**AWARDS ISSUED PRIOR TO JANUARY 20, 2025, WERE FUNDED UNDER PREVIOUS ADMINISTRATIONS AND MAY NOT REFLECT THE PRIORITIES AND POLICIES OF THE CURRENT ADMINISTRATION.** THE OVERARCHING GOAL OF THIS RESEARCH IS TO DEVELOP A DEEP-LEARNING FRAMEWORK FOR THE AUTOMATED SELECTION OF OPTIMAL SOIL SAMPLING SITES BASED ON LANDSCAPE POSITION. SOIL SAMPLING IS ONE OF THE MOST FUNDAMENTAL PROCESSES IN AGRICULTURE: IT IS THE CRUCIAL FIRST STEP IN SOIL TESTING TO DETERMINE SOIL HEALTH. A SOIL ANALYSIS, WHICH PROVIDES INFORMATION IMPORTANT TO MAXIMIZE NUTRIENT USE EFFICIENCY AND AGRICULTURAL PRODUCTIVITY, CAN ONLY BE AS GOOD AS THE SAMPLES SENT TO THE LAB. GOOD SAMPLES REQUIRE SAMPLING AT MULTIPLE OPTIMAL SITES IN THE FIELD. IN THE CURRENT PRACTICE, FARMERS COLLECT AND POOL SAMPLES THEMSELVES AND SEND THEM TO A LAB FOR ANALYSIS. THERE ARE SCIENTIFIC METHODS WHERE SAMPLES CAN BE POOLED BASED ON LANDSCAPE POSITION AND KNOWLEDGE ABOUT HOW NUTRIENTS MOVE IN SOIL DUE TO DIFFERENCES IN PROPERTIES. HOWEVER, MOST PRODUCERS DO NOT FOLLOW THOSE PROCEDURES AS THEY CAN BE COMPLEX AND VARY FROM FIELD TO FIELD. POOLED SOIL SAMPLES DO NOT REPRESENT THE ACTUAL VARIATION IN SOIL PROPERTIES. PRODUCERS CAN PULL MULTIPLE SOIL SAMPLES BUT HAVE NO ASSURANCE OF THE EXTENT OF THEIR FIELDS WHICH EACH SAMPLE ACTUALLY REPRESENTS. AS A RESULT, THERE ARE NO RELIABLE METHODS FOR FARMERS TO ACCURATELY UTILIZE KNOWLEDGE OF SOIL VARIATION WITH THEIR PRECISION AGRICULTURE TECHNOLOGIES. THERE IS AN URGENT NEED FOR AN AUTOMATED TOOL THAT WILL HELP PRODUCERS IDENTIFY WHICH SPOTS ARE OPTIMAL FOR SAMPLING AND CAN BE POOLED TO GET ACCURATE SOIL ANALYSIS RESULTS.TO FILL THIS CRITICAL NEED, WE AIM TO DEVELOP A DEEP-LEARNING TOOL THAT OUTPUTS LANDSCAPE ZONES WITH POSITION ELEVATION AND IDENTIFIES OPTIMAL SAMPLING SPOTS FOR EACH ZONE. THE TRAINING LANDSCAPE DATA AND SCIENTIFIC METHODOLOGY WILL BE PROVIDED BY THE SOIL PEDOLOGIST CO-PI. THE ARTIFICIAL INTELLIGENCE (AI) -ENABLED TOOL WILL BE DEVELOPED WITH GPS GUIDANCE TO GO FROM SPOT TO SPOT. IT WILL ALSO ALLOW FOR THE MIXING OF APPROPRIATE SAMPLES AND PROVIDE INFORMATION ON THE ESTIMATED COST OF ANALYSIS. IF THE PRODUCER CHOOSES A PRICE CAP, THE TOOL COULD INFORM THEM OF THE NUMBER OF SAMPLES THAT COULD BE ANALYZED WITHIN THAT PRICE RANGE AND HOW ACCURATE THE RESULTS WOULD BE.OUR CENTRAL HYPOTHESIS IS THAT THE USE OF ADVANCED DEEP-LEARNING TECHNIQUES TO ANALYZE AND REFINE LANDSCAPE DATA WILL ENABLE PRECISE AND RELIABLE RECOMMENDATIONS OF OPTIMAL SOIL SAMPLING SPOTS. IN ADDITION, THE FRAMEWORK WILL PRODUCE CONSISTENT RESULTS UNDER UNCERTAIN VARIABLE CONDITIONS FROM FIELD TO FIELD. THE RATIONALE IS THAT DEEP LEARNING EXTRACTS MEANINGS FROM LANDSCAPE DATA AND HUMAN-LABELED DATA TO TRAIN MULTI-LAYER NEURAL NETWORKS TO INCREASE THE RELIABILITY AND ACCURACY OF SAMPLING LOCATION SELECTION. OUR TEAM IS PARTICULARLY WELL PREPARED TO UNDERTAKE THE PROPOSED RESEARCH BECAUSE OF OUR EXTENSIVE AND SUCCESSFUL TRACK RECORD OF AI-ENABLED AND DATA-DRIVEN RESEARCH IN PRECISION AGRICULTURE AND SOIL-LANDSCAPE ANALYSIS.WE PLAN TO TEST THE CENTRAL HYPOTHESIS BY PURSUING THE FOLLOWING THREE SPECIFIC OBJECTIVES:ESTABLISH A CYBERINFRASTRUCTURE OF,LANDSCAPE DATA AND SOIL SAMPLING ANNOTATIONS TO TRAIN DEEP CONVOLUTIONAL NEURAL NETWORKS.DEVELOP A DEEP-LEARNING PIPELINE TO LEARN, ANALYZE, AND REFINE LANDSCAPE DATA FOR AUTOMATED SELECTION OF SOIL SAMPLING LOCATIONS. THE FRAMEWORK TAKES LANDSCAPE DATA AS INPUT AND OUTPUTS OPTIMAL SOIL SAMPLING SPOTS.DESIGN AND IMPLEMENT A SET OF METRICS TO ASSESS THE SUCCESS RATE OF THE OPTIMAL SAMPLING SITE PREDICTION TOOL.THE PROPOSED RESEARCH IS ORIGINAL AND TRANSFORMATIVE BECAUSE IT WILL CREATE AN ADVANCED TOOL FOR A CRUCIAL AND CHALLENGING PRECISION AGRICULTURE PROBLEM, NAMELY AUTOMATED AND RELIABLE SELECTION OF SOIL SAMPLING SITES. THIS TOOL IS CURRENTLY MISSING, AND IT WILL ENABLE A SIGNIFICANT IMPROVEMENT IN SOIL SAMPLING AND ANALYSIS, WHICH WILL LEAD TO A BETTER UNDERSTANDING OF SOIL HEALTH. IT WILL LAY A FOUNDATION FOR NOVEL APPLICATIONS OF DATA SCIENCE AND AI TECHNOLOGIES TO SOLVE AGRICULTURAL PROBLEMS.