donsutherland1

Tomorrow will be blustery and cool. Temperatures will likely top out in the upper 40s. It will then turn much warmer on Thursday. The mercury could soar to 70° or above in New York City. The temperature could make a run a 80° from Washington, DC to Philadelphia. Friday will remain warm, but showers and perhaps a thundershower are possible as a cold front moves through the region. It will turn cooler on Saturday and then milder again on Sunday. Showers are possible during the weekend. The ENSO Region 1+2 anomaly was +0.9°C and the Region 3.4 anomaly was -0.1°C for the week centered around March 19. For the past six weeks, the ENSO Region 1+2 anomaly has averaged +1.00°C and the ENSO Region 3.4 anomaly has averaged -0.18°C. Neutral ENSO conditions will likely continue into at least late spring. The SOI was +26.75 today. The preliminary Arctic Oscillation (AO) was +1.227 today.

I know. I have more hope for AI down the road than the GFS as it currently stands.

AI has an enormous amount of learning to do with snowfall forecasts (the GFS's problem is structural) and such forecasts will grow more skillful with the additional learning. I could see Boston having a small amount of measurable snowfall (probably < 1" monthly total), but don't see the kind of pattern that brings anything close to what that map is showing. Historically, there are no snowstorms on record that saw 6" or more in Boston and 4" or more in both New York City and Philadelphia on or after April 10th. The latest such storm (and only on after the first week of April) occurred during April 8-9, 1917. Prior to regular recordkeeping, a storm that met and exceeded such criteria occurred on April 12-13, 1841.

New York City is finishing March with a mean temperature of 46.9° (4.1° above normal). That makes March 2025 the 12th warmest March on record. A cold front will move slowly across the region tonight into tomorrow. It could be accompanied by strong thunderstorms, damaging winds, and heavy rainfall. A general 0.50"-1.50" rainfall with locally higher amounts is possible. In its wake, it will turn much cooler for tomorrow and Wednesday. It will again become unseasonably warm on Thursday. A cooler period will likely develop during the weekend. An AI random forest model currently suggests that April could wind up near or even somewhat below normal in New York City. A random forest model is a machine learning algorithm that combines the output of multiple decision trees to handle non-linear relationships to reach a single result. The ENSO Region 1+2 anomaly was +0.9°C and the Region 3.4 anomaly was -0.1°C for the week centered around March 19. For the past six weeks, the ENSO Region 1+2 anomaly has averaged +1.00°C and the ENSO Region 3.4 anomaly has averaged -0.18°C. Neutral ENSO conditions will likely continue into at least late spring. The SOI was +0.19 today. The preliminary Arctic Oscillation (AO) was +1.174 today.

Consensus often does better than most forecasters on a consistent basis. It will be interesting to compare the consensus with the AI forecast for Boston, New York City, and Phoenix once the picks are in. The AI forecast anomalies were: Boston: +0.8, New York City: 0.0, and Phoenix: +0.9. In probabilistic terms, AI was broadly in line with today's CPC idea of "equal chances" for Boston and New York City and "above normal" for Phoenix. In terms of rainfall, "equal chances" applied to NYC and Phoenix, but CPC had below normal precipitation for Phoenix. Given the wet anomalies shown for Arkansas, I generated the AI forecast for Little Rock's April precipitation (6.96" vs. 5.59" normal value): Let's see how AI does finishes. Unfortunately, I didn't run the numbers for all of the cities, as I wasn't thinking of the contest at the time. I did look at a range of variable for New York City, as well as some variables for additional cities. Precipitation is tough to forecast. Maybe next month, I'll ask it for numbers for all of the contest cities to see how AI performs. I personally think AI likely has a long way to go when it comes to consistently skillful long-range forecasts e.g., monthly numbers. Maybe the AI-generated numbers will prove consistently good. We'll see.

DCA _ NYC _ BOS __ ORD _ ATL _ IAH ___ DEN _ PHX _ SEA -0.2 -0.2 0.3 -0.4 1.3 1.3 0.7 1.0 0.4

Before a return to cooler weather, Atlanta could experience a historic early-season heat event this week.

Final outcome: The period was warmer than had been shown on the guidance. The high temperatures well far higher than the values suggested on the guidance.

Temperatures will rebound tomorrow. It will be unseasonably warm with highs in the middle and upper 60s. A cold front will move slowly across the region tomorrow and tomorrow night. It could be accompanied by strong thunderstorms, damaging winds, and heavy rainfall. A general 0.50"-1.50" rainfall with locally higher amounts is possible from late tomorrow afternoon through Tuesday morning. The temperature roller coaster will again dip for Tuesday and Wednesday before unseasonable warmth returns on Thursday. An AI random forest model currently suggests that April could wind up near or even somewhat below normal in New York City. A random forest model is a machine learning algorithm that combines the output of multiple decision trees to handle non-linear relationships to reach a single result. The ENSO Region 1+2 anomaly was +0.9°C and the Region 3.4 anomaly was -0.1°C for the week centered around March 19. For the past six weeks, the ENSO Region 1+2 anomaly has averaged +1.00°C and the ENSO Region 3.4 anomaly has averaged -0.18°C. Neutral ENSO conditions will likely continue into at least late spring. The SOI was +2.54 today. The preliminary Arctic Oscillation (AO) was +1.260 today. Based on sensitivity analysis applied to the latest guidance, there is an implied near 100% probability that New York City will have a warmer than normal March (1991-2020 normal). March will likely finish with a mean temperature near 46.9° (4.2° above normal). That would tie March 1979 as the 11th warmest March on record.

Artificial Intelligence (AI) has garnered significant attention this winter and early spring in forecasting. Multiple AI initiatives are ongoing. In some cases, such as forecast 500 mb patterns, AI has shown improvement over the best-of-class operational ECMWF model. One area that has not garnered as much attention is longer-range forecasting. Can AI make advances that push the frontiers of long-range forecasting? Chaos and non-linearity become more important as one expands the forecasting timeframe. To take a closer look, I constructed some AI models to forecast the coming April. The AI-generated standard error and coefficient of determination hint at a large improvement over linear forecasting techniques. I say "hint," because I remain cautious about a hindcasting problem where things look good against past data only to break down in an actual forecasting situation. Consider, for example, the initial promise and more limited impact of using the Siberian Snow Advance Index (SAI) for winter forecasting. I will share the results here for New York City (Central Park). April 2025 AI Forecast: Monthly mean temperature: 53.7° Monthly maximum temperature: 84° Monthly minimum temperature: 35° Monthly precipitation: 4.32" Those are equivalent to ensemble means. Some additional information: Monthly mean temperature: Below 52°: 17% trees or members Below 53°: 41% of members 54° or above: 51% of members 55° or above: 28% of members 56° or above: 4% of members This indicates that although somewhat more members show an above normal temperature, the smaller number of cool members are sufficiently cool to skew the overall average. Monthly precipitation: Below 3": 25% 4" or above: 61% 5" or above: 21% Monthly Snowfall: No measurable snowfall: 86% of members Trace of snow: 13% of members Measurable snow: 1% of members I also inquired about how many members showed an April freeze. 29% had a monthly minimum temperature of 32° or below. 1991-2020 Baselines: Monthly Mean: 53.7° Monthly Maximum: 83° Monthly Minimum: 33° Monthly Precipitation: 4.09" Monthly Snowfall: 0.4" Comparison: CFSv2 Forecasts: Miscellaneous: The AI model suggested that both Phoenix (above normal on the CFSv2) and Boston (normal on the CFSv2) would be above normal with April mean temperatures of 74.1° and 49.4° respectively. The AI model also suggested that Phoenix would have an April monthly maximum of 100° (range of scenarios: 97°-105° with 47% of members at 100° or above). In terms of rainfall, the mean forecast for Phoenix was 0.22" (normal). 33% of members showed no measurable rainfall; 67% had measurable rainfall; 28% had 0.25" or more; 13% had 0.50" or more; and, 8% had 1.00" or more. My expectation: I am agnostic about the AI forecasts. Aside from my own thinking that an April freeze is unlikely and that NYC has very likely seen its last measurable snowfall, I have no strong views about the possible outcomes. I note that this time around, the AI forecasts seem to be broadly in line with the CFSv2 ideas. Outcome: The outcome will be verified at the end of the month.

White Plains has now seen the temperature fall 41F (22.8C) from its daily high. That breaks the March record of 40F (22.2C) set on March 4, 1972.

I revised my thinking to mid 50s for tomorrow.

So far, Central Park has seen its largest one- and two-hour drops on record for March: 26 degrees and 29 degrees respectively.

The temperature surged into the 80s in parts of the region today. New York City saw its first 80° or above March temperature since March 26, 2021 when the thermometer peaked at 82°. March 29th, which accounts for 5 of New York City's 17 80° or above March temperatures, is the most common date for such temperatures in March. High temperatures included: Islip: 82° (old record: 81°, 1998) ***Tied March monthly record*** New York City-Central Park: 81° New York city-JFK Airport: 81° New York City-LaGuardia Airport: 82° Newark: 85° White Plains: 80° So far, the high temperature at Burlington is 31°. If Burlington's high temperature finishes at 32° or below, today would be the first time on record that Burlington had a high of 32° or below when New York City had a high of 80° or above. Currently, the coldest high at Burlington when New York City reached 80° or above is 33°. That occurred on April 7, 1942 when New York City had a maximum temperature of 82°. Tomorrow will be noticeably cooler with highs in the middle to perhaps upper 50s in New York City, but Monday could be another unseasonably warm day with highs in the middle and upper 60s. The temperature roller coaster will again dip for Tuesday and Wednesday before unseasonable warmth returns on Thursday. An AI random forest model currently suggests that April could wind up near or even somewhat below normal in New York City. A random forest model is a machine learning algorithm that combines the output of multiple decision trees to handle non-linear relationships to reach a single result. The ENSO Region 1+2 anomaly was +0.9°C and the Region 3.4 anomaly was -0.1°C for the week centered around March 19. For the past six weeks, the ENSO Region 1+2 anomaly has averaged +1.00°C and the ENSO Region 3.4 anomaly has averaged -0.18°C. Neutral ENSO conditions will likely continue into at least late spring. The SOI was +16.94 yesterday. The preliminary Arctic Oscillation (AO) was +1.694 today. Based on sensitivity analysis applied to the latest guidance, there is an implied near 100% probability that New York City will have a warmer than normal March (1991-2020 normal). March will likely finish with a mean temperature near 47.2° (4.4° above normal). That would tie March 2000 as the 10th warmest March on record.