong learner and collaborator. Builds success jointly with teammates, peers, and customers. Christa is a certified GIS Professional and holds a graduate degree in Geography.","sameAs":["https:\/\/www.esri.com\/en-us\/industries\/water\/overview","https:\/\/www.linkedin.com\/in\/crstacampbell\/","https:\/\/x.com\/CrstaCampbell"],"gender":"Female","jobTitle":"Director Industry Solutions, Water","worksFor":"Esri","url":""}]}},"text_date":"January 16, 2026","author_name":"Esri Water","author_page":false,"custom_image":"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/SAWPA_Banner.jpg","primary_product":"Water Resources","tag_data":[{"term_id":3922,"name":"Analysis","slug":"analysis","term_group":0,"term_taxonomy_id":3922,"taxonomy":"post_tag","description":"","parent":0,"count":2,"filter":"raw"},{"term_id":451,"name":"ArcGIS","slug":"arcgis","term_group":0,"term_taxonomy_id":451,"taxonomy":"post_tag","description":"","parent":0,"count":80,"filter":"raw"},{"term_id":6509,"name":"Deep learning","slug":"deep-learning","term_group":0,"term_taxonomy_id":6509,"taxonomy":"post_tag","description":"","parent":0,"count":1,"filter":"raw"},{"term_id":841,"name":"environment","slug":"environment","term_group":0,"term_taxonomy_id":841,"taxonomy":"post_tag","description":"","parent":0,"count":25,"filter":"raw"},{"term_id":6511,"name":"watershed","slug":"watershed","term_group":0,"term_taxonomy_id":6511,"taxonomy":"post_tag","description":"","parent":0,"count":1,"filter":"raw"}],"category_data":[{"term_id":3872,"name":"Water Resources","slug":"water-resources","term_group":0,"term_taxonomy_id":3872,"taxonomy":"category","description":"","parent":191,"count":6,"filter":"raw"}],"product_data":{"errors":{"invalid_taxonomy":["Invalid taxonomy."]},"error_data":[]},"primary_product_link":"https:\/\/www.esri.com\/en-us\/industries\/blog\/industry\/water-resources","short_description":"Deep learning identified Arundo in a 56,000-acre study area within the Santa Ana River Watershed, saving hundreds of staff hours. ","image":"https:\/\/www.esri.com\/en-us\/industries\/blog\/app\/uploads\/2026\/01\/SAWPA_Card.jpg","_links":{"self":[{"href":"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-json\/wp\/v2\/article\/581255","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-json\/wp\/v2\/article"}],"about":[{"href":"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-json\/wp\/v2\/types\/blog"}],"author":[{"embeddable":true,"href":"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-json\/wp\/v2\/users\/131"}],"version-history":[{"count":0,"href":"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-json\/wp\/v2\/article\/581255\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-json\/wp\/v2\/media?parent=581255"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-json\/wp\/v2\/categories?post=581255"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.esri.com\/en-us\/industries\/blog\/wp-json\/wp\/v2\/tags?post=581255"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}